o


[PDF]o - Rackcdn.comhttps://f5250d95e2766cce1a42-7463c38e8494c97b9c7c48562e9646e9.ssl.cf3.rackcd...

7 downloads 224 Views 2MB Size

II III Illllllllllll lllllllllll Mil I I l l US005632272A

United States Patent [i9]

[ii] Patent Number: [45] Date of Patent:

Diab et al.

4,869,254 4,883,353 4,892,101 4,907,594 4,911,167 4,927,264 4,928,692 4,948,248 4,955,379 4,956,867 5,057,695 5,273,036 5,458,128

[54] SIGNAL PROCESSING APPARATUS [75] Inventors: Mohamed K. Diab; Esmaiel Kiani-Azarbayjany; Ibrahim M. Elfadel, all of Laguna Niguel; Rex J. McCarthy, Mission Viejo; Walter M. Weber, Los Angeles; Robert A. Smith, Corona, all of Calif. [73] Assignee: Masimo Corporation, Mission Viejo, Calif. [21] Appl. No.: 320,154 [22] FUed:

Related U.S. Application Data [63] Continuation-in-part of Ser. No. 132,812, Oct 6, 1993, Pat. No. 5,490,505, and a continuation-in-part of Ser. No. 249, 690, May 26, 1994, Pat. No. 5,482,036, which is a continuation of Ser. No. 666,060, Mar. 7, 1991, abandoned. [51] Int. Cl.6 A61B 5/00 [52] U.S. Cl 128/633; 128/666 [58] Field of Search 128/633-634, 128/664-667, 672, 687-688, 716; 356/39-41 References Cited U.S.

PATENT DOCUMENTS

3,647,299 3/1972 Lavallee . 3,704,706 12/1972 Herczfdd et al. 4,063,551 12/1977 Sweeney . 4,086,915 5/1978 Kofsky et al. . 4,095,117 6/1978 Nagy . 4,407,290 10/1983 Wilber. 4,537,200 8/1985 Widrow . 4,649,505 3/1987 Zinser, Jr. et al. 4,773,422 9/1988 Isaacson et al. . 4,799,493 1/1989 DeFault. 4,800,495 1/1989 Smith. 4,824^42 4/1989 Frick et al. . 4,848,901 7/1989 Hood, Jr. . 4,860,759 8/1989 Kahn et al. . 4,863,265 9/1989 Flower et al. . 4,867,571 9/1989 Frick et al. . 4,869,253 9/1989 Craig, Jr. et al. .

/

Stone et al. . Hausman . Cheung et al. . Muz . Corenman et al. . Shiga et al. . Goodman et al. . Lehman . HaU . Zurek et a!. . Hirao et al. . Kronberg et al. . Polanyi et al

Rabiner, Lawrence et al. Theory and Application of Digital Signal Processing, p. 260, 1975. Tremper, Kevin et al.. Advances in Oxygen Monitoring, pp. 137-153. 1987. Harris, Fred et al., "Digital Signal Processing with Efficient Polyphase Recursive All-Pass Filters", Presented at International Conference on Signal Processing, Florence, Italy, Sep. 4-6, 1991, 6 pages. (List contmued on next page.) Primary Examiner—Angela D. Sykes Attorney, Agent, or Firm—Knobbe, Martens, Olson & Bear, LLP. ABSTRACT

[57]

The present invention involves method and apparatus for analyzing two measured signals that are modeled as containing primary and secondary portions. Coefficients relate the two signals according to a model deflned in accordance with the present invention. In one embodiment, the present invention involves utilizing a transformation which evaluates a plurality of possible signal coefficients in order to find appropriate coefficients. Alternatively, the present invention involves using statistical functions or Fourier transform and windowing techniques to determine the coefficients relating to two measured signals. Use of this invention is described in particular detail with respect to blood oximetry measurements. 23 Claims, 37 Drawing Sheets

SENSOR CAIN CONTROL

•j RED EMITTER]

/302 -llRED EMITTERl

.. 128/633 X

OTHER PUBLICATIONS

Oct. 7,1994

[56]

9/1989 11/1989 1/1990 3/1990 3/1990 5/1990 5/1990 8/1990 9/1990 9/1990 10/1991 12/1993 10/1995

5,632,272 May 27, 1997

FRONT END: ANALOG SIONAL CONDI TIONINC-

EMITTER CURRENT DRIVERS

OXYGEN SATURATION ANALOG TO DIGITAL CONVERSION

DIGITAL TO ANALOG CONVERSION

DIGITAL SIGNAL PROCESSING AND SIGNAL EXTRACTION

EMITTER CURRENT CONTROL

PLEmVSMOGRAPHIC

5,632,272 Page 2

OTHER PUBLICATIONS Haykin, Simon, Adaptive Filter Theory, Prentice Hall, Englewood Cliffs, NJ, 1985. Widrow, Bernard, Adaptive Signal Processing, Prentice Hall, Englewood Cliffs, NJ 1985. Brown, David P., "Evaluation of Pulse Oximeters using Theoretical Models and Experimental Studies", Master's thesis. University of Washington, Nov. 25,1987, pp. 1-142. Cohen, Amon, "Volume I Time and Frequency Domains Analysis", Biomedical Signal Processing, CRC Press, Inc., jBoca Raton, Florida, pp. 152-159. Severinghaus, J.W, "Pulse Oximetry Uses and Limitations", pp. 1-4, ASA Convention, New Orleans, 1989.

Mook, G.A., et al., "Spectrophotometirc determination of Oxygen saturation of blood independent of the presence of indocyanine green", Cardiovascular Research, vol. 13, pp. 233-237, 1979. Neuman, Michael R., "Pulse Oximetry: Physical Principles, Technical Realization and Present Limitations", Continuous Transcutaneous Monitoring, Plenum Press, New York, 1987, pp. 135-144. Mook, G.A., et al., "Wavelength dependency of the spectrophotometric determination of blood oxygen saturation", Clinical Chemistry Acta, vol. 26, pp. 170-173, 1969. Klimasauskas, Casey, "Neural Nets and Noise Filtering", Dr. Dobb's Journal, Jan. 1989, p. 32. Melnikof, S. "Neural Networks for Signal Processing: A Case Study", Dr. Dobbs Journal, Jan. 1989. pp. 36-37.

U.S.

Patent

May 27, 1997

Sheet 1 of 37

5,632,272

FIG.I

-^t

BONE MUSCLE TISSUE ARTERIAL BLOOD VENOUS BLOOD

FIG.2

I

FIG.3 S=s+n

U.S. Patent

May 27, 1997

5,632,272

Sheet 2 of 37

r 24b _£_

22b

20

7

S

s

XF Ab

hn

JZSIGNAL CONDITIONER

Xb

22a

r-

Aa

Aa

Aa

A/D 24a

SIGNAL CONDITIONER

n'(t)= nAa(t) - ranXb(t)

i

n

CORRELATION CANCELER

FIG. 4a

. z i 28

27

DISPLAY s

Xa(0

U.S. Patent

May 27, 1997

5,632,272

Sheet 3 of 37

r

20

J

S

s

hn

_£j

Aa

Aa

Jl.

SIGNAL CONDITIONER

Al5" Xb Ab

Aa

22b

22a

SIGNAL CONDITIONER

JL

24b

A/D 24a

A/D

S

'W= sXa(t)-rvsXb(t)

i

JZ.27

CORRELATION CANCELER

FIG. 4b

JL 28 DISPLAY

U.S. Patent

May 27, 1997

5,632,272

Sheet 4 of 37

30-

hP^xP**^

^0

OR S

X^= s x£ ) + n Xb ( t )

- sx a (t) OR s y t )

bXa (t)h 0 R

JC?

b

^

Xb ( t )

L n'(t)-

FIG. 5a

Z

JO^-X^'^Xa'1' n ^ t ) QR n-xb(t)

OR

syi>=.x£>+nAb(i>

^

cxa ( l ) OR JJ? ^ t f 1 '

1 s(t)-

7

f/G. 56

U.S. Patent

May 27, 1997

FREQUENCY ( f )

Sheet 5 of 37

5,632,272

FIG.5C

U.S. Patent

May 27, 1997

5,632,272

Sheet 6 of 37

M

Io V

L

42

Ai Ai Ai Ai Ai Ai Ai A2A2 A2 A2 A2 A2A2 An An An An An An An M

FIG. 6a

Io

V

V

r<*

Ai Ai Ai Ai Ai Ai Ai A2A2 A2 A2 A2 A2A2 A3 A3 A3 A3 A3 A3 As k

A

A

A

A

A

A

As A,. Ae Ae Ae Ai As A 5 A 6 As Ae As An An An An An An An

FIG. 6b

M

Io V

V 42

A M , A, Ai A, AiAi A2A2 A2 A2 A2 A2A2 A3 A3, A3 A4 A4 A3A4 A A3 A4 A4 A4 . A6 Ae AeAe A 5 A:5 As A 5 ;^6 AsJ As ;

/

(An An An An An An An)

FIG. 6C

M

V

M

May 27, 1997

Sheet 7 of 37

5,632,272 *s

s V see IDO

3h

^no

EXTR DETE

U.S. Patent

!^ V

N ^

Qi

o 1— <

ct:

o LJ

1—

z

U.S.

Patent

May 27, 1997

5,632,272

Sheet 8 of 37

i o

1-^

•43

I?

•r *?

l/l I— Z UJ

o £>

o o <

—u

\-tf

c1 —u

RELATIVE CORRELATION CANCELER ENERGY OUTPUT

z o

U.S. Patent

May 27, 1997

5,632,272

Sheet 9 of 37

i o

tf

tf

K

^

m

tf

o Li_ LL.

tf

UJ

O

CJ

tf

tf

tf

RELATIVE CORRELATION CANCELER ENERGY OUTPUT

(J If)

UJS. Patent

May 27, 1997

LEAST SQUARES LATTICE PREDICTOR

REGRESSION FILTER

A

r

ED

K

S

§

5,632,272

Sheet 10 of 37

§

;iE-

(0

\

O

I

5^ VJ

co

s'

^Sy

N Ss

f^ •*> V

X

&)0 QC <

,x+'

I

/

IN

y

w

U.S. Patent

May 27, 1997

QRD-LSL PREDICATOR

y

Sheet 11 of 37 QRD-LSL FILTER

5,632,272

U.S. Patent

May 27, 1997

Sheet 12 of 37

5,632,272

?20 INITIALIZE NOISE CANCELLER

JZ. UO

INPUT NEW SAMPLES [ S X a ( t ) AND S X b ( t ) ]

TIME UPDATE OF [Z - 1 ] ELEMENTS

I

JZ. /40

CALCULATE REFERENCE [ n ' (t) or s' (t)] FOR TWO MEASURED SIGNAL SAMPLES

JZ /50 ZERO-STAGE UPDATE

JZ ?60 m = 0

JZ ?70 ORDER UPDATE mth-STAGE OF LSL-PREDICTOR

JZ 180 ORDER UPDATE mth-STAGE OF REGRESSION FILTER(S) m = m + 1

200 CALCULATE OUTPUT

JZ 2/0 TO DISPLAY

FIG.9

U.S. Patent

May 27, 1997

Sheet 13 of 37

5,632,272

JZ f20a INITIALIZE NOISE CANCELLER INPUT NEW SAMPLES fJOa

t S X a O ) AND S X b ( t ) ]

TIME UPDATE OF [Z~1] ELEMENTS

JZ f40a CALCULATE REFERENCE [ n ' (t) or s' (t)] FOR TWO MEASURED SIGNAL SAMPLES

JZ /50a ZERO-STAGE UPDATE

JJ /60a m = 0

t

r- *M a

ORDER UPDATE m th -STAGE OF QRD-LSL-PREDICTOR

JZ /80a ORDER UPDATE m th -STAGE OF QRD-LSL FILTER(S) m = m + 1

200a CALCULATE OUTPUT

I TO DISPLAY

r-2Wa

FIG. 9a

U.S. Patent

May 27, 1997

LEAST SQURES LATTICE PREDICTOR

Sheet 14 of 37 REGRESSION FILTER

5,632,272 REGRESSION FILTER

U.S. Patent

May 27, 1997 QRD-LSL PREDICATOR

Sheet 15 of 37

5,632,272

FILTER 1 FILTER 2

k

cn

tn

U.S. Patent

May 27, 1997

Sheet 16 of 37

5,632,272

>< _J s

CL

^

/ ^ h ^

$ ^

2

O

o

1-

X CL

<

<

t-

V

2

z

12

O

o

CK X IU 1— UJ ZJ CL I

y > <

<

>- < X o

8

U.

s

o 2 2 WO K

\

?

0- X Ld —I < _l 2 <

12 UJ

[

cc cc

0 2

^ >

3

v50

V

o

1

j "

_1 CCO LU^

t z so

Q

LU O

ANALOG TO DIGITAL CONVERSION

2 O

o

(

2

<

V 1

^ ^

o z .5

Q 2 0

^.

^

V\

h

—1

a:

^

DIGITAL TO ANALOG CONVERSION

i

o

FRONT E ANALO SIGNAI CONDITIO!

\

DC O O

<

t— 2 UJ

cc cc tn 3 ce OLU > ccg j—

UJ bJ

(— !

!• o KI

wili

§

v

^

91 1

UJ Q

^ ^

1

?

)y — j

K y-

^

*N,

i^

\

\

!

\ \

—1—



Q: UJ

CK LU 1=

UJ

\ !i

i \

Q UJ Qf

T

i— t-

\

LU

Q UJ CC

L -d



^ ^

U.S. Patent

May 27, 1997

Sheet 17 of 37

5,632,272

^

U.S. Patent

May 27,1997 <

Sheet 18 of 37

_

i—

Q_

^ ^ Q co K. .

Kl

5,632,272

m Q

O

oi—

xo ^ 1o^^ _ ^ A__

s

s

\

•N

I

K!

^

or

UJ

1 UJ cc

Li_

_i Q_ CL

<

go O

C/)

U.S. Patent

May 27, 1997

5,632,272

Sheet 19 of 37 i— 2 Ul DC CC 3 O

1 O ir i— _l

S •n

DC O UJC*

z or •> 2


o

t5 ^

Ul CO

U.S. Patent

May 27, 1997

Sheet 20 of 37

^ ^ >

*

^

5,632,272

U.S. Patent

May 27, 1997

5,632,272

Sheet 21 of 37

ft

CD <

CO Ul _l

CL < CO

UJ CD


<

^

+

Q UJ OC

O CO

UJ CD

Q Ul

CO

co ^ ^ V

CO

)

3 Q O 2 UJ DQ

<

X 3

+

Q Ul CC

o

ui _i Ul W)

Q Ul

N

<

5

O < y2 D <

U.S. Patent

May 27, 1997

5,632,272

Sheet 22 of 37

O

o

V) N CL X

X CO

2(/)
>

Q d U l CD

LU CC

<

Ii

or u. ^i

i

DC ^ U l CO

U_ Ul Li

1

CT'-^ U l 00 LL. UJ U. —1 3 CL CD S

3 0 CD 2

<

•(/) 3 O K

0\-

<

^ CO 3 CL O 1- K

3 in

o^ 1

3 m

o--' .

z o

p o < ^ ^>-

O CD Ul Q

^ ^ Y

^

2 O

PO < ' ^ ^

y

O CD Ul Q

^

i

5!

CC UJ h-^-. - 1 CO [ 1 LLJ _l CO a . 0 0 ^

< < Q . CO

5 ^ o 2 -JLO ce^'

CC Ul 1- ^—N - 1 CO

< < 1 0. C/J ^

C7) _J in DC^ L-

o

LL.

Ul

< ^

Ul Z <

N X

x m O
1

U. Ul _l tn 0 . CO ^

X O

N X

m Q Csl Ul CD

Cf—

<

cc

U.S. Patent

May 27, 1997

*

V

CO 2

CO

Q Ul

cc

Sheet 23 of 37

5,632,272

1

""*" DC

1 o z

RATI ATIO

c^ ^j

Z

3

3

Io < Ul CO

FT o < cc UJ < CO CO

co m Q

ui

N X

DC < in DC Csl

CD

!<

<

z

N X

CO UJ

Q

Cs|

Ul

CD

0!

^

U.S. Patent

May 27, 1997

Sheet 24 of 37

5,632,272

z o

SJ ^

U.S. Patent

May 27, 1997

o 1z <

1I



UJ cc

O 3 >- hX <

§

Q -,_

z E

<5

O CO

>

y

j

t

^ ^

2

5,632,272

Sheet 25 of 37

•o

o

o

oxa

UJ 1— >



^1

Ul h" L^

H ^ o-J o i—

o

CO ^ LL-

ui x

co

,

t o z

Ul

>-

o

=N 5§

o o

z

§

CD i

,

1

! «

CD UJ

K

y =>

599 / /

x UJ i-

HIGI 'NFID TES

^1 ^

P o < o

tt CO Ul >

DC

3 O CC

C/) CJ 1-

c o 1—

Ul

<

5 o

co

z

CD

CL

CD

z

^

^

U.S. Patent

May 27, 1997

Sheet 26 of 37

5,632,272

s ^

UJ -r <

N

5 COwCD

U.S. Patent

May 27, 1997

Sheet 27 of 37

5,632,272

U.S. Patent

May 27, 1997

Sheet 28 of 37

5,632,272

S ^

^

U.S. Patent

May 27, 1997

Sheet 29 of 37

OUTPUT ENERGY

5,632,272

U.S. Patent

May 27, 1997

5,632,272

Sheet 30 of 37 CO

z o cr co o

oS

<0

3 ^

U.S. Patent

May 27, 1997

5,632,272

Sheet 31 of 37

m o

xxxxxx X

xx_ CO MJ < >

X X X X

xxxx

<

X < CO

NUMBER OF OCCURENCES

U.S. Patent

May 27, 1997

5,632,272

Sheet 32 of 37 CL <

O ^

n

i

'P

CO

<

O >X

o

LC _J 1— <. CO

T -z. < 1—

UJ h i _J _J

o

QL

UJ CO _i 3 CL

<

cr

x

§

s

U.S. Patent

May 27, 1997

5,632,272

Sheet 33 of 37

o UJ H -

(/) _J < _) r> n a. o

K

i^

'—^^

/

\ UJ _1 UJ CO

M

1 \ \

UJ tfl

O

< X

OL

^ ir;

§^

" \

X UJ

X UJ

it

it

1—

o

it o

S U-

o o

o

\

1

s

UJ UJ tfl

s

T
Q

\-

V

2

<

L. rs lis

Q


fcl _ . o

<

w y

ig

a!

LJ

^ o

to UJ

<

O I

^\

&

UJ

UJ

tn u. < U X Q CL

cc

*

Cl

Q UJ CC

o: J : _! u.

Q

cc < cc

UJ O C/>I

U2

X UJ CL CC X t-

< (/)

1

^0

Q UJ

_J <

_J

/

< ^ O

UJ

( CC

r ^

UJ CC

tn

'

2

UJ O CC _ 1

^

i

c-i

•-

S tn

<

Ci CJ

APSH z, 56

D SN z, 56

9?

-jLUCC

IJ-lfc (/) cc => |—

W UJ i CL

UJ

s<

s \ \

cn $ s y) tn

2 -1 O

^

< < (/) /-

/ c\i

/

y

Z

RIAL RATIC

^

(

<

< l CC< 3 3 l-O < LU

^

> ^>

Q UJ

I—

FO

68.

2

0 2

h

< m

UJ

in

cc <

^cj

gci

<^

SS

^3

^ ^

^N

1

T 2

1

\ Q _l O X

1—

v

\

0 : 1 - 1

<0

X

£

Q;

<

cc

X

•-

a \

Q 3 h2 O

UJ

cc

Q UJ

/

i

Q _l

LJ

ri l

I

Li_

I

LU

2

iQ

/ ^

X CL

»

LU

\

8| /

UJ tfl

r

1" <0

^ 1 \ ^

U.S. Patent

May 27, 1997

Sheet 34 of 37

x CL <

o o m >-

X Ld _l CL Z

\—

LJ h<

cr u Ul _i

<

UJ _J

3

OL

o >

^

0.2

Z^3 1

V) •^

^

^

U.

o

S^ 9o iz ^3 b_

§

2

o

f— <

cr 3 I— < (f)

1— Li_

2 O

cru

_l < y

cr

UJ H-

cr <

1

VERSE INDOW NOTION

^

z0

J

5,632,272

U.S. Patent

May 27, 1997

Sheet 35 of 37

*?

5,632,272

AaW = SAred(t)

FIG.26

S

Ab(t) = S AIRM

•~t

FIG.27

•-t

FIG.28

U.S.

Patent

May 27, 1997

Sheet 36 of 37

5,632,272

s

Xa( t ) = s Ared(t)

FIG. 29

\

FIG. 30

/' s Xb(t) = s XlR(t)

U.S. Patent

May 27, 1997

to

Q

<

Ui

OJ.E TJ OC

O

±.2 O l_ 0 ) •*-

— QJc UJ O -m,-

^

o u. -"llHl

oC

o o UJ O

5,632,272

Sheet 37 of 37

CO

to

5,632,272 1 SIGNAL PROCESSING APPARATUS

2

corresponding primary or secondary signal portions, they have a frequency spectrum which is similar to that of the Reference to Prior Related AppUcation primary or secondary signal portions. This is a continuation-in-part application of U.S. patent In many cases, nothing or very little is known about the application Ser. No. 08/132,812 filed Oct. 6, 1993, and 5 secondary and/or primary signal portions. One area where entitled "Signal Processmg Apparatus" now U.S. Pat. No. measured signals comprising a primaiy signal portion and a 5,490,505 and a continuation-in-part apphcation of U.S. secondary signal portion about which no infonnation can patent apphcation Ser. No. 08/249,690 filed May 26, 1994 ^ ^ b e determined is physiological monitoring. Physientitled "Signal Processing Apparatus and Method", now U. °loZ1C3} monitoring generally involves measured signals S. Pat. No. 5.482,036 which is a continuation of U.S. patent io denved from a physiological system such as the human apphcation Ser. No. 07/666,060 filed Mar. 7, 1991, now ^ Measurements which are typically taken with physit A A ological momtonng systems include electrocardiographs, blood pressure, blood gas saturation (such as oxygen BACKGROUND OF THE INVENTION saturation), capnographs, other blood constituent monitoring, heart rate, respiration rate, electro1. Field of the Invention encephalograph (EEG) and depth of anesthesia, for example. The present invention relates to the field of signal proOther types of measurements include those which measure cessing. More specifically, the present invention relates to the pressure and quantity of a substance within the body the processing of measured signals, containing a primary such as cardiac output, venous oxygen saturation, arterial signal portion and a secondary signal portion, for the oxygen saturation, bilirubin, total hemoglobin, breathalyzer removal or derivation of either the primary or secondary 20 testing, drug testing, cholesterol testing, glucose testing, signal portion when little is known about either of these extra vasation, and carbon dioxide testing, protem testing, components. More particularly, the present invention relates carbon monoxide testing, and other in-vivo measurements, to modeUng the measured signals in a novel way which for example. CompUcations arising in these measurements fadMtates minimizing the conelation between the primary are often due to motion of the patient, both extemai and signal portion and the secondary signal portion in order to 2 internal (muscle movement, vessel movement, and probe produce a primary and/or secondary signal. The present movement, for example), during the measurement process, invention is especiaUy useful for physiological monitoring M a n y types 0 f physiological measurements can be made systems including blood oxygen saturation systems. by using the known properties of energy attenuation as a 2. Description of the Related Art 30 selected form of energy passes through a medium. Signal processors are typicaUy employed to remove or A blood gas monitor is one example of a physiological derive either the primaiy or secondary signal portion from a monitoring system which is based upon the measurement of composite measured signal including a primary signal porenergy attenuated by biological tissues or substances. Blood tion and a secondary signal portion. For example, a comgas monitors transmit Ught into the test medium and meaposite signal may contain noise and desirable portions. If the 35 sure the attenuation of the Ught as a function of time. The secondary signal portion occupies a different frequency output signal of a blood gas monitor which is sensitive to the spectrum than the primary signal portion, then conventional arterial blood flow contains a component which is a wavefiltering techniques such as low pass, band pass, and high form representative of the patient's arterial pulse. This type pass filtering are available to remove or derive either the of signal, which contains a component related to the primary or the secondary signal portion from the total signal. 4 0 patient's pulse, is caUed a plethysmographic wave, and is Fixed single or multiple notch filters could also be employed shown in FIG. 1 as curve s. Plethysmographic waveforms if the primary and/or secondary signal portion(s) exist at a are used in blood gas saturation measurements. As the heart fixed frequency(s). beats, the amount of blood in the arteries increases and It is often the case that an overlap in frequency spectrum decreases, causing increases and decreases in energy between the primary and secondary signal portions exists. 45 attenuation, iUustrated by the cydic wave s in FIG. 1. CompUcating matters further, the statistical properties of one Typically, a digit such as a finger, an ear lobe, or other or both of the primary and secondary signal portions change portion of the body where blood flows close to the skin, is with time. In such cases, conventional filtering techniques employed as the medium through which Ught energy is are ineffective in extracting either the primary or secondary transmitted for blood gas attenuation measurements. The signal. If, however, a description of either the primaiy or 50 finger comprises skin, fat, bone, muscle, etc., shown schesecondary signal portion can be derived, conelation matically in FIG. 2, each of which attenuates energy incident canceling, such as adaptive noise canceling, can be on the finger in a generaUy predictable and constant manner, employed to remove either the primary or secondary signal However, when fleshy portions of the finger are compressed portion of the signal isolating the other portion. In other enatically, for example by motion of the finger, energy words, given sufScient information about one of the signal 55 attenuation becomes enatic. portions.that signal portion can be extracted. An example of a more reaUstic measured waveform S is Conventional conelation cancelers, such as adaptive shown in FIG. 3, illustrating the effect of motion. The noise cancelers, dynamicaUy change their transfer function primary plethysmographic waveform portion of the signal s to adapt to and remove portions of a composite signal. is the waveform representative of the pulse, conesponding However, conelation cancelers require either a secondary 60 to the sawtooth-like pattem wave in FIG. 1. The large, reference or a primary reference which conelates to either secondary motion-induced excursions in signal ampUtude the secondary signal portion only or the primary signal obscure the primary plethysmographic signal s. Even smaU portion only. For instance, for a measured signal containing variations in ampUtude make it difficult to distinguish the noise and desirable signal, the noise can be removed with a primary signal component s in the presence of a secondary conelation canceler if a noise reference is avaUable. This is 65 signal component n. often the case. Although the ampUtude of the reference A pulse oximeter is a type of blood gas monitor which signals are not necessarily the same as the ampUtude of the non-invasively measures the arterial saturation of oxygen in

5,632,272 3

4

the blood. The pumping of the heart forces freshly oxygentogether with either the first or second measured signals for ated blood into the arteries causing greater energy attenuacomputation of, respectively, either the first or second prition. As weU understood in the art, the arterial saturation of mary signal portions. oxygenated blood may be determined from the depth of the Physiological monitors can benefit from signal processors vaUeys relative to the peaks of two plethysmographic wave- 5 of the present invention. Often in physiological measureforms measured at separate wavelengths. Patient movement ments a first signal comprising a first primary portion and a introduces motion artifacts to the composite signal as Ulus- first secondary portion and a second signal comprising a secon trated in the plethysmographic waveform iUustrated in FIG. d primary portion and a second secondary portion are 3. These motion artifacts distort the measured signal. acquired. The signals may be acquired by propagating 10 energy through a patient's body (or a material which is SUMMARY OF THE INVENTION derived from the body, such as breath, blood, or tissue, for example) or inside a vessel and measuring an attenuated This invention provides improvements upon the methods s fg n a l aftej- transmission or reflection. Alternatively, the and apparatus disclosed in U.S. patent appUcation Ser. No. s jg n a i m a y be acquired by measuring energy generated by a 08/132,812. filed Oct. 6. 1993, entitled Signal Processing i s patient's body, such as in electrocardiography. The signals Apparatus, which earUer appUcation has been assigned to are processed via the signal processor of the present inventhe assignee of the instant appUcation. The present invention t i o l l t 0 acquire either a secondary reference or a primary involves several different embodiments using the novel reference which is input to a conelation canceler, such as an signal model in accordance with the present invention to adaptive noise canceler. isolate either a primary signal portion or a secondaiy signal 0 n e p h y s i o i 0 g i c a l monitoring apparatus which benefits portion of a composite measured signal. In one embodiment, from ^ ^ Mention is a monitoring system which a signal processor acquires a first measured signal and a determines a signal which is representative of the arterial second measured signal that is conelated to the first meap u l s e > c a l l e d a plethysmographic wave. This signal can be sured signal. The first signal comprises a first primary signal US ed in blood pressure calculations, blood constituent portion and a first secondary signal portion. The second 2 5 measurements, etc. A specific example of such a use is in signal compnses a second pnmary signal portion and a p u l s e o x i m e t r y . p ^ ^ oximetry involves determining the second secondary signal portion. The signals may be saturation of oxygen in the blood. In this configuration, the acquired by propagating energy through a medium and ^ ^ o f ±e s i g n a l i s ^ arterial blood contribupvhaary measuring an attenuated signal after transmission or reflect i o n t 0 attenU ation of energy as it passes through a portion tion. Alternatively, the signals may be acquired by measur- 3o o f ^ b o d y w h e r e b l o o d flows c l o s e t 0 ±e. s k i n T h e ing energy generated by the medium. pumping of the heart causes blood flow to increase and In one embodiment, the first and second measured signals decrease in the arteries in a periodic fashion, causing periare processed to generate a secondary reference which does odic attenuation wherein the periodic waveform is the not contain the primary signal portions from either of the plethysmographic waveform representative of the arterial first or second measured signals. This secondary reference is 35 pulse. The secondaiy portion is noise. In accordance with the conelated to the secondary signal portion of each of the first present invention, the measured signals are modeled such and second measured signals. The secondary reference is that this secondary portion of the signal is related to the used to remove the secondary portion of each of the first and venous blood contribution to attenuation of energy as it second measured signals via a conelation canceler, such as passes through the body. The secondary portion also an adaptive noise canceler. The conelation canceler is a 4 0 includes artifacts due to patient movement which causes the device which takes a first and second input and removes venous blood to flow in an unpredictable manner, causing from the first input aU signal components which are correunpredictable attenuation and conupting the otherwise perflated to the second input. Any unit which performs or nearly odic plethysmographic waveform. Respiration also causes performs this function is herein considered to be a conelathe secondary or noise portion to vary, although typicaUy at tion canceler. 4 5 a lower frequency than the patients pulse rate. Accordingly, An adaptive conelation canceler can be described by the measured signal which forms a plethysmographic waveanalogy to a dynamic multiple notch filter which dynamiform is modeled in accordance with the present invention caUy changes its transfer function in response to a reference such that the primary portion of the signal is representative signal and the measured signals to remove frequencies from of arterial blood contribution to attenuation and the secondthe measured signals that are also present in the reference 50 ary portion is due to several other parameters, signal. Thus, a typical adaptive conelation canceler receives A physiological monitor particularly adapted to pulse the signal from which it is desired to remove a component oximetry oxygen saturation measurement comprises two and receives a reference signal of the undesired portion. The Ught emitting diodes (LED's) which emit Ught at different output of the conelation canceler is a good approximation to wavelengths to produce first and second signals. A detector the desired signal with the undesired component removed. 55 registers the attenuation of the two different energy signals Alternatively, the first and second measured signals may after each passes through an absorptive media, for example be processed to generate a primary reference which does not a digit such as a finger, or an earlobe. The attenuated signals contain the secondary signal portions from either of the first generaUy comprise both primary (arterial attenuator) and or second measured signals. The primaiy reference may then secondary (noise) signal portions. A static filtering system, be used to remove the primary portion of each of the first and 60 such as a bandpassfilter,removes a portion of the secondary second measured signals via a conelation canceler. The signal which is outside of a known bandwidth of interest output of the conelation canceler is a good approximation to leaving an enatic or random secondary signal portion, often the secondary signal with the primary signal removed and caused by motion and often difficult to remove, along with may be used for subsequent processing in the same instruthe primary signal portion. ment or an auxUiary instrument. In this capadty, the 65 A processor in accordance with one embodiment of the approximation to the secondary signal may be used as a present invention removes the primary signal portions from reference signal for input to a second conelation canceler the measured signals yielding a secondary reference which

5,632,272 5

6

is a combination of the remaining secondary signal portions. plethysmograph, the method further comprises the step of The secondary reference is conelated to both of the secondcalculating the pulse rate. ary signal portions. The secondary reference and at least one Another aspect of the present invention involves a physiof the measured signals are input to a conelation canceler, ological monitor. The monitor has a first input configured to such as an adaptive noise canceler, which removes the 5 receive a first measured signal Si having a primary portion, random or enatic portion of the secondary signal. This Sj, and a secondary portion n ^ The monitor also has a yields a good approximation to a primary plethysmographic second input configured to receive a second measured signal signal as measured at one of the measured signal waveS 2 having a primary portion s 2 and a secondary portion E 2 . lengths. As is known in the art, quantitative measurements of Advantageously, the first and the second measured signals the amount of oxygenated arterial blood in the body can be io Si and 83 are in accordance with the foUowing relationship: detenmned from the plethysmographic signal in a variety of ways. Si=wi+ni The processor of the present invention may also remove 53=12+% the secondary signal portions from the measured signals yielding a primary reference which is a combination of the 15 w h e r e s i ^ d s 2 , and n, and n 2 are related by: s =r s a n remaining primary signal portions. The primary reference is i a 2 d ttr^t^ conelated to both of the primary signal portions. The and where ia and iv are coefficients, primary reference and at least one of the measured signals The monitor further has a scan reference processor, the are input to a correlation canceler which removes the priscan reference processor responds to a pluraUty of possible mary portions of the measured signals. This yields a good 20 values for ia to multiply the second measured signal by each approximation to the secondary signal at one of the meaof the possible values for ia and for each of the resulting sured signal wavelengths. This signal may be useful for values, to subtract the resulting values from the first mearemoving secondary signals from an auxUiary instrument as sured signal to provide a pluraUty of output signals. A weU as determining venous blood oxygen saturation. conelation canceler having a first input configured to receive 25 t h e In accordance with the signal model of the present *** measured signal, and having a second input coninvention, the two measured signals each having primary ^ ^ t 0 receive the plurality of output signals from the and secondary signal portions can be related by coefficients. saturation scan reference processor, provides a pluraUty of out ut vectors By relating the two equations with respect to coefficients P conespondmg to the conelation canceUation defined in accordance with the present invention, the coefbetween the pluraUty of output signals and ffie first measured ficients provide information about the arterial oxygen satu- 3 0 s l g n a L ^ integrator having an input configured to receive ration and aboutthe noise (the venous oxygen saturation and ^ P l u r a l l t y o f <»*P* v e ? o r s *om ^ correlation canceler ls other parameters). In accordance with this aspect of the responsive to the pluraUty of output vectors to determine a present invention, the coefficients can be determined by conespondmg power for each output vector An extremum minimizing the conelation between the primary and secdetector is coupled at its mput to the output of the integrator, ondary signal portions as defined in the model. Accordingly, 35 T h e extremum detector is responsive to the conespondmg the signal model of the present invention can be utilized in P o w e r f o r e a c h 0 U t P u t v f t o r t 0 detecl a ^ f e d power, many ways in order to obtain information about the mea^ o n e embodiment, the pluraUty of possible values corres on sured signals as wiU be further apparent in the detaUed P d to a pluraUty of possible values for a selected Wood description of the prefened embodiments. constituent. In one embodiment the. the selected blood ^ , , . . . , ,. . 4 0 constituent is artenal blood oxygen saturation. In another One aspect of the present invention is a method for use in embodiment, the selected blood constituent is venous blood a signal processor for processmg at least two measured saturation. In yet another embodiment, the selected signals Si and S 2 each containing a pnmary signal portion b l o o d c o n s t i t U e n t is carbon monoxide, s and a secondary sigmd portion n, the signals S} and S 2 ^ ^ Another of ^ t ^ involves a h bemg in accordance with the foUowing relationship: ^ has a ^ ^ 45 o l o g i c a l m o n £ o r T h e ^ c o n f l g u r e d to receive a first measured signal Sj having a primary portion, s=s+n 1 1 1 Si, and a secondary portion, %. The monitor also has a s2=*2+n2 second input configured to received a second measured signal S 2 having a primary portion s 2 and a secondary where Sj and s 2 , and ^ and n 2 are related by: 50 portion n 2 . The first and the second measured signals Sj and Si=T a s 2 and n^i^ S 2 are in accordance with the foUowing relationship: and where ia and r,, are coefficients. The method comprises a number of steps. A value of Sr^+n, coefficient ia is determined which minimize conelation s2=^2+n2 between Si and n ^ Then, at least one of the first and second 55 signals is processed usmg the determined value for ia to where s1 and s 2 , and ^ and n 2 are related by: significantly reduce n from at least one of the first or second s1=rnS2 and ^ = ^ 2 measured signal to form a clean signal. and where r a and ry are coefficients. In one embodiment, the clean signal is displayed on a A transform module is responsive to the first and the display. In another embodiment, wherein the first and second 60 second measured signals and responsive to a plurality of signals are physiological signals, the method further compossible values for ia to provide at least one power curve as prises the step of processing the clean signal to determine a an output An extremum calculation module is responsive to physiological parameter from the first or second measured the at least one power curve to select a value for rfl which signals. In one embodiment, the parameter is arterial oxygen minimizes the conelation between s and n, and to calculate saturation. In another embodiment, the parameter is an E C G 65 from the value for ra a conesponding saturation value as an signal. In yet another embodiment, wherein the first portion output. A display module is responsive to the output of of the measured signals is indicative of a heart saturation calculation to display the saturation value.

5,632,272 7

8

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 Ulustrates an ideal plethysmographic waveform.

FIG. 14 Ulustrates additional detaU of the operations P e r m e d ^ the digital signal processing circuity of FIG.

FIG. 2 schematicaUy Ulustrates a typical finger. ™-,,^.„ .... , , ., ,. , , ™~, * -n t. t •. t ,. c ... FIG. 15 Ulustrates additional detaUregardmg the demoduFIG. 3 lUustrates a plethysmographic waveform which 5 l a t i o n m o d u l e of FIG. 14. includes a motion-induced enatic signal partion. ,^_, , . _, T^^. * -n t. t u t-Ac u •1 • 1 FIG. 16 Ulustrates additional detaU regarding the decimaFIG. 4a lUustrates a schematic diagram of a physiological tion module of FIG 14 monitor to compute primary physiological signals. TTT/-' AU-II t-t u _ *• Ac i . - i - i FIG. 17 represents a more detaUed block diagram of the FIG. 4b illustrates a schematic diagram of a physiological .. /.. t t- tA 1 c r^n IA 14 monitor to compute secondary signds. 10 0 P e r a t l 0 I l s o f t h e s t a t l s t l c s m o d u l e o f ^ FIG. 5a Ulustrates an example of an adaptive noise HG 18 Ulustrates a block diagram of the operations of canceler which could be employed in a physiological ^ e embodiment of the saturation transform module of FIG. monitor, to compute primary physiological signals. FIG. Sb Ulustrates an example of an adaptive noise HG. 19 Ulustrates a block diagram of the operation of the canceler which could be employed in a physiological 15 saturation calculation module of FIG. 14. monitor, to compute secondary motion artifact signals. FTG. 20 Ulustrates a block diagram of the operations of the FIG. 5c Ulustrates the transfer function of a multiple notch P u l s e r a t e calculation module of FIG. 14. filter. FIG. 21 Ulustrates a block diagram of the operations of the FIG. 6a Ulustrates a schematic of absorbing material 20 motion artifact suppression module of FIG. 20. comprising N constituents within the absorbing material. FIG. 21a Ulustrates an alternative block diagram for the FIG. 6b Ulustrates another schematic of absorbing mateoperations ofthe motion artifact suppression module of FIG. 2 rial comprising N constituents, including one mixed layer, ®within the absorbing material. FIG. 22 Ulustrates a saturation transform curve in accorFIG. 6c Ulustrates another schematic of absorbing mate- 25 dance with the prindples of the present mvention. rial comprising N constituents, including two mixed layers, FIG. 23 Ulustrates a block diagram of an alternative within the absorbing material. embodiment to the saturation transform in order to obtain a FIG. 7a Ulustrates a schematic diagram of a monitor, to saturation value, compute primary and secondary signals in accordance with FIG. 24 Ulustrates a histogram saturation transform in 30 one aspect of the present invention. accordance with the alternative embodiment of FIG. 23. FIG. 7b Ulustrates the ideal conelation canceler energy or FIGS. 25A. 25B and 25C Ulustrate yet another alternative power output as a function of the signal coefficients r^ r 2 ,. embodiment in order to obtam the saturation. . . r n . In this particular example. r3=ra and r7=rv. n G . 26 Ulustrates a signal measured at a red wavelength FIG. 7c Ulustrates the non-ideal conelation canceler 3 5 Xa=Ared=660 nm for use in a processor of the present energy or power output as a function of the signal coeffiinvention for determining the secondaiy reference n'(t) or cients r 1 ,r 2 ....r n . In this particular example, r3=r a andr 7 =r v . the primary reference s'(t) and for use in a conelation FIG. 8 is a schematic model of a joint process estimator canceler. The measured signal comprises a primary portion comprising a least-squares lattice predictor and a regression s ^ t ) and a secondary portion n ^ t ) . "Iter. FIG. 27 Ulustrates a signal measured at an infrared 40 FIG. 8a is a schematic model of a joint process estimator wavelength ^ = ^ ^ = 9 1 0 nm for use in a processor of the comprising a QRD least-squares lattice (LSL) predictor and present invention for determining the secondary reference a regression filter. n'(t) or the primary reference s'(t) and for use in a conelation FIG. 9 is a flowchart representing a subroutine for implecanceler. The measured signal comprises a primary portion menting in software a joint process estimator as modeled in 45 s ^ t ) and a secondary partion n ^ t ) . FIG. 8. FIG. 28 Ulustrates the secondaiy reference n'(t) deterFIG. 9a is a flowchart representing a subroutine for mined by a processor of the present invention, implementing in software a joint process estunator as modFIG. 29 Ulustrates a good approxUnation s"Xa(t) to the eled in FIG. 8a. primary portion s^(t) of the signal 8^(1) measured at FIG. 10 is a schematic model of a joint process estimator 50 Aa=Ared=660 nm estimated by conelation canceUation with with a least-squares lattice predictor and two regression a secondary reference n'(t). ^tersFIG. 30 Ulustrates a good approximation s ' ^ t ) to the FIG. 10a is a schematic model of a joint process estimator primary portion 5^,(1) of the signal S ^ t ) measured at with a QRD least-squares lattice predictor and two regres>.b=AlR=910 nm estimated by conelation canceUation with 55 sion filters. a secondary reference n'(t). FIG. 11 is an example of a physiological monitor in FIG. 31 depicts a set of 3 concentric electrodes, i.e., a accordance with the teachings of one aspect of the present tripolar electrode sensor, to derive electrocardiography invention. (ECG) signals, denoted as S^ S 2 and Sj, for use with the FIG. l l a Ulustrates an example of a low noise emitter 60 present invention. Each of the ECG signals contains a cunent driver with accompanying digital to analog conprimary portion and a secondary portion, verter. FIG. 12 Ulustrates the front end analog signal conditionDETAILED DESCRIPTION OF THE ing circuitry and the analog to digital conversion circuitry of INVENTION the physiological monitor of FIG. 11. 65 T he present invention involves a system which utUizes FIG. 13 Ulustrates further detaU of the digital signal first and second measured signals that each contain a priprocessing circuitry of FIG. 11. mary signal portion and a secondary signal portion. In other

5,632,272 9 words, given a first and second composite signals S^t^Si (fj+n^t) and S2(t)=s2(t)+n2(t), the system of the present invention can be used to isolate either the primary signal portion s(t) or the secondary signal portion n(t). FoUowing piocessing, the output of the system provides a good approximation n"(t) to the secondary signal portion n(t) or a good approximation s"(t) to the primary signal portion s(t). The system of the present invention is particularly useful where the primary and/or secondary signal portion n(t) may contain one or more of a constant portion, a predictable portion, an enatic portion, a random portion, etc. The primary signal approximation s"(t) or secondary signal approximation n"(t) is derived by removing as many of the secondary signal portions n(t) or primary signal portions s(t) from the composite signal S(t) as possible. The remauung signal forms either the primary signal approxUnation s"(t) or secondary signal approxUnation n"(t), respectively. The constant portion and predictable portion of the secondary signal n(t) are easUy removed with traditional filtering techniques, such as simple subtraction, low pass, band pass, and high pass filtering. The enatic portion is more difficult to remove due to its unpredictable nature. If something is known about the enatic signal, even statistically, it could be removed, at least partiaUy, from the measured signal via traditional filtering techniques. However, often no information is known about the enatic portion of the secondary signal n(t). In this case, traditional filtering techniques are usuaUy insuffident. In order to remove the secondary signal n(t), a signal model in accordance with the present invention is defined as foUows for the first and second measured signals Si and S2:

10

A block diagram of a generic monitor incorporating a signal processor according to the present invention, and a conelation canceler is shown in FIGS. 4a and 4b. Two measured signals, 8^(1) and Sj^/t), are acquired by a 5 detector 20. One skiUed in the art wiU reaUze that for some physiological measurements, more than one detector may be advantageous. Each signal is conditioned by a signal conditioner 22a and 22b. Conditioning includes, but is not limited to, such procedures as filtering the signals to remove 10 constant portions and ampUfying the signals for ease of manipulation. The signals are then converted to digital data by an analog-to-digital converter 24a and 242>. The first measured signal 8^(1) comprises a first primary signal portion, labeled herein 8^(1), and a first secondary signal portion, labeled herein n ^ t ) . The second measured signal 15 8^(1) is at least partiaUy conelated to the first measured signal S^(t) and comprises a second primary signal portion, labeled herein 5^,(1), and a second secondary signal portion, labeled herein n ^ t ) . TypicaUy the first and second secondary signal portions, n ^ t ) and n ^ t ) , are unconelated and/or 20 enatic with respect to the primary signal portions s ^ t ) and 5^(1). The secondary signal portions n ^ t ) and n ^ t ) are often caused by motion of a patient in physiological measurements. The signals 8^(1) and 8^(1) are input to a reference 25 processor 26. The reference processor 26 multipUes the second measured signal 8^,(1) by either a factor 1^=3^(1)/ s^Xt) or a factor r ^ n ^ t y n ^ t ) and then subtracts the second measured signal 8^(1) from the first measured signal 8^(1). The signal coeffident factors xa and r„ are determined 30 to cause either the primary signal portions 8^(1) and 5^(1) or the secondary signal portions n ^ t ) and n^Xt) to cancel, S\ = si+ni respectively, when the two signals 8^(1) and 8^,(1) are S2 = si + K2 subtracted. Thus, the output of the reference processor 26 is . _ _ either a secondary reference signal n'(t)=nXa(t)-rcnXi(t), in si-r^2 ni-rvn2 35 FIG. 4a, which is conelated to both of the secondary signal or ,•„=— andrv=n1/n2 portions n^(t) and n ^ t ) or a primary reference signal 51 sXt^Sj^tHVSjuXt), in FIG. 4b, which is conelated to both of the primary signal portions s ^ t ) and 8^,(1). A reference where s1 and n^ are at least somewhat (preferably signal n'(t) or s'(t) is input, along with one of the measured substantiaUy) unconelated and s2 and n 2 are at least some- 40 signals 8^(1) or 8,^(1), to a conelation canceler 27 which what (preferably substantiaUy) unconelated. The first and uses the reference signal n'(t) or s'(t) to remove either the second measured signals Sj and 83 are related by conelation secondary signal portions n ^ t ) or n ^ t ) or the primaiy coefficients xa and rv as defined above. The use and selection signal portions 8^(1) or s ^ t ) from the measured signal of these coefficients is described in further detaU below. Sj^t) or 85^(1). The output of the conelation canceler 27 is In accordance with one aspect of the present invention, 45 a good primary signal approximation s"(t) or secondary this signal model is used in combination with a conelation signal approximation n"(t). In one embodiment, the approxicanceler, such as an adaptive noise canceler, to remove or mation s"(t) or n"(t) is displayed on a display 28. derive the enatic portion of the measured signals. In one embodiment, an adaptive noise canceler 30, an GeneraUy, a conelation canceler has two signal inputs and example of which is shown in block diagram form in FIG. one output. One of the inputs is either the secondary 50 5a, is employed as the conelation canceler 27, to remove reference nXt) or the primary reference sXt) which are either one of the enatic, secondary signal portions n ^ t ) and conelated, respectively, to the secondary signal portions n(t) n ^ t ) from the first and second signals S ^ t ) and S ^ t ) . The and the primaiy signal portions s(t) present in the composite adaptive noise canceler 30 in FIG. 5a has as one input a signal S(t). The other input is for the composite signal S(t). sample of the secondary reference nXt) which is conelated IdeaUy, the output of the conelation canceler s"(t) or n'Xt) 55 to the secondary signal portions n ^ t ) and n ^ t ) . The conesponds, respectively, to the primary signal s(t) or the secondary reference nXt) is determined from the two measecondary signal n(t) portions only. Often, the most difficult sured signals S ^ t ) and 8^(1) by the processor 26 of the task in the appUcation of conelation cancelers is determinpresent invention as described herein. A second input to the ing the reference signals nXt) and sXt) which are conelated adaptive noise canceler, is a sample of either the first or to the secondary n(t) and primary s(t) portions, respectively, 60 second composite measured signals SXa(t)=sXa(t)+nXa(t) or of the measured signal S(t) since, as discussed above, these SXi(t)=sXi,(t)+n>i>(t). portions are quite difficult to isolate from the measured The adaptive noise canceler 30, in FIG. Sb. may also be signal S(t). In the signal processor of the present invention, employed to remove either one of primaiy signal portions either a secondary reference nXt) or a primary reference sXt) s ^ t ) and s ^ t ) from the first and second measured signals is determined from two composite signals measured 65 8^(1) and 8^,(1). The adaptive noise canceler 30 has as one simultaneously, or nearly simultaneously, at two different input a sample of the primary reference sXt) which is wavelengths, Aa and Ab. conelated to the primary signal portions s ^ t ) and s ^ t ) .

5,632,272 11 The primary reference sXt) is determined from the two measured signals 8^(1) and 8^(1) by the processor 26 of the present invention as described herein. A second input to the adaptive noise canceler 30 is a sample of either the first or second measured signals S ^ t H ^ Q + n ^ t ) or 8^(0=8^ 5 (O+n^t) Th^ adaptive noise canceler 30 functions to remove frequencies common to both the reference nXt) or s'(t) and the measured signal S^Jt) or 8^,(1). Since the reference signals are conelated to either the secondary signal portions io HJUXO and n ^ t ) or the primary signal portions s^(t) and Sxi(t). the reference signals wiU be conespondingly enatic or weU behaved. The adaptive noise canceler 30 acts in a manner which may be analogized to a dynamic multiple notch filter based on the spectral distribution of the reference 15 signal nXt) or s'(t). FIG. 5c Ulustrates an exemplary transfer function of a multiple notchfilter.The notches, or dips in the ampUtude of the transfer function, indicate frequencies which are attenuated or removed when a signal passes through the notch 20 filter. The output of the notch filter is the composite signal having frequencies at which a notch is present removed. In the analogy to an adaptive noise canceler 30. the frequendes at which notches are present change continuously based upon the inputs to the adaptive noise canceler 30. 25 The adaptive noise canceler 30 (FIGS. 5a and Sb) produces an output signal, labeled herein as s",Jt), s ' ^ t ) , n"jjt) oi iTjJt) which is fed back to an internal processor 32 within the adaptive noise canceler 30. The internal processor 32 automaticaUy adjusts its own transfer function 30 accoiding to a predetermined algorithm such that the output of the internal processor 32 labeled bx(t) in FIG. 5a and cx(t) in FIG. Sb. closely resembles either the secondaiy signal portion n ^ t ) oi n ^ t ) oi the primary signal portion 5^(1) or SjuXt). The output bx(t) of the internal processor 32 in FIG. 35 5a is subtracted from the measured signal, 8^(1) or 8^,(1), yielding a signal output sl,;^(t)=sXa(t)+nxa(t)-bXa(t) or a signal output s ' ^ ^ s ^ t ^ n ^ Q - b ^ t ) . The internal processor optimizes s ' ^ t ) or s ' ^ t ) such that s ' ^ t ) or s ' ^ t ) is approximately equal to the primary signal s ^ t ) or 5^(1), 40 respectively. The output cx(t) of the internal processor 32 in FIG. Sb is subtracted from the measured signal. 8^(1) or S^Xt), yielding a signal output given by ^yjlt^s^Jty+n^,, (t)-C;u(t) oi a signal output given by n ' ^ t ^ s ^ X O + n ^ t ) c ^ t ) The internal processor optimizes n ' ^ t ) or n ' ^ t ) 45 such that n"^(t) or n ' ^ t ) is approximately equal to the secondary signal portion n ^ t ) or n ^ t ) , respectively. One algorithm which may be used for the adjustment of the transfer function of the internal processor 32 is a least-squares algorithm, as described in Chapter 6 and 50 Chapter 12 of the book Adaptive Signal Processing by Bernard Widrow and Samuel Steams. pubUshed by Prentice HaU, copyright 1985. This entire book, including Chapters 6 and 12, is hereby incoiporated herein by reference. Adaptive processors 30 in FIGS. Sa and Sb have been 55 successfuUy appUed to a number of problems including antenna sidelobe canceling, pattem recognition, the elimination of periodic interference in general, and the elimination of echoes on long distance telephone transmission Unes. However, considerable ingenuity is often required to find a 60 suitable reference signal nXt) or sXt) since the portions jUt).. n ^ t ) , s ^ t ) and s^(t) cannot easily be separated from the measured composite signals 8^(1) and 8^(1). If either the actual secondary portion n^(t) or n ^ t ) or the primary signal portion 5^(1) or s ^ t ) were a priori avaUable, 65 techniques such as conelation canceUation would not be necessary.

12 GENERALIZED DETERMINAnON OF PRIMARY AND SECONDARY REFERENCE SIGNALS , , , An explanation which descnbes how the reference signals n W a n d s © n^y b e determined foUows. A first signal is measured at, for example, a wavelength Aa. by a detector yielding a signal S^(t): s

^ *° 8^(1) is the primary signal portion and n ^ t ) is the secondary signal portion. A simUar measurement is taken simultaneously, or nearly simultaneously, at a different wavelength, Ab, yielding:

w here

s^ty^^ty+n^it). Note

as lon

as

(2)

16

^ g t* measurements, 8^(1) and 8^(1), are taken substantiaUy simultaneously, the secondary signal components, n^(t) and n ^ t ) , are conelated because any random or enatic functions affect each measurement in nearl y t*16 s a m e fashion. The substantiaUy predictable primar s y ig n a l components. Sjjt) and s ^ t ) , are also conelated to one another, T o obtahl t*16 reference signals nXt) and s'(t). the measured si nals g S^W a n d sw,(t) are transformed to eliminate, respectively, the primaiy oi secondaiy signal components. In accordance with the present invention one way of doing this is t0 {iRd proportionaUty constants. r a and ry, between the primary signal portions Sjjt) and s ^ t ) and the secondaiy si nal g portions n^(t) and n ^ t ) such that the signals can be modeled as foUows: ^W-Vw.w "jJt^rjhM-

(3)

In accordance with the inventive signal model of the present invention, these proportionaUty relationships can be satisfied in many measurements, includmg but not Umited to absoiption measurements and physiological measurements, AdditionaUy, in accordance with the signal model of the present invention, in most measurements, the proportionality constants i a and iv can be deteimined such that: "xaW^yi^M ^W^xtC)-

(4)

Multiplying equation (2) by xa and then subtracting equation (2) from equation (1) results in a single equation wherem the primary signal terms s ^ t ) and s^Xt) cancel: "'('^s^W-^xiC'^n^W-^n^f); (5a) . , , . , . , a nol . ? - z e r o . s l g n a l ^ c \ l s conelated to each secondary s^Portion n^(t) and n ^ t ) and can be used as the ^condaiy reference n (t) in a conelation canceler such as an ^ p t i v e noise canceler. t. M u t t ^ m g equation (2) by r v and then subtracting equaT ( 2 ) ^om eWf"*. (1) r e s u l t s m a single equation ™hei.em ^ s e c o n d a ry signal terms n^(t) and n ^ t ) cancel, eaving. XMJLfy+AM^Uthr*** (5b) a non-zero signal which is conelated to each of the primary signal portions s ^ t ) and s^Xt) and can be used as the signal leference sXt) in a conelation canceler such as an adaptive noise canceler.

5,632,272 13 EXAMPLE OF DETERMINAnON OF PRIMARY AND SECONDARY REFERENCE SIGNALS IN AN ABSORPTIVE SYSTEM Conelation canceUng is particularly useful in a large number of measurements generaUy described as absorption measurements. An example of an absorption type monitor which can advantageously employ conelation canceUng, such as adaptive noise canceUng, based upon a reference n'(t) or s'(t) determined by a processoi of the present invention is one which determines the concentration of an energy absorbing constituent within an absorbing material when the material is subject to change. Such changes can be caused by foices about which infonnation is desired or , . . . , , , .. . pnmary, or alternatively, by random or eiratic secondary forces such as a mechanical force on the material. Random or enatic interference, such as motion, generates secondary components in the measured signal. These secondary components can be removed or derived by the conelation canceler if a suitable secondary reference nXt) or primary reference s (t) is known. A schematic N constituent absorbing material comprising a container 42 having N different absorbing constituents, labeled Aj, A2, A3, . . . AN, is shown in FIG. 6a. The constituents A1 through A^ in FIG. 6a are ananged in a generaUy orderly, layered fashion withm the container 42. An example of a particular type of absorptive system is one in which Ught energy passes through the container 42 and is absorbed according to the generaUzed Beer-Lambert Law of Ught absorption. For Ught of wavelength Aa, this attenuation may be approximated by:

14

subject to forces, each layer of constituents may be affected by the perturbation differently than other layers. Some perturbations of the optical path lengths of each layer x/t) ma y r e s u l t ^ excursions in the measured signal which 5 represent desired or pnmary information. Other perturbations of the optical path length of each layer x^t) cause undesired or secondary excursions which mask primary information in the measured signal. Secondary signal components associated with secondary excursions must also be io removed to obtain primary infonnation from the measured sl naL g SimUarly, the abiUty to compute secondary signal components caused by secondary excursions directly aUows 0 ne . *? obtal11 . P ™ " ^ a f * } components from the measured signal via simple subtraction, or conelation canceUation techn'aues 15 T h e r e l a t i o n canceler may selectively remove from the c o m p 0 s i t e signal, measured after being transmitted through o r reflected from the absorbing material, either the second^ o r ^ g primary signal components caused by forces w hich perturb or change the material differently from the 20 forces which perturbed or changed the material to cause respectively, either the primary or secondary signal component. For the purposes of Ulustration, it wUl be assumed that the portion of the measured signal which is deemed to be the primary signal s ^ t ) is the attenuation term e5C5xs(t) asso25 ciated with a constituent of interest, namely A 5 , and that the layer of constituent A s is affected by perturbations different than each of the layers of other constituents A1 through A4 and A 6 through A ^ An example of such a situation is when layer A 5 is subject to forces about which mfonnation is 30 deemed to be primary and, additionaUy, the entire material is subject to forces which affect each of the layers. In this ,_. / _w . . . A ^ case, since the total force affecting the layer of constituent V !=1 ' " / A s is different than the total forces affecting each of the other InitiaUy transfonning the signal by taking the natural logal a y e r s «•"} ^ f ^ t i o n is deemed to be primary about the rithm of both sides and manipulating t l m s , the signal is 3 5 f ° r c e s and resultant perturbation of the layer of constituent transformed such that the signal components are combined ^ f*™*™ terms due to constituents A, through A 4 and by addition rather than multipUcation, Le.: t6 f f A ^ ^ F " ^ " Z " ^ ^ ^ ^ Even if the additional forces which affect the entire matenal cause the same perturbation in each layer, including the layer N m Sju, = ln(io/r) = x e.-AoC^i 40 of A 5 , the total forces on the layer of constituent A 5 cause it to have different total perturbation than each of the other where IQ is the incident light energy intensity; I is the layers of constituents A1 through A 4 and A 6 through A^. transmitted Ught energy intensity; e.iXa is the absorption It is often the case that the total pertuibation affecting the coefficient ofthe i"1 constituent at the wavelength Aa; xXt) is layeis associated with the secondaiy signal components is the optical path length of i* layer, i.e., the thickness of 45 caused by random or enatic forces. This causes the thickness material of the i* layer through which optical energy passes; of layers to change enaticaUy and the optical path length of and cXt) is the concentration of the i'h constituent in the each layer, x/t), to change enaticaUy, thereby producing a volume associated with the thickness xXt). The absorption random or enatic secondaiy signal component nXa(t). coefficients e.^ through e^ are known values which are However, regardless of whether or not the secondaiy signal constant at each wavelength. Most concentrations c/t) so portion n ^ t ) is enatic, the secondary signal component through c^t) are typicaUy unknown, as are most of the n ^ t ) can be either removed or derived via a conelation optical path lengths x/t) of each layer. The total optical path canceler, such as an adaptive noise canceler, having as one length is the sum of each of the individual optical path input, respectively, a secondary reference n^t) oi a primaiy lengths xXt) of each layei. reference s'(t) determined by a processor of the present When the material is not subject to any foices which cause 55 invention as long as the perturbation on layers other than the change in the thicknesses of the layers, the optical path layer of constituent A 5 is different than the perturbation on length of each layer. xXt), is generaUy constant. This results the layer of constituent A 5 . The conelation canceler yields a in generaUy constant attenuation of the optical eneigy and good approximation to eithei the primary signal 8^(1) or the thus, a generally constant offset in the measured signal. secondary signal n ^ t ) . In the event that an approximation TypicaUy, this offset portion of the signal is of Uttle interest 60 to the primary signal is obtained, the concentration of the since knowledge about a force which perturbs the material constituent of interest, c5(t), can often be determined since is usuaUy desfred. Any signal portion outside of a known in some physiological measurements, the thickness of the bandwidth of interest, including the constant undesired primary signal component, x5(t) in this example, is known or signal portion resulting from the generaUy constant absorpcan be determined. tion of the constituents when not subject to change, is 65 The conelation canceler utiUzes either the secondary removed. This is easUy accompUshed by traditional band reference n'(t) or the primaiy reference sXt) deteimined from pass filtering techniques. However, when the material is two substantiaUy simultaneously measured signals 8^(1)

5,632,272 15

16

and SjuXt). 8^(1) is determined as above in equation (7). measured signal yielding a good approximation to either the S^jXt) iS determined simUarly at a different wavelength Ab. primary signals s"j^(t>=e s .^5X5(1) or s"w,(t)=e5,^,£5X5(1) or To find either the secondaiy reference nXt) 01 the primary the secondary signals n"xJty=nyj(t) or n"A£,(t>=nw>(t). In the reference sXt). attenuated transmitted energy is measured at event that the primary signals are obtained, the concentrathe two different wavelengths Aa and Ab and transformed via 5 tion cs(t) may then be determined from the approximation to logarithmic conversion. The signals 8^(1) and 8^,(1) can the primary signal s ' ^ t ) or s " ^ ! ) according to: then be written (logarithm converted) as: 4 N S;U» = £5.^5*5©+.X; eu«< : «+ Z e i ^ c « 1=1 (=6 ,, S\4,) = ^sM^si') + n^t)

,m (9)

Swo = SSWWJW + "xiW

" 20

es^VsJu,

(12a)

n

(12b) 25

iko=v«.

where »»«*V»i

(13a)

«5 )*&•& XA-

(13b) 30

It is often the case that both equations (12) and (13) can be simultaneously satisfied. Multiplying equation (11) by ia and subtracting the result from equation (9) yields a non-zero secondary reference which is a linear sum of secondary 35 signal components: (14a) (15a)

40

N

x r&Mcxit) + x rjLittcKit) x ciKiCOleuu. - wa*] + 1=1

1!

X ejJ5(r)[eya - r^osl

1=6

(16a) 45

. .r

.

("b)

As discussed previously, the absorption coefficients are constant at each wavelength Aa and Ab and the thickness of the primary signal component, xs(t) in this example, is often known or can be determined as a function of time, thereby aUowing calculation ofthe concentration c 5 (t) of constituent

(ii)

Furthei transformations of the signals are the proportionaUty relationships in accordance with the signal model of the present invention defining ia and r,, simUar to equation (3), which aUows determination of a noise reference n'(t) and a primary reference s'(t). These are:

4

esCH'u.Cyes^sW

or

!=6

«'(») = S)jt) - rcS)j,(t) = n%jt) - f-eRu(f) 4 Af = ^x eyaCtfO + x^ euacptft) -

(17a)

10

Sua = ^SAJCSJSM + 1 ey^ci*; + x tiMctxi 1=1

Cs&y^Wj/esjjCstt) (8)

DETERMINAnON OF CONCENTRATION OR SATURATION IN A VOLUME CONTAINING MORE THAN ONE CONSTITUENT Referring to FIG. 6b, another material having N different constituents ananged in layers is shown. In this material, two constituents A 5 and A 6 are found within one layer having thickness x 5 6 (t)=x 5 (t)+x 6 (t), located generaUy randomly within the layer. This is analogous to combining the layers of constituents A 5 and A 6 in FIG. 6a. A combination of layers, such as the combination of layers of constituents ^ s and A5, is feasible when the two layers are under the same total forces which result in the same change of the optical path lengths x s (t) and x^t) of the layers, Often it is desirable to find the concentration or the saturation, i.e., a percent concentration, of one constituent within a given thickness which contains more than one constituent and is subject to unique forces. A determination of the concentration or the saturation of a constituent within a given volume may be made with any number of constituents in the volume subject to the same total forces and therefore under the same perturbation or change. To determine the saturation of one constituent in a volume comprising many constituents, as many measured signals as there are constituents which absorb incident Ught energy are necessary. It wiU be understood that constituents which do not absorb Ught energy are not consequential in the determination of saturation. To detennine the concentration, as .

.

,,

. . ,

,

. . . A t

many signals as there are constituents which absorb mcident Ught energy are necessary as weU as information about the suin of Multiplying equation (11) by r„ and subtracting the result concentrations, 50 from equation (9) yields a primary reference which is a It is often the case that a thickness under unique motion linear sum of primary signal components: contains only two constituents. For example, it may be desirable to know the concentration or saturation of A s s '(')=S)i«(f)-''Ai(')=sxt«6%Xocfr<5,6(»>«iXa(t) (i8a) n j j t ) . oi the primaiy portions, S j j t ) or s ^ t ) . of the =^(t)+^(') (18b)

5,632.272 17

18

•5w,(')=e5^Cs-«5,6W+«6^C6*5,6(f)+nM.(')

equations (18) and (19), or alternatively equations (20) and (21), by a conelation canceler The conelation canceler, again, requires a sample of either the primary reference s'(t) or the secondaiy reference n'(t) and a sample of either of the composite signals 8^(1) or 8^(1) of equations (18) and (19).

(19a)

=Su,W+nu,(f)(19b) It is also often the case that there may be two or more thicknesses within a medium each containing the same two constituents but each experiendng a separate motion as in FIG. 6c. For example, it may be desirable to know the DETERMINATION OF PRIMARY AND concentration or saturation of A 5 within a given volume SECONDARY REFERENCE SIGNALS FOR which contains A s and A 6 as weU as the concentration or SATURATION MEASUREMENTS saturation of Ag within a given volume which contains A3 10 One method foi detennining reference signals s'(t) or n'(t) and A4, A3 and A4 having the same constituency as A s and from the measured signals 8^(1) and Sxi(t) in accordance A 6 respectively. In this case, the primary signals 5^(1) and with one aspect of the invention is what wiU be refened to SA*(t) again comprise terms related to both A 5 and A 6 and as the constant saturation approach. In this approach, it is portions of the secondary signals n ^ t ) and n^Xt) comprise assumed that the saturation of A s in the volume containing terms related to both A3 and A 4 . The layers, A3 and A4, do not enter into the primary equation because they are 15 A s and Ag and the saturation of A3 in the volume containing A3 and A4 remains relatively constant over some period of assumed to be perturbed by a different frequency, or random time, i.e.: or enatic secondary forces which are unconelated with the primary force. Since constituents 3 and 5 as weU as conSaturation(A5(f))=C5(f)/[c5(f)+c6(t)] (22a) stituents 4 and 6 are taken to be the same, they have the same absoiption coefficients (i.e., tt:4(f)] (22b) eg xa and e.4_xb=e.6.7j,- GeneiaUy speaking, however, A3 and (23a) SaturationtAsWMl+tc^ycstf)]}- 1 A4 wiU have different concentrations than A5 and Ag and will therefore have a different saturation. Consequently a (23b) Saturation(A3(f))={l+[C4(fyC3(t)]}-1 single constituent within a medium may have one or more saturations associated with it. The primary and secondary 25 are substantiaUy constant over many samples of the measured signals S ^ and Sj^. This assumption is accurate over signals according to this model may be written as: many samples since saturation generaUy changes relatively slowly in physiological systems. (20a) **<.(') = fe.xA + e6^cs] xs_6{t) The constant saturation assumption is equivalent to 2 N (20b) 30 assuming that: n\c(t) = [esjWS + eejirf*] X3,\(t) + . X e ; ^ a Xi(t) + X e u * c; Xi(t) 1=1 "xJ.') = [(f) = [%.«.% + e6Mc6]

1=7 + n^(f)

xs^t)

c s (f)/c 6 (f)=constant,

(24a)

C3(f)/c4(f)=coiistant2

(24b)

(20c) (21a)

since the only other term in equations (23a) and (23b) is a constant, namely the numeral 1. ei^cpc^t). Using this assumption, the proportionaUty constants ia and r,, which aUow determination of the secondary reference (21c) "XAC) = [e5,M,c3 + e 6 A b c ¥ ] x3^(f) + nuit) signal n'(t) and the primary reference signal s'(t) in the where signals n ^ t ) and n^Xt) are simUar to the secondary 40 constant saturation method are: signals n ^ t ) and n ^ t ) except for the omission of the 3, 4 e 5 > cs X5,6(f) + £6^3 C6 X5,6(t) layer. (25a) *$& cs xsjift) + fiM C6 xs,6(t) Any signal portions whether primary or secondary, outside of a known bandwidth of interest, including the constant (26a) swysuit) undesired secondary signal portion resulting from the gen- 45 ^ . X o CS + £6JUC6 eraUy constant absorption of the constituents when not under (27a) *5AbCS+AbC6 perturbation, should be removed to determine an approxis-Ste (cslce) + e6£a mation to either the primary signal or the secondary signal (28a) esjii (C5/C6) + 65.Xi within the bandwidth of interest. This is easUy accompUshed by traditional band pass filtering techniques. As in the 50 (29a) s")M/s"lM = constants; where previous example, it is often the case that the total pertur(30a) nUt) * ra(t)nn{t) bation or change affecting the layers associated with the and secondary signal components is caused by random or enatic forces, causing the thickness of each layer, or the optical ZSte C3 X31,(t) + eoAo C4 XM(f) path length of each layer, x/t), to change enatically, pro- 55 (25b) e5MC3X3.4(f) + Z6MC4X3,4 M ) ducing a random or enatic secondary signal component (26b) nWyiikb(t) nxJt). Regardless of whethei 01 not the secondaiy signal e5.ioC3+e6^>C4 portion nxait) is enatic, the secondary signal component (27b) zsM<:3+e6JJ,C4 nxa(t) can be removed or derived via a conelation canceler, such as an adaptive noise canceler, having as one input a 60 «S.^ (C3/C4) + £ 6 ^ (28b) secondary reference n'(t) or a primary reference s'(t) deteresAi (C3/C4) + enM, mined by a processor of the present invention as long as the (29b) n"idf)lri"u(t) = constantt; where pertuibation in layeis othei than the layei of constituents A 5 (30b) siAfi^r^fisait). and Ag is different than the perturbation in the layer of constituents A 5 and Ag. Either the enatic secondary signal 65 components n ^ t ) and n^Xt) or the primary components In accordance with the present invention, it is often the 5^(1) and s^Xt) may advantageously be removed from case that both equations (26) and (30) can be simultaneously 35

2 JV nu(t) = [ejjiics + €6jiC4] *3,4(f) + Z e y j , cpcit) + X i= l i=7

(21b)

5,632,272 19

20

satisfied to determine the proportionaUty constants ia and rv. In other words, coefficients ia and iv are selected at values AdditionaUy, the absoiption coefficients at each wavelength which reflect the minimum of conelation between the prie s.Xa' e6.Xa- e s.xi' and Cgj^ are constant and the central mary signal portions and the secondary signal portions. In assumption of the constant satuiation method is that c5(t)/ piactice, one does not usuaUy have significant prioi infoiCg(t) and 03(1)^4(1) aie constant ovei many sample periods. 5 mation about either the primary signal portions 8^(1) and Thus, new proportionaUty constants r a and rv may be deter8^,(1) oi the secondary signal portions n ^ t ) and n ^ t ) ofthe mined every few samples from new approximations to either measured signals 8^(1) and 8^(1). The lack of this inforthe primary or secondary signal as output from the conelamation makes it difficult to determine which of the pluraUty tion canceler. Thus, the approximations to either the primary of coefficients r^ r2, . . . r n conespond to the signal signals 8^(1) and 8^,(1) or the secondary signals nXa(t) and 10 coefficients r ^ s ^ Q / s ^ t ) and r ^ n ^ ^ / n ^ ^ ) . n x*(t)- found by the conelation canceler for a substantiaUy One approach to determine the signal coefficients ia and immediately preceding set of samples of the measured r,, from the pluraUty of coefficients r^ r 2 rn employs the signals 8^(1) and Sj^Xt) are used in a processor of the use of a conelation canceler 27, such as an adaptive noise present invention for calculating the proportionality canceler, which takes a first input which conesponds to one constants. ia and rv, for the next set of samples of the 15 of the measured signals 8^(1) or Sj^Xt) and takes a second measured signals 8^(1) and 8^,(1). input which conesponds to successively each one of the Multiplying equation (19) by ia and subtracting the resultpluraUty of reference signals R ' ^ , t), R'(r2, t ) , . . . , RXr„, t) ing equation from equation (18) yields a non-zero secondary as shown in FIG. 7a. For each of the reference signals R'ta, reference signal: t). R(r 2 . t), . . . , R(r„, t) the conesponding output of the 20 correlation canceler 27 is input to a "squares" operation 28 n'(t)=S>Jt)-r^(iiF=n^(t)-rjiu,{t). (31a) w hi c h squares the output of the conelation canceler 27. The Multiplyingequation(19)byr v andsubtractingtheresultoutput of the squares operation 28 is provided to an inteing equation from equation (18) yields a non-zero primary ? * " j 2 ? f o r f o n m f g a cumulative output signal (a summareference sisnalsquares). The cumulative output signal is subse25 quently input to an extremum detector 31. The purpose of s X')=s>J.')-r^u,(t)=s^,(f)-rvs)jl(t). (3ib) the extremum detectoi 31 is to chose signal coefficients i a and iv from the set r^ r 2 , . . . r„ by observing which provide When using the constant saturation method in patient a maximum in the cumulative output signal as in FIGS. 7b monitoring, initial proportionaUty coefficients can be deterand 7c. In other words, coefficients which provide a maximined as further explained below. It is not necessary for the 30 mum integrated output, such as energy or power, from the patient to remain motionless even for an initialization conelation cancelei 27 conespond to the signal coefficients period. With values for the proportionaUty coefficients r a and Ia and Iv w hich relate to a minimum conelation between the rv determined, a conelation canceler may be utilized with a primary signal portions and the secondary signal portions in secondary reference n'(t) or a primary reference sXt). accordance with the signal model of the present mvention. 35 One could also configure a system geometry which would DETERMINAnON OF SIGNAL COEFFICIENTS require one to locate the coefficients from the set r^ r 2 ,. . . FOR PRIMARY AND SECONDARY r„ which provide a minimum or inflection in the cumulative REFERENCE SIGNALS USING THE output signal to identify the signal coefficients ia and rK. CONSTANT SATURATION METHOD Use of a pluraUty of coefficients in the processoi of the In accoidance with one aspect of the piesent mvention, 4 0 present mvention in conjunction with a conelation cancelei the leference processoi 26 of FIGS. 4a and HG. 4b of the 27 to detennine the signal coefficients ia and r„ may be present invention may be configured to multiply the second demonstrated by using the properties of conelation cancelmeasured assumed signal Sw,(t)=sXi(t)+nW)(t) by each of a lation. If x, y and z are taken to be any coUection of three pluraUty of signal coefficients r^ r 2 , . . . r„ and then subtract tune varying signals, then the properties of some conelation each result from the first measured signal S^(t)=s^(t)+n^ 45 cancelers C(x, y) may be defined as foUows: (t) to obtain a pluraUty of reference signals RXr, t^sW)-i-s^ty-nUfyr

n^t)

(32)

for rsrj, r 2 . . . . rn as shown in FIG. 7a. In other words, a pluraUty of signal coefficients are chosen to represent a cross section of possible signal coefficients. In order to determme either the primary reference sXt) or the secondary reference nXt) from the above pluraUty of reference signals of equation (32). signal coefficients ia and r^, are deteimined from the pluiaUty of assumed signal coefficients r^ r2, . . . r n . The coefficients ia and r„ are selected such that they cause either the primary signal portions s^(t) and s^Xt) or the secondary signal portions n^(t) and n ^ t ) to cancel or nearly cancel when they are substituted into the reference function R(r, t), e.g. ^(')=VM,(')

(33a)

nxpO^xtC)

(33b)

n'Cr^ffOv, t)=nUt)-rjixi,(t)

(33c)

*'(*)=K'('v. f&iJty-VtoV)-

(33d)

50

Properly (1) C(x, y)=0 for x, y correlated

(34a)

Properly (2) C(x, y)=x for x, y uncorrelated

(34b)

Property (3) C(.x^-y, z)=C(x, z}+C(y, z)

(34c)

With properties (1), (2) and (3) it is easy to demonstrate that the energy or power output of a conelation canceler with a first input which conesponds to one of the measured signals 55 S^(t) or 8^,(1) and a second input which conesponds to successively each one of a pluiaUty of reference signals R ^ , t), RXr2. t) RXr„. t) can determme the signal coefficients ia and r„ needed to produce the primary reference sXt) and secondary reference nXt). If we take as a first 6Q input to the conelation canceler the measured signal 8^(1) and as a second input the pluraUty of reference signals R'^, t), R'(i2, t ) , . . . , R(i„. t) then the outputs of the conelation cancelei C(SXa(t), R'(r7-,t)) foi j=l, 2 , . . . , n may be written as 65

Cis^i^n^Jt),

sjJO-yxkOWl)-^©)

(35)

where j=l, 2, . . . . n and we have used the expressions

5,632,272 21

22

K'O: t'>=SiJ.t)-rS}M

(36)

Sx«(f>=^(f)+n**(')

The energy functions given above, and shown in FIG. 7b, indicate that the conelation canceler output is usuaUy zero d u e t o c o r r elation between the measured signal S^(t) or

(37a) (37b)

S x ( t ) a n d manLy o f tfae l u r a l i * P ty o f reference signals R ' ^ , t), 5 R'(r 2 , t ) , . . . , R'(r„, t). However, the energy functions are non The use of property (3) aUows one to expand equation (35) zero at values of r. which conespond to canceUation of either into two terms the primary signal portions s^,(t) and s^Xt) or the secondary signal portions n ^ t ) and n w (t) in the reference signal R ' ^ , C(Sja(r),ff(«*))=C(jxo(<')^i<,(»)-wu>(r>Mijz,(»)-mxi(f)) t). These values conespond to the signal coefficients ia and r . +C(n%j.t),sUf>-rsii,(t)+nwymu>(t)) (38) io " ^ s h o u l d b e u n d e r s t o o d t h a t t h ^ e ^ y b e instances in

SwXO^wM+nwW-

so that upon use of properties (1) and (2) the conelation cancelei output is given by

time when either the primaiy signal portions s ^ t ) and s ^ t ) ° r * e secondary signal portions n ^ t ) and n^(t) are identically zero or nearly zero. In these cases, only one signal C(SUW(rj,t)y=sumrrrj»hjit)8(.rrrv) (39) coefficient value wiU provide maximum energy or power 1 5 output of the conelation canceler. where 8(x) is the unit impulse function Since there may be more than one signal coeffident value which provides maximum conelation canceler energy or power output, an ambiguity may arise. It may not be 5(x)=o if xM immediately obvious which signal coefficient together with 5fal=i if ;t=o Md 20 t*16 reference function R'(r, t) provides either the primary or secondary reference. In such cases, it is necessary to conThe time variable, t, of the conelation canceler output sider the constraints of the physical system at hand. For C(S^(t), R(r-, t)) may be eliminated by computing its example, in pulse oximetry, it is known that arterial blood, energy or power. The energy of the conelation canceler whose signature is the primary plethysmographic wave, has output is given bv 25 greater oxygen saturation than venous blood, whose signature is the secondary enatic or random signal. Consequently, E-iJr^C-iSiJfifiXrptjdt in pulse oximetry, the ratio of the primary signals due to arterial pulsation t^s^jfyls-jjit) is the smaller of the two ^O-j-rJ^tJfW^rj-r^iJfidt (41a) s i g n a l c o e ffident values whUe the ratio of the secondary is It should be understood that one could, equally weU, have 30 ^ f chosen the measured signal S

(fl as the first innnt to the

^

d ue t 0 1

1 S the lar

"f^ v e n o u s Wood dynamics r ^ n ^ O / n ^

ger

o f t h e t W 0 S1

g n a l coefficient values, assuming

... , 7 tu 1 t-t. c c • 1 conela ion canceler and the pluraUty of reference signals Kit,, t) R(r 2 , t ) , . . . , R ( r t) as the second input In this event the conelation canceler energy output is 2 E (r)=Jc (s (t),R(r,t)dt

Aa=660 nm and Ab=910 nm. ^ should ^ be understood ^ ^ ^ ^ ^ ctical t a t i o n s o f ^ p l u r a l i t y o f mference signals and cross conglator technique, the ideal features Usted as properties (1), 35 (2) and (3) above wiU not be precisely satisfied but wiU be approximations thereof. Therefore, in practical implemen=5(rrrJs2u(t)dt+&(rrr*)ln2u>(t'>dt (41b) tations of this embodiment of the present invention, the ^ u IA 1 u A t Atu f t: 1 -t t- tu conelation canceler energy curves depicted in FIG. 7b wiU It should also be understood that in practical situations the . . . „. _ .,. , aj , Ti .r .. . . .„ . c .. t tt - 1 u 1 A not consist of infinitely nanow delta functions but w m have use of discrete tune measurement signals may be employed 40 <. ... •,,. • t A -tu tu A • t A • r r o -» „ .. .. t - i A t fimte width associated with them as depicted m FIG. 7c. as weU as continuous time measurement signals. A system _. , , . . , , . . .. . .. . ... . . ,. , , At t- c t t: It should also be understood that it is possible to have which performs a discrete transform r(e.g., a saturation .. . . . „. . . . .. . j . j, . ^.t. t 1s • A -tu tu more than two signal coefficient values which produce e transform in the present example) in accordance with the . . ., , .. t- . A -U A -tu C t rmo maximum energy or power output from a conelation canpresent mvention is descnbed with reference to FIGS. , „,,. .. ~; . , ., , . , . %i
(42a)

55

JW)

=

nAitit)

i' = i,

ri*tj.

8(r,-r,.)At| .^nUti)-0.5(4,(10)+

nUtn)) )

Thus, reference signal techniques together with a cone(42b) EW) = Sfe - r„)At I X &(!•) - 0.5(^((b) + &&,)) } + „ l a tion canceUation, such as an adaptive noise canceler, can 60 I '=0 I be employed to decompose a signal into two or more signal , n .. components each of which is related by a ratio. S(rj-rr)At { ^fad-ostiM + fat,)) J PREFERRED CORRELAHON CANCELER USING A JOINT PROCESS ESTTMATOR where t,- is the i * discrete time, t 0 is the initial time, t„ is the 65 IMPLEMENTATION final time and At is the time between discrete time measureOnce either the secondary reference nXt) or the primary ment samples. reference sXt) is determined by the processor of the present

5,632,272 23

24

invention, the conelation canceler can be implemented in or n ' ^ t ) of the joint process estimator 60 is a very good either hardware or software. The prefened implementation approximation to either the primary signal 5^(1) or the of a conelation canceler is that of an adaptive noise canceler secondary signal n ^ ^ ) . using a joint process estimator. The joint process estimator 60 can be divided into stages, The least mean squares (LMS) implementation of the 5 beginning with a zero-stage and terminating in an m"!-stage, internal processor 32 described above in conjunction with stage, as shown in FIG. 8. Each stage, except foi the the adaptive noise cancelei of FIGS. 5a and FIG. Sb is zero-stage, is identical to every other stage. The zero-stage relatively easy to implement, but lacks the speed of adapis an input stage for the joint process estimator 60. The first tation desirable for most physiological monitoring appUcastage through the m^-stage work on the signal produced in tions of the present invention. Thus, a faster approach for 10 the immediately previous stage, i.e., the (m-l)"'-stage, such adaptive noise canceling. caUed a least-squares lattice joint that a good primary signal approximation s"Xa(t) or secondprocess estimator model, is used in one embodiment. A joint ary signal approximation n ' ^ t ) is produced as output from process estimator 60 is shown diagrammaticaUy in FIG. 8 the m^-stage. and is described in detaU in Chapter 9 of Adaptive Filter The least-squares lattice predictor 70 comprises registers Theory by Simon Haykin, pubUshed by Prentice-HaU, copy- 15 90 and 92, summing elements 100 and 102, and delay right 1986. This entire book, including Chapter 9, is hereby elements 110. The registeis 90 and 92 contain multipUcative incoiporated herein by reference. values of a foiwaid reflection coefficient F ^ t ) and a The function of the joint process estimatoi is to remove backward reflection coefficient Fbm(t) which multiply the eithei the secondaiy signal portions n ^ t ) 01 n^Xt) or the reference signal nXt) or sXt) and signals derived from the primary signal portions 8^(1) or 8^,(1) from the measured 20 reference signal n'(t) or sXt). Each stage of the least-squares signals 8^(1) or Sxi(t), yielding either a primary signal lattice predictor outputs a forward prediction enor fm(t) and approximation s"Xa(t) or 8^,(1) or a secondary signal a backward prediction enor bm(t). The subscript mis indicaapproximation n"^(t) or n ' ^ t ) . Thus, the joint process tive of the stage. estimator estimates either the value of the primary signals For each set of samples, i.e. one sample of the reference 8^(1) or 8^(1) or the secondary signals n^(t) or n ^ t ) . The 25 signal n'(t) or s'(t) derived substantiaUy simultaneously with inputs to the joint process estimator 60 are either the one sample of the measured signal S^(t). the sample of the secondary reference n'(t) or the primary reference sXt) and reference signal n'(t) or sXt) is input to the least-squares the composite measured signal S j j t ) or 8^,(1). The output lattice predictor 70. The zero-stage forward prediction enor is a good approxUnation to the signal S^(t) or 8^(1) with f0(t) and the zero-stage backward prediction enor b0(t) are either the secondary signal or the primary signal removed, 30 set equal to the reference signal nXt) or s'(t). The backward i.e. a good approxUnation to either Syjl), s^Xt), n ^ t ) or prediction enor b0(t) is delayed by one sample period by the nj^Xt)delay element 110 in the first stage of the least-squares The joint process estimator 60 of FIG. 8 utiUzes, in lattice predictor 70. Thus, the immediately previous value of conjunction, a least square lattice predictor 70 and a regresthe reference nXt) or sXt) is used in calculations involving sion filter 80. Either the secondary reference nXt) or the 35 the first-stage delay element 110. The zero-stage forward primaiy reference sXt) is input to the least square lattice prediction enor is added to the delayed zero-stage backward predictor 70 whUe the measured signal 8^(1) or 8^(1) is prediction enor b 0 (t-l) multipUed by the forward reflection input to the regression filter 80. For simpUdty in the coefficient value F ^ t ) register 92 value, to produce a foUowing description. 8^(1) wiU be the measured signal first-stage forward prediction enor f^t). AdditionaUy, the from which either the primaiy portion s ^ t ) 01 the secondaiy 40 zero-stage forward prediction enor f0(t) is multipUed by the portion n ^ t ) wUl be estimated by the joint process estimabackward reflection coeffident r 6 a (t) register 90 value and tor 60. However, it wiU be noted that 8^,(1) could also be added to the delayed zero-stage backward prediction enor input to the regression filter 80 and the primary portion s^Xt) b 0 (t-l) to produce a first-stage backward prediction enor or the secondary portion n ^ t ) of this signal could be b^t). In each subsequent stage, m, of the least square lattice estimated. 45 predictor 70, the previous forward and backward prediction The joint process estimator 60 removes aU frequendes enor values, f ^ ^ t ) and b m _ 1 (t-l), the backward prediction that are present in both the reference nXt) or sXt), and the enor being delayed by one sample period, are used to measured signal 8^(1). The secondary signal portion nj^t) pioduce values of the forward and backward prediction usuaUy comprises frequencies unrelated to those of the enors for the present stage, fOT(t) and bm(t). primary signal portion s ^ t ) . It is improbable that the 50 The backward prediction enor bm(t) is fed to the concursecondary signal portion n ^ t ) would be of exactly the same rent stage, m, of the regression filter 80. There it is input to spectral content as the primary signal portion sXa(t). a register 96. which contains a multipUcative regression However, in the unlikely event that the spectral content of coefficient value KOT ^(t). For example, in the zero-stage of 5^(1) and n ^ t ) are simUar, this approach wiU not yield the regression filter 80, the zero-stage backward prediction accurate results. FunctionaUy, the joint process estimator 60 55 enor b0(t) is multipUed by the zero-stage regression coefcompares the reference mput signal n'(t) or sXt). which is ficient ^ ^ ( t ) register 96 value and subtracted from the conelated to either the secondary signal portion n ^ t ) or the measured value of the signal S ^ t ) at a summing element primary signal portion 8^(1), and input signal 8^(1) and 106 to produce a first stage estimation enor signal e1Xa(t). removes aU frequencies which are identical. Thus, the joint The first-stage estimation enor signal ei:Ka(t) is a first process estimator 60 acts as a dynamic multiple notch filter 60 approximation to either the primary signal or the secondary to remove those frequencies in the secondary signal comsignal. This first-stage estimation enor signal e 1Xa (t) is input ponent n^(t) as they change enaticaUy with the motion of to the first-stage of the regression filter 80. The first-stage the patient or those frequendes in the primaiy signal combackward prediction enor b^t), multipUed by the first-stage ponent 5^(1) as they change with the arterial pulsation of the regression coefficient i q ^ ( t ) register 96 value is subtracted patient. This yields a signal having substantiaUy the same 65 from the first-stage estimation enor signal e1Xa(t) to produce spectral content and ampUtude as either the primary signal the second-stage estimation enor e 2 ^(t). The second-stage Sx^Ct) or the secondary signal n ^ t ) . Thus, the output s ' ^ t ) estimation enor signal e2Xa(t) is a second, somewhat better

5,632,272 25

26

approximation to either the primary signal s ^ t ) or the ence nXt) or the primary reference s'(t) are input to the joint secondary signal n ^ t ) . process estimator 60, as shown in FIG. 8. The forward and The same processes are repeated in the least-squares backward prediction enor signals f0(t) and b 0 (t). and interlattice predictor 70 and the regression filter 80 for each stage mediate variables including the weighted sums of the foruntil a good approximation emXJt), to either the primaiy 5 ward and backward enor signals 3 0 (t) and So(t), and the signal s^Jf) or the secondary signal n^(t) is determined conversion factor y0(t) are calculated foi the zero-stage Each of the signals discussed above, including the forward according to: prediction enor f m (t), the backward prediction enor b m (t), the estimation enoi signal e m ^ ( t ) , is necessary to calculate /cXfH'oW^'W (5ia) 2 the forward reflection coefficient FfJJt), the backward 1° %(t)=&0(t)=KS0(t-i}t-\nXf)\ (52a) reflection coefficient Fb m (t), and the regression coefficient Vx^Ct) register 90, 92, and 96 values in each stage, m. In Yo(«-i)=i (53a) addition to the forward prediction enor f m (t), the backward ^ a s e c o n d a r y reference n'© is used or according to: prediction enor b m (t), and the estimation enor e ^ ^ t ) signals, a number of intermediate variables, not shown in 1 5 /o(fH,o(f)=i'(') (5ib) FIG. 8 but based on the values labeled in FIG. 8, are required to calculate the forward reflection coefficient FfJt), the 3o«=PoW=^5o(f-iH^Wi2 (52b) backward reflection coeffident r 6 j n ( t ) , and the regression Cf-n-i rssbi Yo( coefficient Km ^ ( t ) register 90,92, and 96 values. Intermediate variables include a weighted sum of the 2 0 if a primary reference s'(t) is used where, again, A without a forward prediction enor squares 3 m (t), a weighted sum of wavelength identifier, a or b, is a constant multipUcative the backward prediction enor squares S(t), a scalar paramvalue unrelated to wavelength. eter A J t ) , a conversion factor yjt), and anothei scalar Forward reflection coefficient r / m ( t ) , backward reflection parameter p m A a (t). The weighted sum of the foiward precoefficient TbJt). and regression coeffident K ^ t ) regisdiction enors'3 m (t) is defined as: ter 90, 92 and 96 values in each stage thereafter are set according to the output of the previous stage. The forward t (44) reflection coefficient Ffl(t), backward reflection coefficient 2 3m« = .j^ *•'- M)i ; rj, ?1 (t), and regression coeffident iq ^ ( t ) register 90,92 and 96 values in the first stage are thus set according to the where A without a wavelength identifier, a or b, is a constant 30 ^gorithm u s i n g v a l u e s & fae Z ero-stage of the joint process multipUcative value unrelated to wavelength and is typicaUy estimator 60. In each stage, m g 1, intermediate values and less than or equal to one, i.e., A^ 1. The weighted sum ofthe register values including the parameter A ^ t ) ; the forward backward prediction enors 6 m (t) is defined as: reflection coefficient F ^ t ) register 90 value; the backward reflection coefficient Tbjrl(t) register 92 value; the forward gm(t)= ix'-i\bm(t)P ^ 35 and backward enor signals fm(t) and b m (t); the weighted sum 1=1 of squared forward prediction enors 3 ^ ( 1 ) as manipulated in where, again, A without a wavelength identifier, a or b, is a §9-3 of ^ H ^ 1 1 b o o k ; t h e w e i g h t e d s u m of s ( l u a r e d constant multipUcative value unrelated to wavelength and is backwaid prediction enors S 6 , m (t), as manipulated in §9.3 typicaUy less than or equal to one, i.e.. Ag 1. These weighted 4 0 of the Haykin book; the conversion factor yjt); the paramsum intermediate enor signals can be manipulated such that eter p m , Xa (t); the regression coeffident Km ; J t ) register 96 they are more easUy solved for, as described in Chapter 9, value; and the estimation enor em+1^,(t) value are set §9.3 of the Haykin book referenced above and defined according to: hereinafter in equations (59) and (60). An-t^An-dt-mb^t-lW^m^t-l)} (54) 45 DESCRIPTION OF THE JOINT PROCESS r^f^A^tyS^r-i)} (55) ESTIMATOR The operation of the joint process estimator 60 is as foUows. When the joint process estimator 60 is turned on, the initial values of intermediate variables and signals 50 including the parameter A m _ 1 (t), the weighted sum of the foiwaid prediction enoi signals S ^ t ) , the weighted sum of the backwaid prediction enor signals S ^ . ^ t ) , the parameter p m >Xt). and the zero-stage estimation enor e 0 ^ ( t ) are initiaUzed, some to zero and some to a smaU positive 55 number 8: Am-,(0)=0; 3m-,(0)=5; fim-i(0)=5;

(46)

r^OMA^^fyS^-iW}

(56)

MtW^-dtyr^w^t-i)

(57)

t

r!l

*mW=*m-i( -i)+ W'ym-i(') 3„(r)=3m_1(r)-{iA^i(f)i2/S^i('-i)}

(58) (59)

&m(t'F&m-l(t-i)-{lAm-l(t)\2l3^.l(t)} ym(t-i)=^m,l(t-i)-{\bm_1(i-i)\2/&m.l(t-i)}

(60) (61)

pm^(t)='>^m.Xa('-^Mbm(^*m.^tyUt)}

(62)

^I^.(')=^JU,(')-K*„('W)

(<*)

(47) 60 (48)

Pm,Xtf(0)=0; (49) e (t)^s MfbrtSO (50) "*" *" 65 After initiaUzation, a simultaneous sample of the measured signal 8^(1) or 8^(1) and either the secondary refer-

where a (*) denotes a complex conjugate. These equations cause the enor signals fm(t), b m (t), e m A a (t) to be squared or to be multipUed by one another, in effect squaring the enors, and creating new intermediate enor values, such as A m _ 1 (t). The enor signals and the interme-

5.632.272 27

28

diate enor values are recursively tied together, as shown in the above equations (54) through (64). They interact to minimize the enor signals in the next stage. After a good approximation to either the primary signal SxJt) or the secondary signal n ^ t ) has been determined by the joint process estimator 60, a next set of samples, including a sample of the measured signal Sx^t) and a sample of either the secondary reference n'(t) or the primary reference sXt), are input to the joint process estimator 60. The re-initiaUzation process does not re-occur, such that the forward and backward reflection coefficient F ^ t ) and Fbjn (t) register 90, 92 values and the regression coeffident KmaaCt) register 96 value reflect the multipUcative values required to estimate either the primary signal portion 8^(1) or the secondary signal portion n ^ t ) of the sample of 8^(1) input previously. Thus, information from previous samples is used to estimate either the primary or secondary signal portion of a present set of samples in each stage. In a more numericaUy stable and prefened embodiment of the above described joint process estimator, a normaUzed joint process estimator is used. This version of the joint process estimator normalizes several variables ofthe abovedescribed joint process estimator such that the normaUzed variables faU between - 1 and 1. The derivation of the normalized joint process estimator is motivated in the Haykin text as problem 12 on page 640 by redefining the variables defined according to the foUowing conditions: /»(') =

Yo(0)=l

2. At each instant t ^ l , generate the various zeroth-order variables as foUows: 10

bo(t)=Mt) =

JffiL

15

3. For regression filtering, initialize the algorithm by setting at time index t=0 20

Pm(0)=O

4. At each instant t ^ l , generate the zeroth-order variable «o(rH/(f). 25

30

K(t)Ut) 35

N 3B(t)Pm(r-1)

+ N2(t)

Nm

bUt) N

Yo('-l)=l Ht) = ^Ut-l.)

Ut)

NSnMlUf-l)

b(t) =

Am«

P„-i(0)=5=10^

This transformation aUows the conversion of Equations (54)-(64) to the foUowing normalized equations: 40

AJ^1(») = Am.i(»- iMi-K.-iWPpti-

Accordingly, a normalized joint process estimator can be used for a more stable system. In yet another embodiment, the conelation canceUation is performed with a QRD algorithm as shown diagrammaticaUy in FIG. 8a and as described in detaU in Chapter 18 of Adaptive Filter Theory by Simon Haykin, pubUshed by Prentice-HaU, copyright 1986. The foUowing equations adapted from the Haykin book conespond to the QRD-LSL diagram of FIG. 8a (also adapted from the Haykin book). Computations a. Predictions: For time t=l, 2 , . . . , and prediction order m=l, 2 , . . . . M, where M is the final prediction order, compute: P„-i(f- 1) = X.p„-i(»- 2) + l E i ^ K t - 1)P

ifc^iCf- i)Pp+*>„.!«- iy^i(t)

fcm«=-

45

[6)»-i(f-i)-A„_i(ty„- 1 (t)]

[i-iA^MPmi-i/UwPp

Cb*-l(t+l) =

I

a&c-i)

U^t)-1^(1)1^(1-i)) / * » = •

eL-^f-l)

fc»-i('- 1 ) = - „ 1 „

[1 - IA„-i(»)Pp [1 - iS^-Kr- I)?]"2

50 e/Jf) = e»^i(t - l ) e / ^ i ( t ) - ^ ( r - l ) * . " ^ ^ - 1 )

|3m(r) = [l-IA^i(f)P]p m -i(t-l) T£m-l(>) = Cbjn-l(t- l)\iaTfo.i(t-

l) + sbJr-i(t-

l)e/^-i(f)

Tm(f) = Y'»-i(')[l-l*»»-i(')P] 55 1/2

Pm(0

'm(t)Ut-l) •Jf- i ) M ' ) L Y/nC

ZmtllM = SntU') -

p m (r-l) + fem(r)em(f)

•>!?(*-l) = c» J »-i(*-l)tf£(/-l) 3^iW = X3^i(*- 1) + IG/^IWP

Pjt)bjt)

Cfjn-l(t)=-

60

InitiaUzation of NormaUzed Joint Process Estimatoi Let N(t) be defined as the reference noise input at time index n and U(t) be defined as combined signal plus noise input at time index t the foUowing equations apply (see 65 Haykin, p. 619): 1. To initialize the algorithm, at time t=0 set

*-m3£i(.t- 1) ^l(f)

•r/m-lW = = fv»(') = C/.m-l(f)e b.m-l(t - 1) - &rf(»)k 1 / 2 '£™-l(' - 1)

5,632,272 29

30

-continued

for detennining either the secondary reference nXt) or the primary reference sXt). The flowchart for the joint process estimator is implemented in software. A one-time initiaUzation is performed when the physiological monitor is powered-on, as indicated by an "INITIALIZE NOISE CANCELER" action block 120. The initialization sets aU registers 90, 92. and 96 and delay element variables 110 to the values described above in equations (46) through (50). Next, a set of simultaneous samples of the composite measured signals 8^(1) and 8^,(1) is input to the subroutine represented by the flowchart in FIG. 9. Then a time update of each of the delay element program variables occurs, as indicated in a 'TIME UPDATE OF [Z -1 ] ELEMENTS" action block 130. The value stored in each of the delay element variables 110 is set to the value at the input of the delay element variable 110. Thus, the zero-stage backward prediction enor b0(t) is stored as the first-stage delay element variable, and the first-stage backward prediction enor bjXt) is stored as the second-stage delay element variable, and so on.

' C - l W = cfjo-l/Wtfm-iit-

1) + Wm-lWe »*-l(' - 1)

b. Filtering: For order m = 0, l,...,M—

1; and time

( = 1 , 2 , . ...compute Pm(«) = a.p m (r-1) + l64j„(f)P C6J»(()=-^

10

p^(r-l)

•S'i.Ji.W =

15

e »n(t) = c j ^ e m(r) - 4j()X^pm*(tpm*(t) = ctJfiWpJVf-

1)

1) + J4.m(t)em(r) 20

-^1« = ^(r)Y^«

Then, using the set of measured signal samples 8^(1) and SxfXt), the reference signal is obtained using the ratiometric or the constant saturation methods described above. This is 25 indicated by a "CALCULATE REFERENCE [n\t) or s'(t)] FOR TWO MEASURED SIGNAL SAMPLES" action 5. InitiaUzation block 140. a. AuxiUary parameter initialization: for oidei m=l, 2 , . . A zero-stage order update is performed next as indicated . , M, set in a "ZERO-STAGE UPDATE" action block ISO. The 30 zero-stage backward prediction enor b0(t), and the zero'fe»-i(0)=^».i(0)=0 stage forward prediction enor f0(t) are set equal to the value P™(0)=0 of the reference signal nXt) or sXt). AdditionaUy, the weighted sum of the forward prediction enors 3 m (t) and the b. Soft constraint initialization: For order m = 0 , 1 , . . . , M, weighted sum of backward prediction enors Pm(t) are set 35 set equal to the value defined in equations (47) and (48). Next, a loop counter, m, is initialized as indicated in a P»(-iH "m=0" action block 160. A maximum value of m, defining the total numbei of stages to be used by the subroutine 3 m (0)=5 40 conesponding to the flowchart in FIG. 9, is also defined. where 8 is a smaU positive constant TypicaUy, the loop is constructed such that it stops iterating once a criterion for convergence upon a best approximation c. Data initialization: For t=l, 2 , . . . , compute to either the primary signal or the secondary signal has been met by the joint process estimator 60. AdditionaUy, a maxie/,oW=«i,o(f)=HW 45 mum number of loop iterations may be chosen at which the loop stops iteration. In a prefened embodiment of a physi
5,632,272 31

32

forward and backward prediction enors 3 m (t) and p m (t) are em^^(t)=emM(tyK*^u,(i)bm(ty, and (67) less than a small positive number. An output is calculated > next, as indicated by a "CALCULATE OUTPUT" action «*»»-&»» f O T '=»• (68) block 200. The output is a good approximation to either the The second regression filter has a regression coefficient primary signal or secondary signal, as determined by the 5 ^ ^ ( t ) register 98 value defined simUarly to the first reference processor 26 and joint process estimator 60 subregression filter enor signal values, i.e.: routine conesponding to the flow chart of FIG. 9. This is g „ displayed (or used in a calculation in another subroutine), as iW»('MP»a*(»y Jf)y,ox () mdicated by a "TO DISPLAY" action block 210. A new set of samples of the two measured signals 8^(1) w These values are used in conjunction with those intermediate and SjuXt) is input to the processor and joint process estivariable values, signal values, register and register values mator 60 adaptive noise cancelei subroutine conesponding defined in equations (46) through (64). These signals are to the flowchart of FIG. 9 and the process reiterates for these calculated in an order defined by placing the additional samples. Note, however, that the initiaUzation process does signals immediately adjacent a simUar signal for the wavenot re-occur. New sets of measured signal samples 8^(1) is length Aa. Fo and S;uXt) are continuously input to the reference processor r the constant saturation method, 8^,(1) is input to the 26 and joint process estimator adaptive noise canceler second regression filtei SOb. The output is then a good subroutine. The output foims a chain of samples which is approximation to the primary signal portion s ' ^ t ) or secrepresentative of a continuous wave. This waveform is a ondary signal portion n"^,^). good approxUnation to either the primary signal waveform 20 The addition of the second regression filter 80Z> does not 5^(1) or the secondary waveform n Xa (t) at wavelength Aa. substantiaUy change the computer program subroutine repThe waveform may also be a good approximation to either resented by the flowchart of FIG. 9. Instead of an order the primary signal wavefonn 5^(1) oi the secondaiy waveupdate of the m"1 stage of only one regression filtei, an order form n ' ^ t ) at wavelength Ab. update of the m" 1 stage of both regression filters 80a and 80& 25 i s A conesponding flowchart for the QRD algorithm of FIG. Performed. This is characterized by the plural designation in 8a is depided in FIG. 9a, with reference numeral cone^ " 0 R D E R t ™ ™ 0 F * * S T A G E 0 F ^-G^SSSI0N sponding in number with an 'a' extension FILTER(S)" activity block 180 in FIG. 9. Since the regression filters 80a and 80fc operate independently, indeCALCULAITON OF SATURATION FROM pendent calculations can be performed in the reference CORRELATION CANCELER OUTPUT 30 processor and joint process estunator 60 adaptive noise cancele Physiological monitors may use the approximation of the y s u b r o u t i n e modeled by the flowchart of HG. 9. t o n a t l V e d l primary signals s"jjt) or s ^ t ) or the secondary signals . S 1 ? * ^ J o m t Pf ocess e s t l m a t o r o f I 7 r ^ u mG n - j j t ) or n-jjt) to calculate another quantity, such as the ™' usm\the 9 ^ ^ ^ l ^ haVln/.tW0 regreSs l o n meis 1S saturation of one constituent in a volume containing that f?™ m ^ 10a. This type of joint process constituent plus one or more other constituents. GeneraUy, 35 e u s t l ^ a t o r would be used for conelation canceUation usmg such calculations require information about either a primaiy ^ ^ ^gonthm descnbed m the Haykin book, oi secondaiy signal at two wavelengths. Foi example, the CALCULATION OF SATURATION constant saturation method requires a good approximation of the primary signal portions 5^(1) and 8^,(1) of both meaOnce good approximations to the primary signal portions sured signals Sjjt) and 8^,(1). The arterial saturation is 4 0 s ' ^ t ) and s ' ^ t ) or the secondary signal portion n ' ^ t ) and determined from the approximations to both signals, i.e. n'^^t), have been determined by the joint process estimator s"xa(t) and s ' ^ t ) . The constant saturation method also 60, the saturation of A 5 in a volume containing A 5 and A6, requires a good approximation of the secondary signal for example, may be calculated according to various known portions n ^ t ) or n ^ t ) . An estimate of the venous saturamethods. MathematicaUy, the approximations to the primary tion may be determined from the approximations to these 4 5 signals can be written in terms of Aa and Ab, as: signals i.e. n ' ^ t ) and n ' ^ t ) . s"!j'H5,!u^S*5,6('}+<6,X*C&*5.6('); ^ d (70) A joint process estimator 60 having two regression filters 80a and 80& is shown in FIG. 10. Afirstregression filter 80a s\i(0^5,xiC5*5,6(f>^,x<£6*5.6(')(71) accepts a measured signal 8-^(1). A second regression filter „ . ,„„„ .„,.,, SOb accepts a measured signal S w ( t ) for a use ofthe constant 50 Equation 5 (70) and (71) are equivalent to two equations havin saturation method to determine the reference signal n'(t) or g . t h r e e unknowns, namely c 5 (t), c 6 (t) and x 5 . 6 (t). The sXt). The first and second regression filters 80a and SOb aie sfuration can be determined by acquinng approximations to independent. The backward prediction enor b m (t) is input to the pnmary or secondary signal portions at two different, yet each regression filter 80a and 80b, the input for the second P e n m a t e times t, and t 2 over which the saturation of A 5 in regression filter SOb bypassing the first regression filter 80a. 55 t h e v o l u m e contaming A 5 and A 6 and the saturation of A 3 in rr,, . . ... o A , . „„ the volume containing A , and A 4 does not change substanThe second regression filter SOb compnses registers 98, t- „ „ , % . - i t- t A t A • i t tno A • -i i t tu • t, tially. For example, for tthe pnmaiy signals estimated at and summing elements 108 ananged similarly to those in the times t d t~1 first regression filtei 80a. The second regression filtei SOi ^2" operates via an additional intermediate variable in conjunc- 60 J"v.('i)-%^e5*i,6('i)+«6,!^e^^('i) (72) tion with those defined by equations (54) through (64), i.e.: *V('l)^5,^5*5,6(tl>+<*,UC6*!,6(fl0

p m ,^('Hp^('-iMU<)eW'yY m (f)}; ™d PoAi(0)=0-

(73)

(65) ^(h^jufsXsAhy^J^fitiXsdh)

(74)

*V('2)^,;^5%,6(<2)+%,UAi*S,6(<2)

(75)

(66)

The second regression filter SOb has an enor signal value 65 defined simUar to the first regression filter enor signal Then, difference signals may be detenmned whichrelatethe values, e m + 1 ^ ( t ) . i.e.: signals of equations (72) through (75), i.e.:

5,632,272 33 Asia=s''iJ.'i)-s''}J.t2>*5*<£5te+e6ju£(Ax; and

34

(76) niques. When the patient moves, this causes perturbation such as changing optical pathlength due to movement of A^^Vft^V^Hs.xi^Ax^^c^,(77) background fluids (e.g., venous blood having a different where Ax=xs Sih^ &>• The average saturation at time saturation than the arterial blood). Therefore, the measured 5 sl t+ft -i-t V2 isSnal becomes enatic. Enatic motion mduced noise typi' caUy cannot be predetermined and/or subtracted from the Saturation^ = flMfeM + <*«] (78) measured signal via traditional filtering techniques. Thus, determining the oxygen saturation of arterial blood and e&u-ewtAWAqj,) ^ venous blood becomes more difficult. 1° A schematic of a physiological monitor for pulse oximetry (e 6A* - ^SXbXAska/Asu,) is shown in FIGS. 11-13. FIG. 11 depicts a general hardware block diagram of a pulse oximeter 299. A sensor 300 has two It wiU be understood that the Ax term drops out from the ^ e m i t t e r s 301 and 302 such as LED's. One LED 301 satuiation calculation because ofthe division. Thus, knowlemitting Ught of red wavelengths and another LED 302 edge of the thickness of the primary constituents is not 15 emitting Ught of infrared wavelengths are placed adjacent a required to calculate saturation. 3 1 0 A photodetector 320, which produces an electriflnger cal DTTT c c /-wTn-rcrov ure A CTTDCH^T^TTC signal conespondmg to the attenuated visible and infraPULSE OXIMETRY MEASUREMENTS , . energy signals . , is - ,located . . . opposite -t tu C ™ s 301 IAI red, ,. Ught the TLED A specific example of a physiological monitor utilizing a and 302. The photodetector 320 is connected to front end processor ofthe present mvention to determine a secondary 20 analog signal conditioning circuity 330. reference n'(t) for input to a correlation canceler that The front end analog signal conditioning drcuitry 330 has removes enatic motion-induced secondary signal portions is outputs coupled to analog to digital conversion circuit 332. a pulse oximeter. Pulse oximetry may also be performed The analog to digital conversion circuitry 332 has outputs utiUzing a processor of the present invention to detennine a coupled to a digital signal processing system 334. The primary signal reference sXt) which may be used for display 2 5 digital signal processing system 334 provides the desfred purposes or for input to a conelation canceler to derive parameters as outputs for a display 336. Outputs for display information about patient movement and venous blood are, for example, blood oxygen saturation, heart rate, and a oxygen saturation. clean plethysmographic waveform. A pulse oximeter typicaUy causes energy to propagate The signal processing system also provides an emitter through a medium where blood flows close to the surface for cunent control output 337 to a digital-to-analog converter example, an ear lobe, or a digit such as a finger, a forehead circuit 338 which provides control information for Ught or a fetus' scalp. An attenuated signal is measured after emitter drivers 340. The Ught emitter drivers 340 couple to propagation through or reflected from the medium. The the Ught emitters 301, 302. The digital signal processing pulse oximeter estimates the saturation of oxygenated blood. system 334 also provides a gain control output 342 for the Freshly oxygenated blood is pumped at high pressure front end analog signal conditioning circuitry 330. from the heart into the arteries for use by the body. The FIG. l l a Ulustrates a prefened embodiment for the cornvolume of blood in the arteries varies with the heartbeat, bination of the emitter drivers 340 and the digital to analog giving rise to a variation in absorption of energy at the rate conversion drcuit 338. As depicted in FIG. lla, the driver of the heartbeat, or the pulse. 4 0 comprises first and second input latches 321, 322, a synOxygen depleted, or deoxygenated, blood is returned to chronizing latch 323, a voltage reference 324, a digital to the heart by the veins along with unused oxygenated blood. analog conversion circuit 325, first and second switch banks The volume of blood in the veins varies with the rate of 326,327, first and second voltage to cunent converters 328, breathing, which is typicaUy much slower than the heartbeat. 329 and the LED emitters 301, 302 conesponding to the Thus, when there is no motion mduced variation in the 45 LED emitters 301, 302 of FIG. 11. thickness of the veins, venous blood causes a low frequency The prefened driver depicted in FIG. l l a is advantageous variation in absorption of energy. When there is motion in that the present inventors recognized that much of the induced variation in the thickness of the veins, the low noise in the oximeter 299 of FIG. 11 is caused by the LED frequency variation in absorption is coupled with the enatic emitters 301, 302. Therefore, the emitter driver drcuit of variation in absorption due to motion artifact 50 FIG. l l a is designed to minimize the noise from the emitters In absorption measurements using the transmission of 301, 302. The first and second input latches 321, 324 are energy through a medium, two Ught emitting diodes (LED's) connected directly to the DSP bus. Therefore, these latches are positioned on one side of a portion of the body where significantly mimmizes the bandwidth (resulting in noise) blood flows close to the surface, such as a finger, and a present on the DSP bus which passes through to the driver photodetectoi is positioned on the opposite side of the finger. 55 circuitry of FIG. lla. The output of the first and second mput TypicaUy, in pulse oximetry measurements, one LED emits latches only changes when these latched detect their address a visible wavelength, preferably red. and the other LED on the DSP bus. The first input latch receives the setting for emits an infrared wavelength. However, one skiUed in the art the digital to analog converter circuit 325. The second input wiU reaUze that other wavelength combmations could be latch receives switching control data for the switch banks used. The finger comprises skin, tissue, muscle, both arterial 60 326,327. The synchronizing latch accepts the synchronizing blood and venous blood, fat, etc., each of which absorbs pulses which maintain synchronization between the activaUght energy differently due to different absorption tion of emitters 301,302 and the analog to digital conversion coefficients, different concentrations, different thicknesses, circuit 332. and changing optical pathlengths. When the patient is not The voltage reference is also chosen as a low noise DC moving, absorption is substantiaUy constant except for the 65 voltage reference for the digital to analog conversion circuit flow of blood. The constant attenuation can be determined 325. In addition, in the present embodiment, the voltage and subtracted from the signal via traditional filtering techreference has an lowpass output filter with a very low corner

5,632,272 35

36

frequency (e.g., 1 Hz in the present embodiment). The signal processing system 334, the cunent level foi the red digital to analog converter 325 also has a lowpass filter at its and infrared emitters is maintained constant. It should be output with a very low coinei frequency (e.g., 1 Hz). The undeistood, howevei, that the cunent could be adjusted foi digital to analog converter provides signals for each of the changes in the ambient room Ught and other changes which emitters 301, 302. 5 would effect the voltage input to the front end analog signal In the present embodiment, the output of the voltage to conditioning drcuitry 330. In the present invention, the red cunent converters 328. 329 are switched such that with the and infrared Ught emitters aie modulated as foUows: foi one emitteis 301, 302 connected in back-to-back configuration, complete 625 Hz red cycle, the red emittei 301 is activated only one emittei is active an any given time. In addition, the foi the first quartei cycle, and off foi the remaining thieevoltage to cunent converter for the inactive emitter is 10 quarters cycle; for one complete 625 Hz infrared cycle, the switched off at its input as weU, such that it is completely infrared Ught emitter 302 is activated for one quarter cycle deactivated. This reduces noise from the switching and and is off for the remaining three-quarters cycle. In order to voltage to cunent conversion circuitry. In the present only receive one signal at a time, the emitters are cycled on embodiment, low noise voltage to cunent converters are and off alternatively, in sequence, with each only active for selected (e.g.. Op 27 Op Amps), and the feedback loop is 15 a quarter cycle per 625 Hz cycle and a quarter cycle configured to have a low pass filter to reduce noise. In the separating the active times. The Ught signal is attenuated present embodiment, the low pass filtering function of the (ampUtude modulated) by the pumping of blood through the voltage to cunent converter 328,329 has a corner frequency finger 310 (or other sample medium). The attenuated of just above 625 Hz. which is the switching speed for the (ampUtude modulated) signal is detected by the photodetecemitters, as further discussed below. Accordingly, the pre- 20 tor 320 at the 625 Hz carrier frequency for the red and fened driver circuit of FIG. lla, minimizes the noise of the infrared Ught. Because only a single photodetector is used, emitters 301, 302. the photodetector 320 receives both the red and infrared In general, the red and infrared Ught emitters 301, 302 signals to form a composite time division signal, each emits energy which is absorbed by the finger 310 and The composite time division signal is provided to the received by the photodetector 320. The photodetector 320 25 front analog signal conditioning circuitry 330. Additional produces an electrical signal which conesponds to the detaU regarding the front end analog signal conditioning intensity of the Ught energy striking the photodetector 320. circuitry 330 and the analog to digital converter circuit 332 The front end analog signal conditioning circuitry 330 is iUustrated in FIG. 12. As depicted in FIG. 12, the front end receives the intensity signals and filters and conditions these circuity 330 has a pieamplifiei 342. a high pass filter 344. an signals as further described below foi fiiithei piocessing. 30 amplifier 346, a programmable gain amplifier 348, and a low The resultant signals are provided to the analog-to-digital pass filter 350. The preamplifier 342 is a transunpedance conversion circuitry 332 which converts the analog signals amplifier that converts the composite cunent signal from the to digital signals for further processing by the digital signal photodetector 320 to a conesponding voltage signal, and processing system 334. The digital signal processing system ampUfies the signal. In the present embodunent the pream334 utiUzes the two signals in order to provide a what wiU 35 plifier has a predetermined gain to boost the signal ampUbe caUed herein a "saturation transform." It should be tude for ease of processing. In the present embodiment, the understood, that for parameters other than blood saturation source voltages for the pieamplifiei 342 are -15 VDC and monitoring, the saturation transform could be better termed +15 VDC. As wfll be understood, the attenuated signal as a concentration transform, in-vivo transform, or the like, contains a component representing ambient Ught as weU as depending on the desired parameter. The term saturation 40 the component representing the infrared or the red Ught as transform is used to describe an operation which converts the case may be in time. If there is Ught in the vidnity of the the sample data from time domain to saturation domain sensor 300 other than the red and infrared Ught, this ambient values as wiU be apparent from the discussion below. In the Ught is detected by the photodetector 320. Accordingly, the present embodiment, the output of the digital signal progain of the preamplifier is selected in order to prevent the cessing system 334 provides clean plethysmographic wave- 45 ambient Ught in the signal from saturating the preamplifier forms ofthe detected signals and provides values for oxygen under normal and reasonable operating conditions, saturation and pulse rate to the display 336. in the present embodunent, the preamplifier 342 cornIt should be understood that in different embodiments of prises an Analog Devices AD743JR OpAmp. This transimthe present invention, one or more of the outputs may be pedance amplifier is particularly advantageous in that it provided. The digital signal processing system 334 also 50 exhibits several desired features for the system described, provides control for driving the Ught emitteis 301,302 with such as: low equivalent input voltage noise, low equivalent an emittei cunent control signal on the emitter current input cunent noise, low input bias cunent, high gain bandcontrol output 337. This value is a digital value which is width product, low total harmonic distortion, high common converted by the digital-to-analog conversion circuit 338 mode rejection, high open loop gain, and a high power which provides a control signal to the emitter cunent drivers 55 supply rejection ratio. 340. The emitter cunent drivers 340 provide the appropriate The output of the preampUfier 342 couples as an input to cunent drive for the red emitter 301 and the infrared emitter the high pass filter 344. The output of the preamplifier also 302. Further detaU of the operation of the physiological provides a first input 346 to the analog to digital conveision monitoi foi pulse oximetry is explained below. circuit 332. In the present embodunent, the high pass filter In the present embodiment, the Ught emitters are driven 60 is a single-pole filter with a corner frequency of about Vi-l via the emitter cunent driver 340 to provide Ught transmisHz. However, the corner frequency is readily raised to about sion with digital modulation at 625 Hz. In the present 90 Hz in one embodunent. As wiU be understood, the 625 Hz embodunent. the Ught emitters 301, 302 are driven at a carrier frequency of the red and infrared signals is weU power level which provides an acceptable intensity for above a 90 Hz corneifrequency.The high-pass filtei 344 has detection by the detectoi and foi conditioning by the front 65 an output coupled as an mput to an amplifiei 346. In the end analog signal conditioning circuitry 330. Once this present embodiment the amplifier 346 comprises a unity energy level is determined for a given patient by the digital gam amplifier. However, the gain of the amplifier 346 is

5,632,272 37

38

adjustable by the variation of a single resistor. The gain of to digital converter 356 comprises a single-channel, deltathe amplifier 346 would be increased if the gain of the sigma converter. In the present embodiment, a Crystal preampUfier 342 is decreased to compensate for the effects Semiconductor CS5317-KS delta-sigma analog to digital of ambient Ught. converter is used. Such a converter is advantageous in that The output of the amplifier 346 provides an input to a 5 it is low cost, and exhibits low noise characteristics. More programmable gam amplifier 348. The programmable gam specificaUy, a delta-sigma converter consists of two major amplifier 348 also accepts a programming input from the portions, a noise modulator and a decunation fllter. The digital signal processing system 334 on a gam control signal selected converter uses a second order analog delta-sigma Une 343. The gain of the programmable gain amplifier 348 modulator to provide noise shaping. Noise shaping refers to is digitaUy programmable. The gain is adjusted dynamically 1° changing the noise spectrum from a flat response to a at initiaUzation or sensor placement for changes in the response where noise at the lower frequencies has been medium undei test from patient to patient. Foi example, the reduced by increasing noise at higher frequencies. The signal from different fingers differs somewhat. Therefore, a dedmation filter then cuts out the reshaped, higher fredynamicaUy adjustable amplifier is provided by the proquency noise to provide 16-bit performance at a lower grammable gain ampUfier 348 in order to obtam a signal 15 frequency. The present converter samples the data 128 times suitable for processmg. for every 16 bit data word that it produces. In this manner, the c o n v The programmable gain amplifier is also advantageous in f f P r ° v i d e s e x c \ e U e l l t .^ise rejection, dynamic ran e a n d low h a n an alternative embodiment in which the emitter drive cunent g . ™ ^ ^ ^ helP f c n t l c a l is held constant. In the present embodiment, the emitter measurement situations like low perfusion and electrocauer drive cunent is adjusted for each patient in order to obtain ^' the proper dynamic range at the input ofthe analog to digital In addition, by using a single-channel converter, there is conversion drcuit 332. However, changmg the emitter drive no need to tune two oi more channels to each othei. The cunent can alter the emitter wavelength, which in tum delta-sigma converter is also advantageous in that it exhibits affects the end result in oximetry calculations. Accordingly, noise shaping, for improved noise control. An exemplary it would be advantageous tofixthe emitter drive cunent for 25 analog to digital converter is a Crystal Semiconductor aU patients. In an alternative embodunent of the present CS5317; In the present embodunent, the second analog to mvention, the piogtammable gain ampUfier can be adjusted digital converter 356 samples the signal at a 20 Khz sample by the DSP in order to obtain a signal at the input to the rate. The output ofthe second analog to digital converter 356 analog to digital conversion drcuit which is properly within provides data samples at 20 Khz to the digital signal the dynamic range (+3 v to - 3 v in the present embodiment) 30 processing system 334 (FIG. 11). of the analog to digital conveision circuit 332. In this The digital signal processing system 334 is iUustrated in manner, the emitter drive cunent could be fixed for aU additional detaU in FIG. 13. In the present embodiment, the patients, eliminating the wavelength shift due to emitter digital signal processing system comprises a microcontroUer cunent drive changes. 360, a digital signal processor 362, a program memory 364, The output of the programmable gain ampUfier 348 35 a sample buffer 366, a data memoiy 368, a read only couples as an input to a low-pass filtei 350. Advantageously, memoiy 370 and communication registers 372. In the the low pass filter 350 is a single-pole filter with a corner present embodiment, the digital signal processor 362 is an frequency of approximately 10 Khz in the present embodiAnalog Devices AD 21020. In the present embodunent, the ment. This low pass filter provides anti-aUasing in the microcontroUer 360 comprises a Motorola 68HC05, with present embodiment. bmlt in program memory. In the present embodiment, the s a m l e b u f f e r 3 6 6 s a b u f f e r w h i c h a c c e t s t h e 2Q Khz The output of the low-pass filter 350 provides a second P i P s a m l e data from iile input 352 to the analog-to-digital conversion ciicuit 332. P " " k g t 0 digital conveision circuit 332 for FIG. 12 also depicts additional defect of the analog-toborage in the data memory 368. In the present digital conversion circuit. In the present embodiment, the 4 5 embodiment, the data memory 368 comprises 32 KWords analog-to-digital conversion cUcuit 332 comprises a first (WOTds being 40 bits in the present embodiment) of static r a n d o m access analog-to-digital converter 354 and a second analog-tomemory. digital converter 356. Advantageously, the first analog-toThe microcontroUer 360 is connected to the DSP 362 via digital converter 354 accepts mput from thefirstinput 346 a conventional JTAG Tap Une. The microcontroUer 360 to the analog-to-digital conversion circuit 332, and the 50 transmits the boot loader for the DSP 362 to the program second analog to digital converter 356 accepts input on the memory 364 via the Tap line, and then aUows the DSP 362 second input 352 to the analog-to-digital conversion cirto boot from the program memory 364. The boot loader in cuitry 332. program memory 364 then causes the transfer of the operIn one advantageous embodiment, the first analog-toating instructions for the DSP 362 from the read only digital converter 354 is a diagnostic analog-to-digital con- 55 m e m o r y 3 7 0 to the program memory 364. Advantageously, verter. The diagnostic task (performed by the digital signal the program memory 364 is a very high speed memory for processing system) is to read the output of the detector as the DSP 362. amplified by the preampUfier 342 m order to determine if the The microcontroUer 360 provides the emitter cunent signal is saturating the input to the high-passfilter344. In the control and gain control signals via the communications present embodiment, if the input to the high pass filter 344 60 register 372. becomes saturated, the front end analog signal conditioning FIGS. 14-20 depict functional block diagrams of the circuits 330 provides a '0' output. Alternatively, the first operations of the pulse oximeter 299 carried out by the analog-to-digital converter 354 remains unused. digital signal processing system 334. The signal processing The second analog-to-digital converter 356 accepts the functions described below are carried out by the DSP 362 in conditioned composite analog signal from the front end 65 the present embodiment with the microcontroUer 360 prosignal conditioning circuitry 330 and converts the signal to viding system management. In the present embodiment, the digital form. In the present embodiment, the second analog operation is software/firmware controUed. FIG. 14 depicts a

5,632,272 39

40

generalized functional block diagram for the operations which enhances noise immunity. The sum of the respective performed on the 20 Khz sample data entering the digital ambient Ught samples is then subtracted from the sum of the signal processing system 334. As Ulustrated in FIG. 14, a red and infrared samples, as represented in the subtraction demodulation, as represented in a demodulation module modules 438,440. The subtraction operation provides some 400, is first performed. Decimation, as represented in a 5 attenuation of the ambient Ught signal present in the data. In decimation module 402 is then performed on the resulting the present embodiment, it has been found that approxidata. Certain statistics are calculated, as represented in a mately 20 dB attenuation of the ambient Ught is provided by statistics module 404 and a saturation transform is the operations of the subtraction modules 438, 440. The perfoimed, as represented in a satuiation transform module resultant red and infrared sum values are divided by four, as 406. on the data resulting from the decimation operation. represented in the divide by four modules 442, 444. Each The data subjected to the statistics operations and the data resultant value provides one sample each of the red and subjected to the saturation transfoim operations are forinfrared signals at 625 Hz. warded to saturation operations, as represented by a saturait should be undeistood that the 625 Hz cairiei frequency tion calculation module 408 and pulse rate operations, as has been removed by the demodulation operation 400. The represented in a pulse rate calculation module 410. 15 625 Hz sample data at the output of the demodulation In general, the demodulation operation separates the red operation 400 is sample data without the carrier frequency, and infrared signals from the composite signal and removes In order to satisfy Nyquist sampUng requirements, less than the 625 Hz carrier frequency, leaving raw data points. The 20 Hz is needed (understanding that the human pulse is raw data points are provided at 625 Hz intervals to the about 25 to 250 beats per minute, or about 0.4 Hz-4 Hz), decimation operation which reduces the samples by an order 20 Accordingly, the 625 Hz resolution is reduced to 62.5 Hz in of 10 to samples at 62.5 Hz. The decimation operation also the decimation operation. provides some filtering on the samples. The resulting data is FIG. 16 Ulustrates the operations of the decunation modsubjected to statistics and to the saturation transform operaule 402. The red and infrared sample data is provided at 625 tions in order to calculate a saturation value which is very Hz to respective red and mfrared buffer/filters 450, 452. In tolerant to motion artifacts and other noise in the signal. The 25 the present embodiment, the red and mfrared buffer/filters saturation value is ascertained in the saturation calculation are 519 samples deep. Advantageously, the buffer filters 450, module 408, and a pulse rate and a clean plethysmographic 452 function as continuous first-in, first-out buffers. The 519 waveform is obtained thiough the pulse rate module 410. samples are subjected to low-pass filtering. Preferably, the Additional detaU regarding the various operations is prolow-pass filtering has a cutoff frequency of approximately vided in connection with FIGS. 15-21. 30 7.5 Hz with attenuation of approximately -110 dB. The FIG. 15 Ulustrates the operation of the demodulation buffer/filters 450,452 form a Finite Impulse Response (FIR) module 400. The modulated signal format is depicted in filter with coefficients for 519 taps. In order to reduce the FIG. 15. One fuh 625 Hz cycle of the composite signal is sample frequency by ten, the low-pass filter calculation is depicted in FIG. 15 with the first quarter cycle being the performed every ten samples, as represented in respective active red Ught plus ambient Ught signal, the second quarter 35 red and infrared decimation by 10 modules 454, 456. In cycle being an ambient Ught signal, the third quarter cycle other words, with the transfer of each new ten samples into bemg the active mfrared plus ambient Ught signal, and the the buffer/filters 450, 452, a new low pass filter calculation fourth quarter cycle being an ambient Ught signal. As is performed by multiplying the impulse response depicted in FIG. 15, with a 20 KHz sampling frequency, the (coefficients) by the 519 filter taps. Each filter calculation single fuU cycle at 625 Hz described above comprises 32 40 provides one output sample for respective red and infrared samples of 20 KHz data, eight samples relating to red plus output buffers 458, 460. In the present embodiment, the red ambient Ught, eight samples relating to ambient Ught, eight and infrared output buffers 458, 460 are also continuous samples relating to mfrared plus ambient Ught, and finaUy FIFO buffers that hold 570 samples of data. The 570 samples eight samples related to ambient Ught. provide respective infrared and red samples 01 packets (also Because the signal piocessing system 334 controls the 45 denoted "snapshot" herein) of samples. As depicted in FIG. activation of the Ught emitteis 300, 302, the entire system is 14. the output buffets provide sample data foi the statistics synchronous. The data is synchronously divided (and operation module 404, saturation transform module 406, and thereby demodulated) into four 8-sample packets, with a the pulse rate module 410. time division demultiplexing operation as represented in a FIG. 17 Ulustrates additional functional operation detaUs demultiplexing module 421. One eight-sample packet 422 50 of the statistics module 404. In summary, the statistics represents the red plus ambient Ught signal; a second eightmodule 404 provides first order oximetry calculations and sample packet 424 represents an ambient Ught signal; a thud RMS signal values foi the red and infrared channels. The eight-sample packet 426 represents the attenuated infrared statistics module also provides a cross-conelation output Ught plus ambient Ught signal; and a fourth eight-sample which indicates a cross-conelation between the red and packet 428 represents the ambient Ught signal. A select 55 infrared signals. signal synchronously controls the demultiplexing operation As represented in FIG. 17, the statistics operation accepts so as to divide the time-division multiplexed composite two packets of samples (e.g., 570 samples at 62.5 Hz in the signal at the input of the demultiplexer 421 into its four present embodiment) representing the attenuated infrared subparts. signals, with the carrier frequency removed. The a n ( l lsd A sum of the last four samples from each packet is then 60 respective packets for infrared and red signals are normalcalculated, as represented in the summing operations 430, ized with a log function, as represented in the Log modules 432,434.436 of FIG. 15. In the present embodiment the last 480, 482. The normalization is foUowed by removal of the four samples are used because a low pass filter in the analog DC portion of the signals, as represented in the DC Removal to digital converter 356 of the present embodiment has a modules 484,486. In the present embodiment, the DC settiing time. Thus, coUecting the last four samples from 65 removal involves ascertaining the DC value of the first one each 8-sample packet allows the previous signal to clear. of the samples (or the mean of the first several or the mean This summing operation provides an integration operation of an entire snapshot) from each of the respective red and

5,632,272 41

42

infrared snapshots, and removing this DC value from aU not qualify. Signals which satisfy the signal model wiU have samples in the respective packets. a conelation greater than the threshold. Once the DC signal is removed, the signals are subjected The red and mfrared 120-sample packets are also subto bandpass filtering, as represented in red and infrared jected to a second saturation operation and cross conelation Bandpass Filter modules 488,490. In the present 5 i n ^ e same manner as described above, except the 120 embodiment, with 570 samples in each packet, the bandpass samples are divided mto 5 equal bins of samples (i.e., 5 bins filters are configured with 301 taps to provide a FIR filter of 24 samples each). The RMS, ratio, saturation, and cross with a Unear phase response and Uttle or no distortion. In the conelation operations aie performed on a bin-by-bin basis, present embodiment, the bandpass filter has a pass band These operations are represented in the Divide Into Five from 34 beats/minute to 250 beats/minute. The 301 taps io Equal Bins modules 510, 512. the second red and infrared sUde over the 570 samples in order to obtain 270 filtered RMS modules 514. 516, the second RED-RMS/IR-RMS samples representing the filtered red signal and 270 filtered ratio module 518, the second saturation equation module samples representing the filtered infrared signal. In an ideal 520 and the second cross conelation module 522 in FIG. 17. case, the bandpass filters 488, 490 remove the DC in the mG M m u s t r a t e s additional detaU regarding the saturasignal. However, the DC removal operations 484,486 assist is ^ ^ ^ 0 ^ m o d u l e 4<)6 d e p i c t e d ^ m a 14 A s m u s . m DC removal m the present embodiment. tiated i n mG l s t h e s a t u r a t i o n transform module 406 After filtenng, the last 120 samples from each packet (of comprises a reference processor 530, a conelation canceler now 270 samples in the piesent embodunent) aie selected 5 3 ^ a master power curve module 554, and a bin power for further processing as represented in Select Last 120 curve module 533. The saturation transform module 406 can Samples modules 492, 494. The last 120 samples are 20 b e c o n . e l a t e d to FIG. 7a which has a reference processor 26 selected because, in the present embodiment, the fiist 150 a n d a conelation cancelei 27 and an integiatoi 29 to provide samples faU within the settiing time foi the Saturation a power curve for separate signal coefficients as depicted in Transfer module 406, which processes the same data J ^ Q 7 C T h e saturation transform module 406 obtains a packets, as further discussed below. saturation spectrum from the snapshots of data. In other Conventional saturation equation calculations are pei- 25 words, the saturation transform 406 provides information of formed on the red and infrared 120-sample packets. In the the saturation values present in the snapshots, present embodiment, the conventional saturation calculaAs d icted ^ ^ ls ^ r e f e r e n c e pr0 cessor 530 for tions are performed in two different ways. For one the saturation transform module 406 has a saturation equacalculation, the 120-sample packets are processed to obtam ^ m o d u l e 53^ a r e f e r e n c e g e n e r a t o r m o d u l e 5 3 4 < a D C theu overaU RMS value, as represented m the first red and ^ a b a n d p a s s me]. m o d u l e 5 3 8 . T h e removal module ^ infrared RMS modules 496,498. The resultant RMS values ^ 570.s red and ^ l e a c k e t s from t h e dedmation for red and infrared signals provide input values to a first o p e r a t i o n m p r o v i d e d t 0 ^ reference processor 530. In RED_RMS/IR_RMS ratio operation 500, which provides a d d i t i o n ) a plurality of possible saturation values (the "satuthe RMS red value to RMS infrared value ratio as an input ration ^ scan„ ) ^ provided as ^ 110 ^ saturation to a saturation equation module 502. As weU understood m r e f e r e nce processor 530. In the present embodiment, 117 the art, the ratio of the mtensity of red to infrared attenuated s a t u r a t i o n v a i u e s m provided as the saturation axis scan. In Ught as detected for known red and infrared wavelengths a p r e f e r r e d embodiment, the 117 saturation values range (typicaUy ^ 6 5 0 nm and ^ = 9 1 0 nm) relates to the uniformly from a blood oxygen saturation of 34.8 to 105.0. oxygen saturation ofthe patient. Accordingly, the saturation A c c o r d i n g i y , in the present embodiment, the 117 saturation equation module 502 represents a conventional look-up « values provide an axis scan for the reference processor 530 table or the like which, for predetermined ratios, provides w h i c h g e n e r a t e s a reference signal for use by the conelation known saturation values at its output 504. The red and ^ o t h e r woxdSy ^ r e f e r e n c e processor is canceler ^ infrared RMS values are also provided as outputs of the p r o v i d e d w i t h each of the saturation values, and a resultant statistics operations module 404. reference signal is generated conesponding to the saturation In addition to the conventional saturation operation 502, va iue. The conelation canceler is formed by a joint process the 120-sample packets are subjected to a cross-conelation estimator 550 and a low pass filtei 552 in the piesent operation as represented in a first cross-conelation module embodiment 506 The first cross-correlation module 506 determines if ^ s c a n v a l u e s CO ui d be It s h o u l d b e u n d e r s t o o d ^ good conelation exists between the infrared and red signals. chosen t0 i d e ^ h e r o r l o w e r r e S o l u t i o n tban n7 scan This cross conelation is advantageous for detecting defecvalues T h e scan values could ^ b e non . U niformly spaced. tive or otheiwise malfunctioning detectois. The cross coi„ .„ _ , . . „„ , „ . ^ ,. , .. . , , - A t t u tu • 1 As illustrated in FIG. 18, the saturation equation module relation is also advantageous m detecting when the signal _„. , . . , ^ . . 5 3 2 acce ts model (Le.. the model of Equations (l)-(3)) is satisfied. If P * ? satuf,atl0n ^ s c a n v T a l u e s a s ™ ^ f ^ d , t- u 1 u t ti t u 1 tu provides a ratio 1 ' as an output. In companson to the r conelation becomes too low between the two channels, the ... . ' l ~ „ _ _ ,. . . ,. ,, , , ,,. *_ * T A t A t • tu- tu 55 general discussion of FIG. 7a-7c. this ratio r„ conesponds signal model is not met. In oidei to determine this, the •" f ^ , ,. . , .. , , ? % __ ... , .. . t A u tu to the pluraUty of scan value discussed above m general. The } normalized cross conelation can be computed by the crossK . . .. , • „ „ , . , itJ 1 rA* * u u t c A t r\ u saturation equation simply provides a known ratio r (red/ conelation module 506 for each snapshot of data. One such . . .. M ,. r / f. ^ . . . v, , . f .. . followsinfrared) conesponding to the saturation value received as an input. £ 5lS2 60 The ratio "r„" is provided as an input to the reference ^1 =• generator 534, as are the red and infrared sample packets. N z Si ZS1 rpjjg r e f e r e nce generator 534 multipUes either the red or infrared samples by the ratio "r„" and subtracts the value If the cross conelation is too low, the oximeter 299 from the infrared 01redsamples, respectively. For instance, provides a warning (e.g., audible, visual, etc.) to the opera- 65 in the present embodiment, the reference generator 534 tor. In the present embodiment, if a selected snapshot yields multiplies the red samples by the ratio "r„" and subtracts this a normalized conelation of less than 0.75, the snapshot does value from the infrared samples. The resulting values

5,632,272 43

44

become the output of the reference generator 534. This tions. The number of ceUs input 545 to the joint process operation is completed for each ofthe saturation scan values estimator 550 configures the number of ceUs for the joint (e.g., 117 possible values in the present embodiment). process estimator. In the present embodiment, the number of Accordingly, the resultant data can be described as 117 ceUs for the saturation transform operation 406 is six. As reference signal vectors of 570 data points each, hereinafter 5 weU understood in the art, for each sine wave, the joint refened to as the reference signal vectors. This data can be process estimator requires two ceUs. If there are two sine stored in an array or the like. waves in the 35-250 beats/minute range, six ceUs aUows for In other words, assuming that the red and infrared sample the two heart beat sine waves and one noise sine wave, packets represent the red S^Xt) and mfrared S ^ t ) measured The joint process estimator 550 subjects the first mput signals which have primaiy s(t) and secondaiy n(t) signal io vector on the first input 542 to a conelation canceUation portions, the output of the reference generator becomes the based upon each of the pluraUty of reference signal vectors secondary reference signal nXt), which compUes with the provided in the second input 540 to the conelation canceler signal model defined above, as foUows: 531 (aU 117 reference vectors in sequence in the present embodunent). The conelation cancellation results in a single nxtys^ty-rj^t) 15 o u t p u t v e c t o r f o r e a c h o f t h e 1 1 7 r e f e r e n C e vectors. Each In the present embodiment, the reference signal vectors output vector represents the information that the first input and the infrared signal are provided as input to the DC vector and the conesponding reference signal vector do not removal module 536 of the reference processor 530. The DC have in common. The resulting output vectors are provided removal module 536, like the DC removal modules 484,486 as an output to the joint process estimator, and subjected to in the statistics module 404. ascertains the DC value of the 20 the low pass filter module 552. In the present embodiment, fiist of the samples foi the respective inputs (01 mean of the the low pass filtei 552 comprises a FIR filter with 25 taps and first several or aU samples in a packet) and subtracts the with a comer frequency of 10 Hz with the sampUng frerespective DC baseUne from the sample values. The resultquency of 62.5 Hz (i.e., at the decimation frequency), ing sample values are subjected to a bandpass filter 538. The joint process estimator 550 of the present embodiThe bandpass filter 538 of the reference processor 530 25 ment has a settiing time of 150 data points. Therefore, the perfoims the same type of filtering as the bandpass filters last 120 data points from each 270 point output vector are 488,490 of the statistics module 404. Accordingly, each set used for further processing. In the present embodiment, the of 570 samples subjected to bandpass filtering results in 270 output vectois aie further processed together as a whole, and remaining samples. The resulting data at a first output 542 of are divided into a pluraUty of bins of equal number of data the bandpass filter 538 is one vector of 270 samples 30 points. As depicted in FIG. 18, the output vectors are (representing the filtered infrared signal in the present provided to a master power curve module 554 and to a embodiment). The resulting data at a second output 540 of Divide into five Equal Bins module 556. The Divide into the bandpass filter 538, therefore, is 117 reference signal Five Equal Bins module 556 divides each of the output vectors of 270 data points each, conesponding to each ofthe vectors into five bins of equal numbei of data points (e.g., satuiation axis scan values provided to the saturation refer- 35 with 120 data points per vector, each bin has 24 data points), ence processor 530. Each bin is then provided to the Bin Power Curves module It should be undeistood that the red and infraied sample 558. packets may be switched in their use in the reference The Master Powei Cuive module 554 performs a saturaprocessor 530. In addition, it should be understood that the tion transform as foUows: for each output vector, the sum of DC removal module 536 and the bandpass filter module 538 40 the squares of the data points is ascertained. This provides a can be executed prior to input of the data to the reference sum of squares value conesponding to each output vector processor 530 because the calculations performed in the (each output vector conesponding to one of the saturation reference processor are Unear. This results in a significant scan values). These values provide the basis for a master processing economy. power curve 555, as further represented in FIG. 22. The The outputs of the reference processor 530 provide first 45 horizontal axis of the power curve represents the saturation and second inputs to a joint process estimator 550 of the type axis scan values and the vertical axis represents the sum of described above with reference to FIG. 8. The first input to squares value (or output energy) for each output vector. In the joint process estunator 550 is the 270-sample packet other words, as depicted in FIG. 22, each of the sum of representing the infrared signal in the present embodiment. squares could be plotted with the magnitude of the sum of This signal contains primary and secondary signal portions. 50 squares value plotted on the vertical "energy output" axis at The second input to the joint process estimator is the 117 the point on the horizontal axis of the conesponding satureference signal vectors of 270 samples each. ration scan value which generated that output vector. This The joint process estimator also receives a lambda input results in a master power curve 558, an example of which is 543, a minimum enor input 544 and a number of ceUs depicted in FIG. 22. This provides a saturation transform in configuration input 545. These parameters are weU under- 55 which the spectral content of the attenuated energy is stood in the art. The lambda parameter is often caUed the examined by looking at every possible saturation value and "forgetting parameter" for a joint process estimator. The examining the output value for the assumed saturation value, lambda input 543 provides control for the rate of canceUaAs wiU be understood, where the first and second inputs to tion for the joint process estimator. In the present the conelation canceler 531 are mostly conelated, the sum embodiment, lambda is set to a low value such as 0.8. 60 of squares for the conesponding output vector of the corBecause statistics of the signal are non-stationaiy, a low relation cancelei 531 wiU be veiy low. Conversely, where value improves tracking. The minimum enor input 544 the conelation between the first and second inputs to the provides an initialization parameter (conventionaUy known conelation canceler 531 are not significantly conelated, the as the "initiaUzation value") for the joint process estimator sum of squares of the output vector will be high. 550. In the present embodiment, the minimum enor value is 65 Accordingly, where the spectral content of the reference 10 -6 . This initialization parameter prevents the joint process signal and the first input to the conelation canceler are made estimator 500 from dividing by zero upon initial calculaup mostly of physiological (e.g., movement of venous blood

5,632,272 45

46

due to respuation) and nonphysiological (e.g., motion induced) noise, the output energy wiU be low. Where the spectral content of the reference signal and the first input to the conelation canceler are not conelated, the output energy wiU be much higher. A conesponding transform is completed by the Bin Power Curves module 558, except a saturation transform power curve is generated for each bin. The resulting power curves are provided as the outputs of the saturation transform module 406. In general, in accordance with the signal model of the present invention, there wiU be two peaks in the power curves, as depicted in FIG. 22. One peak conesponds to the arterial oxygen satuiation of the blood, and one peak coiresponds to the venous oxygen concentration of the blood. With reference to the signal model of the present invention, the peak conesponding to the highest saturation value (not necessarily the peak with the greatest magnitude) conesponds to the proportionaUty coefficient r a . In other words, the proportionaUty coefficient ia conesponds to the red/ infrared ratio which wiU be measured for the arterial saturation. SimUarly, peak that conesponds to the lowest saturation value (not necessarily the peak with the lowest magnitude) wiU generaUy conespond to the venous oxygen saturation, which conesponds to the proportionaUty coeffident rv in the signal model of the present invention. Therefore, the proportionaUty coefficient r,, wiU be a red/ infrared ratio conesponding to the venous oxygen saturation. In order to obtain arterial oxygen saturation, the peak in the power curves conesponding to the highest saturation value could be selected. However, to improve confidence in the value, further processing is completed. FIG. 19 Ulustrates the operation ofthe saturation calculation module 408 based upon the output of the saturation transform module 406 and the output of the statistics module 404. As depicted in FIG. 19, the bin power curves and the bin statistics are provided to the saturation calculation module 408. In the present embodiment, the master power curves are not provided to the saturation module 408 but can be displayed for a visual check on system operation. The bin statistics contain the red and infrared RMS values, the seed saturation value, and a value representing the cross-conelation between the red and infrared signals from the statistics module 404. The saturation calculation module 408 first determines a pluraUty of bin attributes as represented by the Compute Bin Attributes module 560. The Compute Bin Attributes module 560 coUects a data bin from the infonnation from the bin power curves and the infonnation from the bin statistics. In the present embodiment, this operation involves plaring the saturation value of the peak from each power curve conesponding to the highest saturation value in the data bin. In the present embodiment, the selection of the highest peak is performed by first computing the first derivative of the power curve in question by convolving the power curve with a smoothing differentiator filter function. In the present embodiment, the smoothing diffeientiatoi filter function (using a FIR filter) has the foUowing coefficients: 0.014964670230367 0.098294046682706 0 204468276324813 „' 1 7 1 S 9 , f i . 9 . 1 S 1 » 2. 826 4 5.704485606695227 0.000000000000000 -5.704482606695227 -2.717182664241813

-0.204468276324813 -0.098294046682706 -0.014964670230367 This filter performs the differentiation and smoothing. Next, each point in the original power curve in question is evaluated and determined to be a possible peak if the foUowing conditions are met: (1) the point is at least 2% of the maximum value in the power curve; (2) the value of the first derivative changes from greater than zero to less than or equal to zero. For each point that is found to be a possible peak, the neighboring points are examined andthe largest of the three points is considered to be the true peak, The peak width foi these selected peaks is also calculated, The peak width of a powei curve in question is computed by summing aU the points in the power curve and subtracting the product of the minimum value in the power curve and the number of points in the powei curve. In the present embodiment, the peak width calculation is appUed to each of the bin power curves. The maximum value is selected as the peak width. In addition, the infrared RMS value from the entire snapshot, the red RMS value, the seed saturation value for each bin, and the cross conelation between the red and infrared signals from the statistics module 404 are also placed in the data bin. The attributes are then used to determine whether the data bin consists of acceptable data, as represented in a Bin Qualifying Logic module 562. If the conelation between the red and infrared signals is too low, the bin is discarded. If the saturation value of the selected peak for a given bin is lower than the seed saturation for the same bin, the peak is replaced with the seed saturation value. If either red or infrared RMS value is below a very smaU threshold, the bins are aU discarded, and no saturation value is provided, because the measured signals are considered to be too smaU to obtain meaningful data. If no bins contain acceptable data, the exception handling module 563 provides a message to the display 336 that the data is enoneous. If some bins qualify, those bins that qualify as having acceptable data are selected, and those that do not qualify are replaced with the average of the bins that are accepted. Each bin is given a time stamp in oidei to maintain the time sequence. A voter operation 565 examines each of the bins and selects the three highest saturation values. These values are forwarded to a cUp and smooth operation 566. The cUp and smooth operation 566 basicaUy perfoims averaging with a low pass filter. The low pass filter provides adjustable smoothing as selected by a Select Smoothing FUter module 568. The Select Smoothing FUter module 568 perfoims its operation based upon a confidence determination performed by a High Confidence Test module 570. The high confidence test is an examination of the peak width for the bin power curves. The width of the peaks provides some indication of motion by the patient—wider peaks indicating motion. Therefore, if the peaks are wide, the smoothing filter is slowed down. If peaks are nanow, the smoothing filter speed is increased. Accordingly, the smoothing filter 566 is adjusted based on the confidence level. The output ofthe cUp and smooth module 566 provides the oxygen saturation values in accordance with the present invention. ^ t ^ e Presently prefened embodiment, the cUp and smooth filter 566 takes each new saturation value and compares it to the cunent saturation value. If the magnitude difference is less than 16 (percent oxygen saturation) of ^ then the value is pass. Otherwise, if the new saturation value is less than the filtered saturation value, the new saturation value is changed to 16 less than the filtered saturation value.

5

10

15

20

25

30

35

40

45

so

55

60

65

5,632,272 47

48

If the new satuiation value is greater than the filtered saturation value, then the new saturation value is changed to 16 more than the filtered saturation value. During high confidence (no motion), the smoothing filter is a simple one-pole or exponential smoothing filter which is 5 computed as foUows: y(n)=o.6*x(nyi0.4*y(n-i)

module 592. The slew rate limiting module 592 connects to the output filter 594. The output filter 594 also receives an input from the output filter coefficient module 596. The output filter 594 provides the filtered heart rate for the display 336 (FIG. 11). In the case of no motion, one of the signals (the infrared signal in the present embodiment) is subjected to DC removal and bandpass filtering as represented in the DC removal and bandpass filter module 578. The DC removal

. . where x(n) is the dipped new satuiation value, and y(n) is 10 a n d bandpass mex module 5 7 8 pmvide t h e s a m e mtelilig a s the filtered satuiation value. , „ „ , . „ . the DC removal and bandpass filtei modules 5536. 538. Dunng motion condition, a thiee-pole IIR (infinite Bwhl n o m o t i o n conditions, the filtered infraied signal is impulse response) filtei is used. Its charactenstics are con^ estimation module 586. ided t0 the first troUed by three tune constants t a . tb, and tc with values of ^ the n t embodiment the spectral estimation com0.985, 0.900, and 0.94 respectively. The coefficients for a 15 ^ ^ a c b ^ z ^ t h a t ro vi des a frequency specduect form I, ER filter are computed from these time trum of heart rate information. The ChiipZ transform is used constants usmg the foUowmg relationships: r a t h e r t h a n a COnVentional Fouriei Transfoim because a frequency range for the desired output can be designated in a0=0 a Chirp Z tiansform. Accoidingly, in the piesent ^k+fcXfo+y 20 embodiment, a frequency spectrum of the heart rate is provided between 30 and 250 beats/minute. In the present «2=(-<»XO('«+'fKO(f«)) embodiment, the frequency spectrum is provided to a spectrum analysis module 590 wliich selects the first harmonic from the spectrum as the pulse rate. UsuaUy, the first fco=i-'i-(y('o+(fc)('i,)) 25 harmonic is the peak in the frequency spectrum that has the greatest magnitude and represents the pulse rate. However, i- (WcX'o- ) i n certain conditions, the second or third harmonic can exhibit the greater magnitude. With this understanding, in b _/, v,«, + ( t •)(•, yjt ^)/, y_t -, order to select the first harmonic, the first peak that has an FIG. 20 and 21 Ulustrate the pulse rate module 410 (FIG. 30 ampUtude of at least 1/20 th of the largest peak in the 14) in greater detaU. As iUustrated in FIG. 20. the heart rate spectrum is selected). This minimizes the possibUity of module 410 has a transient removal and bandpass filtei Selecting as the heart rate a peak in the Chirp Z transfoim module 578, a motion artifact suppression module 580, a caused by noise. saturation equation module 582. a motion status module In the case of motion, a motion artifact suppression is 584, first and second spectral estimation modules 586, 588, 35 completed on the snapshot with the motion artifact suppresa spectrum analysis module 590, a slew rate limiting module sion module 580. The motion artifact suppression module 592, an output filter 594, and an output filter coefficient 580 is depicted in greater detaU in FIG. 21. As can be seen module 596. in FIG. 21. the motion artifact suppression module 580 is As further depicted in FIG. 20, the heart rate module nearly identical to the saturation transform module 406 accepts the infrared and red 570-sample snapshots from the 40 (FIG. 18). Accordingly, the motion artifact suppression output of the decimation module 402. The heart rate module module has a motion artifact reference processor 570 and a 410 further accepts the saturation value which is output from motion artifact conelation canceler 571. the saturation calculation module 408. In addition, the The motion artifact reference processor 570 is the same as maximum peak width as calculated by the confidence test the reference processoi 530 of the saturation transform module 570 (same as peak width calculation described 45 module 406. However, the reference processor 570 utilizes above) is also provided as an input to the heart rate module the satuiation value from the saturation module 408, rather 410. The infrared and red sample packets, the satuiation than completing an entire satuiation transfoim with the 117 value and the output of the motion status module aie satuiation scan values. The reference processor 570, provided to the motion artifact suppression module 580. therefore, has a saturation equation module 581, a reference The average peak width value provides an input to a 50 generator 582. a DC removal module 583, and a bandpass motion status module 584. In the present embodiment, if the filter module 585. These modules are the same as conepeaks are wide, this is taken as an indication of motion. If sponding modules in the saturation transform reference motion is not detected, spectral estimation on the signals is processor 530. In the present embodiment, the saturation carried out directly without motion artifact suppression. equation module 581 receives the arterial saturation value Li the case of motion, motion artifacts are suppressed 55 from the saturation calculation module 408 rather than doing using the motion artifact suppression module 580. The a saturation axis scan as in the saturation transform module motion artifact suppression module 580 is nearly identical to 406. This is because the arterial saturation has been selected, the saturation transform module 406. The motion artifact and there is no need to perform an axis scan. Accordingly, suppression module 580 provides an output which connects the output of the saturation equation module 581 coneas an input to the second spectral estimation module 588. 60 sponds to the proportionaUty constant ia (i.e., the expected The first and second spectral estimation modules 586, 588 red to infrared ratio for the arterial saturation value), have outputs which provide inputs to the spectrum analysis Otherwise, the reference processoi 570 performs the same module 590. The spectrum analysis module 590 also function as the reference processor 530 of the saturation receives an input which is the output of the motion status transform module 406. module 584. The output of the spectrum analysis module 65 The motion artifact conelation cancelei 571 is also simi590 is the initial heart rate determination of the heart rate lar to the saturation transform conelation canceler 531 (FIG. module 410 and is provided as input to the slew rate limiting 18). However, the motion artifact suppression conelation

5,632,272 49 canceler 571 uses a sUghtly different motion artifact joint process estimator 572. Accordingly, the motion artifact suppression conelation canceler 571 has a joint process estimator 572 and a low-pass filter 573. The motion artifact joint process estimator 572 differs from the saturation transform joint process estimator 550 in that there are a different number of cells (between 6 and 10 in the present embodiment), as selected by the Number of CeUs input 574, in that the forgetting parameter differs (0.98 in the present embodiment), and in that the time delay due to adaptation diffeis. The low-pass filter 573 is the same as the low pass filter 552 of the saturation transform conelation canceler 531. Because only one satuiation value is provided to the reference processoi, only one output vector of 270 samples results at the output of the motion artifact suppression conelation canceler 571 for each input packet of 570 samples. In the present embodiment, where the infrared wavelength is provided as a first input to the conelation canceler, the output of the conelation canceler 571 provides a clean infrared waveform. It should be understood that, as described above, the infrared and red wavelength signals could be switched such that a clean red waveform is provided at the output of the motion artifact suppression correlation canceler 571. The output of the conelation canceler 571 is a clean waveform because the actual saturation value of the patient is known which aUows the reference processor 570 to generate a noise reference for inputting to the conelation canceler 571 as the reference signal. The clean waveform at the output of the motion artifact suppression module 580 is a clean plethysmograph waveform which can be forwarded to the display 336. As described above, an alternative joint process estimator uses the QRD least squares lattice approach (FIG. 8a, 9a and 10a). Accordingly, the joint process estimator 573 (as weU as the joint process estimator 550) could be replaced with a joint process estimator executing the QRD least squares lattice operation. FIG. 21a depicts an alternative embodunent of the motion artifact suppression module with a joint process estimator 572a replacing the joint process estimatoi 572. The joint piocess estimatoi 572a comprises a QRD least squares lattice system as in FIG. 10a. In accordance with this embodiment, different initiaUzation parameters are used as necessary for the QRD algorithm. The initialization parameters are referenced in FIG. 21a as "Number of CeUs," "Lambda," "MinSumEn," "Gamslnit," and "SumEnlnit" Numbei of CeUs and Lambda conespond to like parameters in the joint process estimator 572. Gamslnit conesponds to the y initialization variable for aU stages except the zero order stage, which as set forth in the QRD equations above is initialized to ' 1 ' . SummEnlnit provides the 5 initialization paiametei referenced above in the QRD equations. In order to avoid overflow, the larger of the actual calculated denominator in each division in the QRD equations and MinSumEn is used. In the piesent embodiment, the prefened initiaUzation parameters are as foUows: Number of CeUs=6 T hda-0 8 _20 MinSumErr=10 Gainslnit=10_ SumEnInit=10~6. frequency The clean waveform output from the motion artifact suppression module 580 also provides an input to the second spectral estimation module 588. The second spectral estimation module 588 perfoims the same Chirp Z transfoim as

50 the first spectral estimation module 586. In the case of no motion, the output from the first spectral estimation module 586 is provided to the spectrum analysis module 586; in the case of motion, the output from the second spectral estima5 tion module 588 is provided to a spectrum analysis module 590. The spectrum analysis module 590 examines the frequency spectrum from the appropriate spectral estimation module to determine the pulse rate. In the case of motion, the spectrum analysis module 590 selects the peak in the specio trum with the highest ampUtude, because the motion artifact suppression module 580 attenuates aU other frequencies to a value below the actual heart rate peak. In the case of no motion, the spectrum analysis module selects the first harmonic in the spectrum as the heart rate as described above, 15 The output of the spectrum analysis module 590 provides the raw heart rate as an input to the slew rate limiting module 592, which provides an input to an output filter 594. In the present embodiment, the slew rate limiting module 592 prevents changes greater that 20 beats/minute per 2 second 20 interval. The output filter 594 comprises an exponential smoothing filter simUar to the exponential smoothing filter described above with respect to the cUp and smooth filter 566. The output filter is controUed via an output filter coeffident 25 module 596. If motion is large, this filter is slowed down, if there is Uttle or no motion, this filter can sample much faster and stUl maintain a clean value. The output from the output filter 594 is the pulse of the patient, which is advantageously provided to the display 336. 30 Alternative To Saturation Transform Module—Bank Of FUters An alternative to the saturation transform of the saturation transform module 406 can be implemented with a bank of filters as depicted in FIG. 23. As seen in FIG. 23, two banks 35 of filters, a first filter bank 600 and a second filter bank 602 are provided. The first filter bank 600 receives a first measured signal Sxi(t) (the infrared signal samples in the present embodiment) on a conesponding first filter bank input 604, and the second filter bank 602 receives a second 40 measured signal 8^(1) (the red samples in the present embodiment) on a conesponding second filtei bank input 606. In a prefened embodiment, the first and second filter banks utiUze static recursive polyphase bandpass filters with fixed center frequencies and comer frequencies. Recursive 45 polyphase filters are described in an article Hams, et. al. "Digital Signal Processing With Efficient Polyphase Recursive AU-Pass fllters" attached hereto as Appendix A. Howevei, adaptive implementations aie also possible. In the present implementation, the recursire polyphase bandpass so filter elements are each designed to include a specific center frequency and bandwidth. There are N filter elements in each filtei bank. Each ofthe filter elements in the first filter bank 600 have a matching (i.e., same center frequency and bandwidth) filter element in 55 the second filter bank 602. The center frequencies and the coinei frequendes of N elements are each designed to occupy N frequency ranges, 0 to Fj, Fi-F2, F 2 -F3, F3-F4.. • FAT-I-F^ as shown in FIG. 23. ^ s l l o u id be understood that the number of filter elements 60 can range from 1 to infinity. However, in the present embodiment, there are approximately 120 separate filtei elements with center frequencies spread evenly across a range of 25 beats/minute-250 beats/minute. The outputs of the filters contain information about the 65 primary and secondary signals for the first and second measured signals (red and infrared in the present example) at the specified frequendes. The outputs for each pair of

5,632,272 51

52

matching filters (one in the first filter bank 600 and one in the s^r^+rj^ (red) (92) second filter bank 602) are provided to saturation determination modules 610. H G . 2 3 depicts only one saturation N o t e 1bat g ^ a n d SIR aie used in the model of equations determination module 610 for ease of Ulustration. However. (89)-(92). This is because the discussion below is particua saturation determination module can be provided for each 5 l a r l y directed to blood oximetry. Sred and SIR conespond to matched pair of filter elements for paraUel processing. Each S i a nd S 2 in the preceding text, and the discussion that saturation determination module has aratio module 616 and fonov,s conld ^ generaUzed for any measure signal Si and a saturation equation module 618. g The ratio module 616 forms a ratio of the second output to the first output. For instance, in the present example, a io As explained above, detennining ia and i„ (which coneratio of each red RMS value to each conesponding infrared spond to arterial and venous blood oxygen saturation via a RMS value (Red/IR) is completed in the ratio module 616. saturation equation) can be accompUshed using the saturaThe output of the ratio module 616 provides an mput to the tion transform described above doing a scan of many saturation equation module 618 which references a conepossible coefficients. Another method to obtain r a and r„ sponding saturation value for the input ratio. The output of is based on red and infraied data is to look foi r a and r„ which the saturation equation modules 618 are coUected (as repminimize the conelation between s* and n^ assuming s* is resented in the histogram module 620) for each of the at least somewhat (and preferably substantiaUy) unconematched filter pairs. However, the data coUected is initiaUy lated with % (where k = l or 2). These values can be found a function of frequency and saturation. In order to form a by minimizing the foUowing statistical calculation function saturation transform curve simUar to the curve depicted in 20 for k=2: FIG. 22, a histogram or the like is generated as in FIG. 24. The horizontal axis represents the saturation value, and the Correlation^,^) = E «(S«
,

.

.

.

.

.

.

2



~ r —r

Wred — ry HIR)

be chosen as the artenal saturation and the entry in the " '' histogram ' b ' with the lowest saturation value would be i n chosen as the venous or background saturation. 55 i=-p—f^(-Srta+rcSm) ALTERNATIVE DETERMINAnON OF COEFFICIENTS rfl AND r„ Preferably, the conelation of equation (93) is enhanced with As explained above, in accoidance with the piesent a usei specified window function as foUows: invention, primaiy and secondaiy signal portions, particu- 60 laily for pulse oximetry, can be modeled as foUows: N (93a) S ^ ^ + n , (red)

(89)

SiR=s2+'t2 (infrared)

(90)

Correlationtsvs) = I £ WiS2(SndeSlRl,r0,rv)n2(S„d,ARerc,rv)\ '= 1

The Blackman Window is the presently prefened embodi^ , - V J and nj-zvnj (91) 65 m e n t I t s h o u l d b e understood that there aie many additional Substituting Equation (91) into Equation (89) provides the functions which minimize the conelation between signal foUowing: and noise. The function above is simply one. Thus,

5,632,272 53 Correlationfe, n2 )

=

54

I N P w: Z I !=1 L (r 0 -/- v ) 2

(93b) (S„d.'

ENERGY^)

^ / R , ) (-Sreol, - '•oS/J^.)

(^-rv) 2 W

I

N 1 £(S^-rA)2 L _ ^j.(S^-r&tf 1 (ra-r,) !=1 1

=



- I (Srei.)2H'i + (r0 + i=l AT

^v2 . 1 (SfRi.> 10 11 -—[Ri-2r + rrlR2] [Ri-2rvRu vR12 v%] (r. - r,,)2

=

In oidei to implement the minimization on a pluraUty of discrete data points, the sum of the squares of the red sample points, the sum of the squares of the infrared sample points, and the sum of the product of the red times the infrared sample points are first calculated (including the window function, w,):

is ^ e n e r g y o f ^ r e d signal< ^ is ^ e n e r g y o f ^ ^ s i g n a l a n d R i ^ i s t h e correlation between the red a n d infixed signals. T h e comlation b e t w e e n s 2 and n 2 is given by

±c

20

1

Correlation^,^) ~

n=,I.(Sm.)2Wi 25

IRR = Y (Sm) (Sred) w "=1 ' 30

These values are used in the conelation equation (93b). Thus, the conelation equation becomes an equation in terms of two variables, r a and rv. To obtam ia and rv, an exhaustive scan is executed for a good cross-section of possible values for ia and r,, (e.g., 20-50 values each conespondmg to saturation values ranging from 30-105). The minimum of the conelation function is then selected and the values of r0 and rv which resulted in the minimum are chosen as ia and

Once ia and rv have been obtained, arterial oxygen saturation and venous oxygen saturation can be determined by provided ia and r„ to a saturation equation, such as the saturation equation 502 of the statistics module 404 which provides an oxygen saturation value conesponding to the ratios i„ and rv.

(96)

where

N RR= J (Sr,d)hvi i=l

r . "

(95)

2r„ E ( S ^ / R . ) +

1

rv) 2 SiRSredm+ . 2 (SfflJ2^,1=1 !=1 I

(94)

=

2

(ra-rv) 1 (r„-rv)2



N (97) 2 (Snj,- rA-,) (Sred, + ^SIR) 1 i=l ••[-Rl + (r<, +

rr)Ru-rScR2]

As explained above, the constraint is that s^ and n^. (k =2 for f16 P rese iit example) are unconelated. This "deconelation constraint" is obtained by setting the conelation of Equation (97) to zero as foUows: -RtHra+r^yRa-rj-^O

(98)

^ ofter words, the goal is to maxunize equation (94) under the constramt of equation (98). ^ order to obtam the goal, a cost function is defined (e.g., a Lagrangian optimization in the present embodunent) as foUows: KWvM) = 40

i =--— [Kl - 2rvRu + r?Ri + H(-fil + (rc + r,)Ri2 (rc - nO2

(") wRz)}

where |i is the Lagrange multipUer. Finding the value of ia, ^ and n that solve the cost function can be accompUshed usto a 8 constrained optimization method such as described in Luenberger, Linear & Nonlinear Programming, AddisonWesley, 2d Ed., 1984. Along the same lines, if we assume that the red and 50 infrared signals S red and SIR are non-static, the functions R^ R 2 and R 12 defined above are time dependent. Accordingly, with two equations, two unknowns can be obtained by expressing the deconelation constraint set forth in equation ( 9 8 ) a t ^ m ^ v & ^ ^ T h e deconelation constraint can 55 be expressed at two different times, t, and t* as foUows: -R(t) + (r +r)R (t)-rrvR(t) = o (ioo)

45

In a further implementation to obtain ia and r,,, the same signal model set forth above is again used. In order to determiner andr in accordance with this implementation, the energy in the signal s2 is maximized under the constraint that s2 is unconelated with n2. Again, this implementation is based upon minimizing the conelation between s and n and on the signal model of the present mvention where the signal s relates to the arterial pulse and the signal n is the noise -Riih) + (re + rv)R12(t2) - rar^t2) = o (ioi) (containing infonnation on the venous blood, as weU as 60 B e c a u s e e q u a t i o n s ^ (101) m ^ . ^ ^ ^ r ^ motion artifacts and other noise); ta is the ratio (RED/IR) ^ a c h a n g e 0 f valiailes ^ ^ ^ u s e of Unear techniques related to arterial saturation and rv is the ratio (RED/IR) t 0 S oi v e these two equations. Accordingly, with x=ra+rv; related to venous saturation. Accordingly, in this implemeny=r j equations (100) and (101) become tation of the present mvention, ia and rv are determined such that the energy of the signal s 2 is maximized where s 2 and 65 *i2('i)»-*z('i)y=Si('i) (i02) n2 are unconelated. The energy of the signal s2 is given by the foUowing equation: fi^fe^^fe^fiife) (i03)

5,632,272 55

56

These equation (102) and (103) can be solved for x and y. Then, solving for ia and iv from the changes of variables equations provides the foUowing: rv + 2. rv

= x => r2 -Xrv+y=o

(104)

Solving equation (104) results in two values for r,,. In the present embodiment the iv value that results in x ^ r ^ O is selected. If both values of r„ result in x2-r1,y>0, the r, that maximizes the energy of s2 (Energy(s2)) at t 2 is selected, rv is then substituted into the equations above to obtain r j Altematively ra can be found directly in the same manner iv A . • A was aetermmea Alternative To Saturation Transfoim-Complex FFT The blood oxygen saturation, pulse rate and a clean plethysmographic waveform of a patient can also be obtained using the signal model of the present invention usmg a complex FFT, as explained further with reference to FIGS. 25A-25C. In general, by utiUzing the signal model of equations (89)-(92) with two measured signals, each with a first portion and a second portion, where the first portion represents a desired portion of the signal and the second portion represents the undesired portion of the signal, and where the measured signals can be conelated with coeffidents i a and i,,, a fast satuiation transform on a discrete basis can be used on the sample points from the output of the decunation operation 402. FIG. 25A conesponds generaUy to FIG. 14. with the fast saturation transfoim replacing the previously described saturation transfoim. In other words, the operations of FIG. 25A can replace the operations of FIG. 14. As depicted in FIG. 25A, the fast saturation transform is represented in a fast saturation transform/pulse rate calculation module 630. As in FIG. 14. the outputs are arterial oxygen saturation, a clean plethysmographic wavefoim. and pulse rate. HG. 25B and 25C Ulustrate additional detaU regaiding the fast satuiation transform/pulse rate calculation module 630. As depicted in FIG. 25B, the fast saturation transform module 630 has infrared log and red log modules 640. 642 to perfoim a log normalization as in the infrared and red log modules 480, 482 of FIG. 17. SimUarly, there are infrared DC removal and red DC removal modules 644, 646. In addition, there are infrared and red high-pass filter modules 645, 647, window function modules 648,640. complex FFT modules 652.654. select modules 653, 655, magnitude modules 656, 658. threshold modules 660. 662, a point-by-point ratio module 670. a saturation equation module 672, and a select saturation module 680. There are also phase modules 690, 692, a phase difference module 694. and a phase threshold module 696. The output of the select saturation module 680 provides the arterial saturation on an arterial saturation output line 682 In this alternative embodiment, the snapshot for red and infrared signals is 562 samples from the decimation module 402. The infrared DC removal module 644 and the red DC removal module 646 are sUghtly different from the infrared and red DC removal modules 484, 486 of FIG. 17. In the mfrared and red DC removal modules 644,646 of FIG. 25B. the mean of aU 563 sample points for each respective channel is calculated. This mean is then removed from each individual sample point in the respective snapshot in order to remove the baseUne DC from each sample. The outputs of the infrared and red DC removal modules 644, 646 provide inputs to respective infrared high-pass filter module 645 and red high-pass filtei module 647. The high-pass filter modules 645. 647 comprise FIR filters with 51 taps for coefficients. Preferably, the high-pass

filters comprise Chebychev filters with a side-lobe level parameter of 30 and a corner frequency of 0.5 Hz (i.e., 30 beats/minute). It wiU be understood that this filter could be varied for performance. With 562 sample points entering the 5 high-pass filters, and with 51 taps for coefBcients, there are 512 samples provided from these respective infrared and red snapshots at the output of the high-pass filter modules. The 0Ut ut of t h e P Wgh-pass filter modules provides an input to * e w^dow function modules 648, 650 for each respective f 1 n n t i f i fM

io

=r . . , .A , ,**. ,** * ^ e 7 " ^ ^ ^ m ° d u l e A s ^ 8 - 6 5 0 Pform a f i c o n Z ^ ^ T 8 T ^ T^^ ^ tion is used m the present embodiment. The functions throughout FIG. 25B maintain a point-by-point analysis. In the p r e s e n t ^ 0 ^ ^ , the time bandwidfli product for the 15 Kaiser window function is 7. The output of the window function modules provides an input to the respective cornplex Fast Fourier Transform (FFT) modules 652, 654. The complex FFT modules 652,654 perform complex FFTs on respective infrared and red channels on the data 20 snapshots. The data fromthe complex FFTs is then analyzed in two paths, once which examines the magnitude and one which examines the phase from the complex FFT data points. However, prior to further processing, the data is provided to respective mfrared and red select modules 653, 25 655 because the output of the FFT operation wiU provide repetitive infonnation from O-Vi the sampUng rate and from 1/2 the sampUng rate to the sampUng rate. The select modules select only samples from CM/2 the sampling rate (e.g.. 0-31.25 Hz in the present embodunent) and then select from 30 those samples to cover a frequency range of the heart rate and one 01 more harmonics of the heart rate. In the present embodiment, samples which faU in the frequency range of 20 beats per minute to 500 beats per minute are selected, This value can be varied in order to obtam harmonics of the 35 heart rate as desued. Accoidingly. the output of the select modules results in less than 256 samples. In the piesent embodiment, the sample pomts 2-68 of the outputs of the FFTs are utiUzed for further processing, In the first path of piocessing. the output from the select 40 modules 653,655 are provided to respective infrared and red magnitude modules 656, 658. The magnitude modules 656, 658 perform a magnitude function wherein the magnitude on a point-by-pomt basis of the complex FFT points is selected for each of the respective channels. The outputs of the 45 magnitude modules 656, 658 provide an mput to mfrared and red threshold modules 660, 662. The threshold modules 660, 662 examine the sample pomts, on a pomt-by-point basis, to select those points where the magnitude of an individual point is above a 50 particular threshold which is set at a percentage of the maximum magnitude detected among aU the remaining points in the snapshots. In the present embodiment, the percentage for the threshold operation is selected as 1% of the maximum magnitude. 55 After thresholding, the data pomts are forwarded to a pomt-by-point ratio module 670. The point-by-point ratio module takes the red over infrared ratio of the values on a pomt-by-point basis. However, a further test is performed to qualify the points for which a ratio is taken. As seen in FIG. 60 25B. the sample pomts output from the select modules 653, 655 are also provided to infrared and red phase modules 690, 692. The phase modules 690,692 select the phase value from the complex FFT points. The output of the phase modules 690,692 is then presented to a phase difference module 694. 65 The phase difference module 694 calculates the difference in phase between the conesponding data points from the phase modules 690, 692. If the magnitude of the phase v

5,632,272 57

58

difference between any two conesponding points is less than embodiment, the foUowing windowing function is selected: a particular threshold (e.g., 0.1 radians) in the present embodiment), then the sample points qualify. If the phase of _ _ _^ (105) two conesponding sample points is too far apart, then the i°" " — sample points are not used. The output of the phase threshold 5 module 696 provides an enable input to the RED/IR rate where SAr„ equals the saturation value conesponding to module 670. Accordingly, in order for the ratio of a particueach particular frequency for the sample points and SXran lar pair of sample pomts to be taken, the three tests are represents the arterial saturation as chosen at the output of executed: the select arterial saturation module 680. This window 6. the red sample must pass the red threshold 660; 10 function is appUed to the window function input represent7. the infrared sample must pass the infrared threshold ing the complex FFT of either the red or the infrared signal. 662; and The output of the window function module 700 is a red or 8. the phase between the two points must be less than the infrared signal represented with a frequency spectrum as predefined threshold as detenmned in the phase threshdetermined by the FFT, with motion artifacts removed by the old 696. 15 windowing fimction. It should be understood that many For those sample pomts which qualify, a ratio is taken in possible window functions can be provided. In addition, the ratio module 670. For those points which do not quaUfy, with the window function described above, it should be the saturation is set to zero at the output of the satuiation undeistood that using a higher power wiU provide more equation 672. noise suppression. The resulting ratios are provided to a saturation equation 20 In order to obtain pulse rate, the output pointsfromthe module which is the same as the saturation equation modwindow function module 700 are provided to a spectrum ules 502, 520 in the statistics module 504. In other words, analysis module 702. The spectrum analysis module 702 is the saturation equation module 672 accepts the ratio on a the same as the spectrum analysis module 590 of FIG. 20. In pomt-by-point basis and provides as an output a conespondother words, the spectrum analysis module 702 determines ing saturation value conesponding to the discrete ratio 25 the pulse rate by determining the first harmonic in the points. The saturation pomts output from the saturation frequency spectrum represented by the output pomts of the equation module 672 provide a series of saturation points windowing function 700. The output of spectrum analysis which could be plotted as saturation with respect to fre- module 702 is the pulse rate. quency. The frequency reference was entered into the points In order to obtain a clean plethysmographic waveform, at the complex FFT stage. 30 the output of the windowing function 700 is appUed to an The arterial (and the venous) saturation can then be inverse wmdow function module 704. The inverse wmdow selected, as represented in the select arterial saturation function module 704 completes an inverse of the Kaiser module 680, in one of two methods according to the present window function of the wmdow function module 648 or 650 mvention. According to one method, the arterial saturation of FIG. 25B. In other words, the inverse window function value can be selected simply as the point conesponding to 35 704 does a pomt-by-point inverse ofthe Kaiser function for the largest saturation value for aU points output from the points that are stiU defined. The output is a clean plethyssaturation equation module 672 for a packet. Alternatively, mographic wavefoim. a histogram siimlai to the histogiam of FIG. 22 can be Accordingly, by using a complex FFT and windowing generated in which the number of saturation values at functions, the noise can be suppressed from the plethysmodiffeient frequendes (pomts) aie summed to foim a histo- 40 graphic waveform in order to obtain the arterial saturation, gram of the number of occunences for each particular the pulse rate, and a dean plethysmographic waveform. It saturation value. In either method, the arterial saturation can should be understood that although the above description be obtained and provided as an output to the select arterial relates to operations primarily in the frequency domain, saturation module on the arterial saturation output Une 682. operations that obtain sunUar results could also be accomIn order to obtain the venous saturation, the minimum 45 pUshed in the time domain, arterial saturation value, of points that exhibit non-zero Relation to Generalized Equations value, is selected rather than the maximum arterial saturation The measurements described for pulse oximetry above are value. The saturation can be provided to the display 336. now related back to the more generaUzed discussion above. The fast saturation transform information can also be used The signals (logarithm converted) transmitted through the to provide the pulse rate and the clean plethysmographic 50 Anger 310 at each wavelength ka and Xb aie: wave form as further iUustrated in FIG. 25C. In order to Sx {t obtain the pulse rate and a clean plethysmographic wave °f^™%*™^^^%^^ ^^ nm», 1 AA-tic t* • ^ » « » * W + 6 f l V " - c Hb* W + n j J , ' ) ; (l05a) form, several additionai functions are necessary. As seen in i ,i FIG. 25C, the pulse rate and clean plethysmographic wave sXa(t)=erlb02j^ Hb02^(ty^Hbj[/,c Hb^(t}i-n^,(t); (105b) form are determined using a window function module 700, 55 a spectrum analysis module 702 and an inverse window Sv,W=-^(»)+nx»('); (i05c) , .

' -__,

,

.

,

. ,

r

Sxb(t)^%nJ2(t)^BbO2,XbC^HbO2x(t}i*Hb?jJ^Hbx(t}^Hb02.

As depicted in FIG. 25C, the input to the wmdow function u>cvHb02>•(f)jnHb>jfiyHbxv(i)+n-ljt)\ (106a) module 700 is obtained from the output of the complex FFT modules 652 or 654. In the present embodunent, only one 60 s^ty^tMo^Bu^ty^m^g^ity^^jt) (106b) measured signal is necessary. Another input to the window ,t,+n ,t, (106c) s ^ function module 700 is the arterial saturation obtained from the output of the select arterial saturation module 680. The variables above are best understood as conelated to The window function module performs a windowing FIG. 6c as foUows: assume the layer in FIG. 6c containing function selected to pass those frequencies that significantly 65 A3 and A 4 represents venous blood in the test medium, with conelate to the frequencies which exhibited saturation valA3 representing deoxygenated hemoglobin (Hb) and A 4 ues very close to the arterial saturation value. In the present representing oxygenated hemoglobin (HB02) in the venous

5,632,272 59 blood. Similarly, assume that the layei in FIG. 6c containing A 5 and Ag represents arterial blood in the test medium, with As representing deoxygenated hemoglobin (Hb) and Ag representing oxygenated hemoglobin (HB02) in the arterial blood. Accordingly, c''Hb02 represents the concentration of 5 oxygenated hemoglobin in the venous blood, c l l b represents the concentration of deoxygenated hemoglobin in the venous blood, x" represents the thickness of the venous blood (e.g., the thickness the layer containing A3 and A 4 ). Similarly. c A Hb02 represents the concentration of oxygen- 10 ated hemoglobin in the arterial blood, c^Hb represents the concentration of deoxygenated hemoglobin in the arterial blood, and xA represents the thickness of the arterial blood (e.g.. the thickness of the layer containing A5 and A6) The wavelengths chosen are typicaUy one in the visible 15 red range, i.e., ta, and one in the infrared range, i.e., ^.b. Typical wavelength values chosen are Aa=660 nm and A.b =910 nm. In accordance with the constant saturation method, it is assumed that cAiJ602(t)/CAffl,(t)=constant1 and cV Hao2(tycVH&(t) =constant2. TTie oxygen saturation of arte- 20 rial and venous blood changes slowly, if at aU, with respert to the sample rate, making this a vaUd assumption. The proportionaUty coefficients for equations (105) and (106) can then be written as: 25

60

invention which employs the constant saturation method, i.e., the signals SXa(t)=SXn!li(t) and S ^ t ^ S ^ t ) . Afirst segment 26a and 27a of each of the signals is relatively undisturbed by motion artifad, i.e., the patient did not move substantiaUy during the time period in which these segments were measured. These segments 26a and 27a aie thus generaUy representative of the primary plethysmographic waveform at each of the measured wavelengths. A second segment 26b and 27b of each of the signals is affected by motion artifact, i.e., the patient did move during the time period in which these segments were measured. Each of these segments 26b and 27b shows large motion induced excursions in the measured signal. A third segment 26c and 27c of each of the signals is again relatively unaffected by motion artifact and is thus generaUy representative of the primary plethysmographic waveform at each of the measured wavelengths. FIG. 28 shows the secondaiy reference signal n'(t)=n5UI (t)-tanu(t). as determined by a reference processor of the present invention. Again, the secondary reference signal n'(t) is conelated to the secondary signal portions n^and n ^ . Thus, a first segment 28a of tiie secondary reference signal n'(t) is generaUy flat, conesponding to the fact that there is very Uttle motion induced noise in the first segments 26a and 27a of each signal. A second segment 28Z> of the secondary (107) reference signal n'(t) exhibits large excursions, conespondemo2.w$KH-Jt)-ic(t)SH,(t) (110a) same for each figure to better Ulustrate changes in each signal. FIGS. 29 and 30 Ulustrate the effect of conelation v v v v -%«2,^ HM2* (0-KH 4 ,^ as (ty-nijt) canceUation using the secondary reference signal n'(t) as •-rM^HbtaMfiVHK,^v(fy^ab^cvHbxv(t)+n>J,(t)i (ilia) determined by the reference processor. Segments 29b and 45 30Z> are not dominated by motion induced noise as were Multiplying equation (106) by r^t) and then subtracting segments 26b and 27b of the measured signals. AdditionaUy, equation (106) from equation (105), a non-zero primary segments 29a. 30a, 29c, and 30c have not been substantiaUy reference signal s'(t) is determined by: changed from the measured signal segments 26a, 27a, 26c, and 27c where there was no motion induced noise. ^ify=SyJf)-rv(t)S^(t) (110b) so It should be understood that approximation n ' ^ t ) and n ^('HvCKiW (mb) "w'( t ) t o the secondary signals n ^ t ) and n ^ t ) as estimated by a conelation canceler using a primary reference signal The constant saturation assumption does not cause the s'(t) can also be determined in accordance with the present venous contribution to the absorption to be canceled along invention, with the primary signal portions 5^(1) and 8^,(1). Thus, 55 frequendes associated with both the low frequency moduMETHOD FOR ESTTMATING PRIMARY AND lated absorption due to venous absorption when the patient SECONDARY SIGNAL PORTIONS OF is stiU and the modulated absoiption due to venous absorpMEASURED SIGNALS IN A PULSE OXIMErER tion when the patient is moving are represented in the Implementing the various embodunents of the conelation secondary reference signal n'(t). Thus, the conelation can- 60 canceler described above in software is relatively straightceler or other methods described above remove or derive forward given the equations set forth above, and the detaUed both enaticaUy modulated absorption due to venous blood in description above. However, a copy of a computei program the fingei under motion and the constant low frequency subroutine, written in the C programming language, which cydic absorption of venous blood. calculates a primary reference s'(t) using the constant satuTo Ulustrate the operation ofthe oximeter of FIG. 11 to 65 rationmethodand,usingajointpiocessestimator572which obtain clean waveform, FIGS. 26 and 27 depict signals implements a joint process estimator using the equations measured foi input to a reference processoi of the present (54)_(64) is set forth in Appendix B. This joint process

5,632,272 61 estimator estimates a good approximation to the primary signal portions of two measured signals, each having a primary portion which is conelated to the primary reference signal s'(t) and a secondary portion which is conelated to the secondary reference signal n'(t). This subroutine is another way to implement the steps iUustrated in the flowchart of FIG. 9 for a monitor particularly adapted for pulse oximetry. The two signals are measured at two different wavelengths Aa and Ab, where Aa is typicaUy in the visible region and Ab is typicaUy in the infrared region. For example, in one embodiment of the present invention, tailored specificaUy to perform pulse oximetry using the constant saturation method, Aa=660 nm and Ab=940 nm.

62 SataruriJf)

( e HbM ~

20

25

(112b)

A

H402Ai)( WAlU.)

(113a) CHb02(t) +

cUt)

A third portion of the subroutine calculates the primary reference or secondary reference, as in the "CALCULATE PRIMARY OR SECONDARY REFERENCE (s'(t) or n'(t)) FOR TWO MEASURED SIGNAL SAMPLES" action block 140 for the signals 8^(1) and 8^,(1) using the proportionaUty constants ra(t) and r„ (t) determined by the constant saturation method as in equation (3). The saturation is calculated in a separate subroutine and a value of ra(t) or r^t) is imported to the present subroutine for estimating either the primaiy portions 8^(1) and 8^(1) 01 the secondaiy portions n ^ t ) and n ^ t ) of the composite measured signals

S,Jt) and SuHt).

30

3m(t)=nc[m] .Fswsqi pm(t)=nc[m].Bswsqr Ym(t)=nc[m] .Gamma 35

Pm.x*(t)=nc[m].Roh_a Pm,xi(t)=nc[m] .Roh_b em.xa(t)=nc[m]-en:_a 40

Km,^(t)=nc[m].K_a K

e

(113b)

fm(t)=nc[m].fen

m.X6(t)=nc[m].en_b

^Hb.>JAs>JAsu,)

ZHblc-^HbOlte(emM - GHKBAiXAnxoMnu,)

6.m(t)=nc[m].bref

e

(112a)

4b(.t)

Sat^Jt) 10

r

bm(t)=nc[m].ben

4•Hbm(t) +

^Hbte-^HbOZte-

Am(t)=nc[m]. Delta F ^ (t)=nc[m].fref

4.02(f)

ZHbte -

15

The conespondence of the program variables to the variables defined in equations (54)-(64) in the discussion of the joint process estimator is as foUows:

=

m.XA(t)=nc[m].K_b A first portion of the program performs the initialization 45 of the registers 90, 92, 96, and 98 and intermediate variable values as in the "INITIALIZED CORRELATION CANCELER" action block 120. A second portion of the program performs the time updates ofthe delay element variables UO with the value at the input of each delay element variable 50 110 is stored in the delay element variable 110 as in the 'TIME UPDATE OF LEFT [Z -1 ] ELEMENTS" action block 130. The calculation of saturation is performed in a separate module. Various methods for calculation of the 55 oxygen saturation are known to those skiUed in the art. One such calculation is described in the articles by G. A. Mook, et al, and Michael R. Neuman dted above. Once the concentration of oxygenated hemoglobin and deoxygenated hemoglobin are determined, the value of the saturation is 60 determined simUatly to equations (72) through (79) wherein measurements at times ^ and t 2 are made at different, yet proximate times over which the saturation is relatively constant For pulse oximetry, the average saturation at time 65 t=(t1+t2)/2 is then determined by:

A fourth portion of the program perfoims Z-stage update as in the "ZERO STAGE UPDATE" action block 150 where the Z-stage forward prediction enor F^t) and Z-stage backward prediction enor b0(t) are set equal to the value of the reference signal n'(t) or s'(t) just calculated. AdditionaUy zero-stage values of intermediate variables 3,, and (i0(t)(nc [mj.Fswsqr and nc[m].Bswsqr in the program) are calculated for use in setting registeis 90, 92, 96, and 98 values in the least-squares lattice predictor 70 in the regression filters 80a and SOb. A fifth portion of the program is an iterative loop wherein the loop counter, M, is reset to zero with a maximum of m=NC_CELLS, as in the "m=0" action block 160 in FIG. 9. NC_CELLS is a predetermined maximum value of . iterations for the loop. A typical value for NC_CELLS is between 6 and 10, for example. The conditions of the loop are set such that the loop iterates a minimum of five times and continues to iterate until a test for conversion is met or m -NC__CELLS. The test for conversion is whether or not the sum of the weighted sum of four prediction enors plus the weighted sum of backward prediction enors is less than a smaU number, typicaUy 0.00001 (i.e., 3 m (t)+pm(t) gO.00001). A sixth portion of the program calculates the forward and backward reflection coefficient rmJt) and F ^ ^ t ) register 90 and 92 values (nc[m].fref and nc[m].bref in the program) as in the "ORDER UPDATE m^-STAGE OF LSLPREDICTOR" action block 170. Then forward and backward prediction enors fm(t) and bm(t) (nc[m].fen and nc[m] .berr in the program) are calculated. Additionally, intermediate variables 3 m (t), pm(t), and Y(t) (nc[m].Fswsqr. nc[m].Bswsqi, nc[m]. gamma in the program) are calculated. The first cycle of the loop uses the value for nc[0] .Fswsqr and nc[0].Bswsqr calculated in the ZERO STAGE UPDATE portion of the program. A seventh portion of the program, stiU within the loop begun in the fifth portion of the program, calculates the regression coefficient register 96 and 98 values Km ^(t) and Km ^(t) (nc[m].K_a and nc[m] .BL_b in the program) in both regression filters, as in the "ORDER UPDATE m'" STAGE

5,632,272 63 OF REGRESSION F1LTER(S)" action block 180. Intermediate enoi signals and variables e mXa (t), emXi,(t), p^j^t), and pm,xi,(t) (nc[m].en_a and nc[m].en__b, nc[m]. roh_a, and nc[m].roh_b in the subroutine) are also calculated. The loop iterates until the test for convergence is passed. The test for convergence of the joint process estimator is perfoimed each time the loop iterates analogously to the "DONE" action block 190. If the sum of the weighted sums of the forward and backward prediction enors 3m(t)+(3m(t) is less than or equal to 0.00001, the loop terminates, Otherwise, sixth and seventh portions ofthe program repeat. The output of the present subroutine is a good approximation to the primary signals s ' ^ t ) and s ' ^ t ) or the secondary signals n ' ^ t ) and n ' ^ t ) for the set of samples S-^Jt) and 8^,(1) input to the program. After approximations to the primaiy signal portions or the secondary signals portions of many sets of measured signal samples are estimated by the joint process estimator, a compUation of the outputs provides waves which are good approximations to the plethysmographic wave or motion artifact at each wavelength, Aa and Ab. It should be understood that the subroutine of Appendix B is merely one embodiment which implements the equations (54)-(64). Although implementation of the normaUzed and QRD-LSL equations is also straightforward, a subroutine for the normaUzed equations is attached as Appendix C, and a subroutine foi the QRD-LSL algorithm is attached as Appendix D. WhUe one embodiment of a physiological monitor incorporating a processor of the present invention for determining a reference signal for use in a conelation canceler, such as an adaptive noise canceler. to lemove or derive primary and secondary components from a physiological measurement has been described in the form of a pulse oximeter, it wUl be obvious to one skiUed in the art that other types of physiological monitors may also employ the above described techniques. Furthermore, the signal processing techniques described in the present invention may be used to compute the arterial and venous blood oxygen saturations of a physiological system on a continuous or nearly continuous time basis. These calculations may be performed, regardless of whether or not the physiological system undergoes voluntary motion. Furthermore, it wiU be understood that transformations of measured signals other than logarithmic conversion and determination of a proportionaUty factor which aUows removal or derivation of the primary or secondary signal portions for determination of a reference signal are pos sible. AdditionaUy, although the proportionaUty factor r has been described herein as a ratio of a portion of a first signal to a portion of a second signal, a simUar proportionaUty constant determined as a ratio of a portion of a second signal to a portion of a first signal could equaUy weU be utilized in the processor of the present invention. In the latter case, a secondary reference signal would generaUy resemble n'(t)= ixiOO-mju^t). Furthermore, it wiU be understood that conelation canceUation techniques other than joint process estimation may be used togethei with the reference signals of the piesent invention. These may include but are not limited to least mean square algorithms, wavelet transforms, spectral estimation techniques, neural networks, Weiner and Kalman filters among others. One skUled in the art wffl reaUze that many different types of physiological monitors may employ the teachings of the present invention. Other types of physiological monitors include, but are in not limited to, electro cardiographs, blood

64 pressure monitors, blood constituent monitors (other than oxygen satuiation) monitois, capnogiaphs, heart rate monitors, respuation monitors, or depth of anesthesia monitors. AdditionaUy, monitors which measure the pressure and 5 quantity of a substance within the body such as a breathalizer, a drug monitor, a cholesterol monitor, a glucose monitoi, a carbon dioxide monitor, a glucose monitor, or a carbon monoxide monitor may also employ the above described techniques. io Furthermore, one skiUed in the art wiU realize that the above described techniques of primaiy or secondary signal removal or derivation from a composite signal including both primary and secondary components can also be performed on electrocardiography (ECG) signals which are 15 derived from positions on the body which are close and highly conelated to each other. It should be understood that a tripolar Laplacian electrode sensor such as that depicted in FIG. 31 which is a modification of a bipolar Laplacian electrode sensor discussed in the article "Body Surface 20 Lapladan ECG Mapping" by Bin He and Richard J. Cohen contained in the journal IEEE Transactions on Biomedical Engineering, Vol. 39, No. 11, November 1992 could be used as an ECG sensor. It must also be understood that there are a myriad of possible ECG sensor geometry's that may be 25 used to satisfy the requirements of the present invention. The same type of sensoi could also be used foi EEG and EMG measurements. Furiheimoie, one skiUed in the art wiU realize that the above described techniques can also be performed on signals 30 made up of reflected energy, rather than transmitted energy, One skUled in the art will also realize that a primary or secondaiy portion of a measured signal of any type of energy, including but not limited to sound energy. X-ray energy, gamma ray energy, or Ught energy can be estimated 35 by the techniques described above. Thus, one skiUed in the art wiU reaUze that the techniques of the piesent invention can be appUed in such monitois as those using ultrasound where a signal is transmitted through a portion of the body and reflected back from within the body back through this 40 portion of the body. AdditionaUy, monitors such as echo cardiographs may also utiUze the techniques of the present invention since they too rely on transmission and reflection, WhUe the present invention has been described in terms of a physiological monitor, one skiUed in the art wfll realize 45 that the signal processing techniques of the present invention can be appUed in many areas, including but not limited to the processing of a physiological signal. The present invention may be appUed in any situation where a signal processor comprising a detector receives a first signal which includes 50 a first primary signal portion and a first secondary signal portion and a second signal which includes a second primary signal portion and a second secondaiy signal portion. Thus, the signal processor of the present invention is readUy appUcable to numerous signal processing areas, 55 What is claimed is: 1. In a signal processor for processing at least two measured signals Sj and S 2 each containing a primary signal portion s and a secondary signal portion n, said signals Si and S 2 being in accordance with the foUowing relationship: 60 S^+n, S^+BJ w

65

h e r e Si and s2, and ^ and n 2 are related by: s i=r a s 2 and n ^ , ^ and where ia and iv are coefficients, a method comprising the steps of:

5,632,272 65

66

determining a value for the coeffident ia which minimizes conelation between Sj^ and n ^ calculating the blood oxygen saturation from said value of r a ; and displaying the blood oxygen saturation on a display. 2. In a signal processor for processing at least first and second measured signals Sj and S 2 each containing a primary signal portion s and a secondary signal portion n, said signals Si and S 2 being in accordance with the foUowing relationship: 10 Si=tsi+ni

where s1 and s 2 , and n^ and n 2 are related by: 15 51=Tas2 and ^=1^2 and where ia and r„ are coefficients, a method comprising the steps of: detennining a value for the coefficient ia which minimizes conelation between Si and n ^ and 20 processing at least one of the first and second signals using the determined value for ia to significantly reduce n from at least one of the first or second measured signal to form a clean signal. 3. The method of claim 2, further comprising the step of displaying the resulting clean signal on a display. 4. The method of claim 2, wherein said first and second signals aie physiological signals, said method furthei comprising the step of piocessing said clean signal to detennine a physiological parameter from said first and second measured signals. 5. The method of claim 4, wherein said physiological parameter is arterial oxygen saturation. 6. The method of claim 4, wherein said physiological parameter is an ECG signal. 7. The method of claim 2, wherein the primary signal portion of said measured signals is indicative of a heart plethysmograph, said method further comprising the step of calculating the pulse rate. 8. A physiological monitor comprising: a first input configured to receive a first measured signal Si having a primaiy portion, s^ and a secondary portion njj a second input configured to received a second measured signal S 2 having a primary portion s 2 and a secondary portion n 2 , said first and said second measured signals Sj and S 2 being in accordance with the foUowing relationship: Sr^i+n, u^:lS2'Tft2

25

30

35

40

45

50

9. In a signal processor for processing at least firs and second measured signals, each containing a primary signal portion and a secondary signal portion, said first and second signals substantiaUy adhering to a predefined signal model, a method comprising the steps of: sampUng said first and second signals over a period to obtain a first series of data points representing said first signal over said period and a second series of data points representing said second signal over said period; transforming said first series of data points into a first transformed series of points having at least a frequency component and a magnitude component and transf oiming said second series of data points into a second transfoimed series of points having at least a frequency component and a magnitude component; comparing said first and second transfoimed series of points to obtain a third series of comparison values having a magnitude component and at least a frequency component; selecting at least one of said comparison values that has a magnitude within a selected threshold; and from said selected at least one comparison value, determining a resulting value consistent with the predefined signal model. 10. The method of claim 9, wherein said step of comparing comprises determining a series of ratios on a point-by point basis of the first transformed series of points to said second transformed series of points, and wherein said step of selecting at least one of said comparison values comprises the step of selecting the lower of the ratios. 11. The method of claim 10, wherein said step of determining a resulting value comprises calculating a blood oxygen saturation from the selected lowei of the ratios. 12. The method of claim 9, wherein said resulting value is blood oxygen saturation. 13. The method of claim 9, wherein said resulting value is pulse rate. 14. In a signal processor for processing at least first and second measured signals, each containing a primary signal portion and a secondary signal portion, said first and second signals substantiaUy adhering to a signal model for blood constituent satuiation, a method comprising the steps of: sampUng said first and second signals over a period to obtain a first series of data points representing said first signal over said period and a second series of data points representing said second signal over said period; transforming said first and second series of data points from time domain to frequency domain to obtain a first transfoimed series of points and a second transfoimed series of points, said fiist and second transformed series of points having a magnitude component and at least a frequency component;

where S-L and s 2 , and % and n 2 are related by: Si=r a s 2 and n ^ ^ determining a series of ratios of magnitudes with respect and where ia and iy aie coefficients; 55 to frequency of ones of said first transformed series of a transform module responsive to said first and said points to ones of said second transformed series of second measured signals and responsive to a pluraUty points; of possible values for r a to provide at least one power curve as an output; selecting at least one of the ratios from said series of ratios that has a magnitude within a selected threshold; and an extremum calculation module responsive to said at 60 least one power curve to select a value for ia which from said selected at least one of the ratios, determining minimizes the conelation between s and n, and to a resulting value consistent with the signal model. calculate from said value for ia a conesponding satu15. The method of claim 14, wherein said ratios coneration value as an output; and spond to blood oxygen saturation, said step of selecting at a display module responsive to the saturation value output 65 least one of said ratios comprises selecting at least one of the of said extremum calculation module to display said ratios conesponding to the higher values of blood oxygen saturation value. saturation.

5,632.272 67 16. The method of claim 15, wherein said step of determining a resulting value comprises calculating the blood oxygen saturation from the selected at least one ofthe ratios. 17. The method of claim 16, further comprising the steps of: 5 combining with a window function at least one of said first transformed series of points or said second transformed series of points with said resulting value; and

68 sampUng said first and second signals over a period to obtain a first series of data points representing said first signal over said period and a second series of data points representing said second signal over said period; performing a fast saturation transform with said first and second series of data points to obtain a series of transformed data points in said frequency domain;

determining a selected saturation value from said series of performing a spectrum analysis on the combination to 10 obtain the pulse rate. transformed data points. 18. The method of claim 16, wherein said resulting value 21. The method of claim 20, wherein said selected satuis blood oxygen saturation. ration value is arterial blood oxygen saturation. 19. The method of claim 16, furthei comprising the steps 22. The method of claim 20, wherein said selected satuof: 15 ration value is venous blood oxygen saturation. using a window function, combining at least one of said 23. The method of claim 20, wherein said step of perfirst transformed series of points or said second transforming said fast satuiation transform comprises calculating formed series of points with said resulting value; and first and second pluraUties of intermediary transformed performing an inverse window function to obtain a plethysmographic waveform. 20 points from said first and second series of data points, said method further comprising the step of determining a pulse 20. In a signal piocessoi foi piocessing at least first and rate from said selected saturation value and from said first second measured signals, each containing a primary signal pluraUty of intermediary transfoimed points. portion and a secondary signal portion, said first and second signals substantiaUy adhering to a signal model, a method comprising the steps of:

UNITED STATES PATENT AND TRADEMARK OFFICE

CERTIFICATE OF CORRECTION PATENT NO. : 5,632,272 DATED : May 27, 1997 INVENTOR(S):Diabeta| It is certified that error appears in the above-identified patent and that said Letters Patent is hereby corrected as shown below:

Column 66, Line 1, "firs" should be deleted and --first--inserted.

Signed and Sealed this Twenty-nintli Day of May, 2 0 0 1

Attest: NICHOLAS P.GODICI Attesting

Officer

Acting Director of the United States Patent and Trademark

Office