CA2221384A1 - Blood glucose monitoring system - Google Patents

Abstract

A blood glucose monitoring system (100) includes a broadband light source (110) and a specially fabricated optical filter (120) for modulating optical radiation to be transmitted through a fleshy medium. Optical radiation (125) which passes through the fleshy medium (130) is detected by an optical detector (140) which generates an electrical signal indicative of the intensity of the detected light. Digital signal processing is performed on the electrical signal to extract those optical characteristics of the electrical signal due to the optical characteristics of the filter and constituents of the fleshy medium other than blood glucose concentration. The monitoring system employs a unique "double-log" transformation to minimize errors due to indeterminate path length variations of the optical radiation through the fleshy medium. The monitoring system (100) further employs specialized signal processing to avoid inaccuracies due to the previously unidentified solvent effect which arises when glucose is dissolved into water.

Description

CA 0222l384 l997-ll-l8 W O 96/41151 PCTnJS9Gi'-50C

BLOOD ~ OSE MONITORING SYSTEM
E ' .' of the Invention Field of the Invention The present invention relates to .systems for blood ~qlucose and other blood r ~titnD~It concentrations.
Dus,,.;"i- of the Related Art In the past, many systems have been developed for " blood cllald~ . For example, devices have been developed which are capable of determining such blood chdidcl~.i~lil,~ as blood oxygenation, nlucose .: . and other blood ~.hal ,~ c However, ~i~, ' ''0 ' have been ~d when 10 attempting to determine blood ~lucose concentration accurately using noninvasive blood monitoring systems.
The difficulty in .': . ~ blood glucose r , 2~ al~ may be attributed to several causes.
First, blood glucose is typically found in very low r _ within the ' ' - ' t.~all. (e.g., on the order of 100 to 1,000 times lower than ' ' ' ) so that such low cr ~ cliuns are difficult to detect r- ~s _'y, and require a very high signal-to-noise ratio. Second, there has been a lack of .. of the kinds of noise and the 15 proper method to use when removing this noise. For example, noise can be classified as deterministic (d ' '' ' or s~ ,u (random) where either of these kinds of noise could be linear (added) or, ' ' ': 'ti,' 1:
K~ $ of the distinction between the various kinds of noise is essential for purposes of using the proper method of removing noise. ~ , the optical Ghàla~,lu.i~tib~ of glucose are very similar to those of water which is found in a very high - aliv~ within the blood. Thus, where optical i ~, systems are used, the optical 20 chal dcl~H~Ii..s of water tend to obscure the Lhdl d~ liL~ of optical si~qnals due to low glucose ~ ll _ within the ~' Qd 1l r. Ih~ll e, since each individual has unique blood properties, each , typically requires : '' dliun for the particular individual.
In an attempt to - alul'? measure blood glucose levels within the '' ' ~" several methods have been used. For example, one method involves drawing blood from the patient and St:lJdl_' _ the glucose from the other 25 - within the blood. Although highly accurate, this method requires drawing the patiant's blood, which is less desirable than .dS;.~ to' , . especially for patients such as small children or anemic patients.
Furthermore, when blood glucose i is used to control the blood glucose level, blood must be drawn three to six times per day, which may be both physically and F~ ' ' 9~ traumatic for a patient. Other methods ,' . , ~, blood glucose c~ di- by means of urinalysis or some other method which involves 30 pumping or diffusing blood fluid from the body through vessel walls. However, such an analysis tends to be less accurate than a direct l~ea~... of glucose within the blood, since the urine, or other blood fluid, has passed through the kidneys. This problem is especially ". ~ in diabetics. r~ Ih~. ~, acquiring urine samples is often .~ ~ 1.
Another proposed method of blood glucose :- is by means of optical pr t~s ~.udlll. In such devices, light of multiple ~ ,lhs may be used to illuminate a relatively thin portion of tissue, such as a fingertip or an earlobe, so that a spectrum analysis can be r b ' to detErmine the, u~,wli.,~

W O 96/41151 PCTrUS9G,'C~SC6 of the blood flowing within the " ' tissue. Although such a method is highly desirable due to its n( ~ ._ _ character and its - .. to the patient, problems are ' with such methods due to the difficulty in isolatin~q each of the elements within the tissue by means of ~,l , analysis. The difficulty in determining blood glucose conL~.~tl_ - is further ~Aa~.l,.b.Jt~,d due to the low ' of glucose within blood, and the fact 5 that glucose in blood has very similar optical Lha~ li.,s to water. Thus, it is very difficult to distinguish the spectral ~.hala~.lL.i~ of glucose where a high amount of water is also found, such as in human blood.
As is well known in the art, different ' ' . typically referred to as ' ' ' . contained within the medium have different optical Lha,aclL.i~liL~ so that they are more or less absorbent at different ~ u~ of light. Thus, by analyzing the chala~ liL~ of the fleshy medium at different ~.u..' ~,lh~, an indication of the 10 L , ~ ' of the fleshy medium may be determined.
Sp ~, , ~ analysis is based in ~art upon the Beer-Lambert law of optical cll~laLlL~ s for different elements. Briefly, Beer-Lambert's law states that the optical intensity of light through any medium c ' ~ a single s -'-~ ~e is ~ to the exponent of the path lengths through the medium times the - ~._ of the s ih~l,, ,r~ within the medium. That is, I I e-(plc) (1) where pl ~ ..ls the path length through the medium and c l~JIti.~llls the ~ ~ '., " of the b~ - - within the medium. For optical media which have several ~ the optical intensity of the light received from the " - ' medium will be ~.~, i ' to the exponent of the path length through the medium times the - - ~1 alion of the first substance times an optical ' ,UI;U.I CG~.rril.;~..lt a~n ;alml with the first C..h,l,..,. D, plus the path length times the L lld" r of the second ~ b ' times the optical dbso"ui' c~rri..;~"l ~sso~: n d with the second ' : etc. That is, I =I e 1 1 pl C2~2 + etc-) (2~

where ~l~lt~..ls the optical abso,~i' L 'r;~;~..l.
Summarv of the Invention 25Due to the pal_ ~ required by the Beer-Lambert law, the ~ ~r ' ' in detecting glucose ~l arise from the difficulty in 1~ ' ', the exact path length through a medium ~resulting from 1~ cu, ~, the multi-path signal to an eq .' single path signal), as well as; r Ld due to low signal strength resultant from a low concl,..lialiu.. of blood glucose. Path length through a medium such as a fingertip or earlobe is very difficult to d~.lL.I~ since not only are optical . ~.. ~,Ihs absorbed by the fleshy medium, but also the signals 30 are scattered within the medium and 1- -- ' through different path lengths. Fi i' I ~, as indicated by the above nr, I . the measured signal intensity does not vary linearly with respect to the path length. Therefore, CA 02221384 l997-ll-l8 W O 96/41151 PCT~U59fl~ C
.3.
~lal in path length of multiple paths of light through the medium will not result in a linear averaging of the multiple path lengths. Thus, it is often very difficult to determine an exact path length through a fingertip or earlobe.
In addition to these ' '' ' s, it has been found that there is ~ difficulty in detecting glucose within water based upon Beer-Lambert's law. Specifically, it has been found that inaccurate are often taken of 5 glucose - r within water when optical - ~, , t are calibrated so as tD assume maximurc Il of optical ~ 'l Vlh~ will occur throu~h pure water without glucose. The present inventors have found that glucose together with water absorbs less than pure water for certain ' I-i bands and absorbs more for other bands.
In addition to the aforementioned difficulties, current optical ;~ . , devices, as identified by the 10 inventors for use in the present invention, often require expensive s ,~ ' filters which are used to generate a pattern of optical signals to be transmitted. One such filter, commonly known as a dichroic filter, comprises a rotating optically coated disk which includes regions of varying optical ;' ' The regions on the dichroic filter are formed in a pattern so that rotation of the optical disk results in the 1- of selected optical bands.
The high precision necessary for optically coating the filter substrate with various i' ~ ' r of optical material on 15 minute portions of the optical disk typically makes this coating process highly expensive. The present invention d ~ases the cost of a rotating dichroic filter by a factor of approximately 100 times by relaxing the sp~.ifi - --of the filter and ~ ' ~ for the relaxatiorl of filter, ~" - through rnore intensive Signdl ~
steps. The filter c :, ' in àcc~..d ~ with the present invention allows from 10 to 100 times as much light to pass while maintaining the same r.
One aspect of the present invention involves a system for ! i.. ~ 'J ~' hly blood glucose sJr ~ within a patient's '' -~ all,. The system has a light source which emits optical radiation at a plurality of ~ ylhS. A ,~b~"ld~ receives a fleshy medium of the patient and an optical detector is pc to receive light from the light source and di' ILd by the fleshy medium. The optical detector is .~ to optical radiation of at least the plurality of ~.u~ ylh;~ to generate an output signal indicative of the intensity of 25 the optical radiation. A signal processor is coupled to the detector to receive the output signal. The signal processor is r , ... to the output signal to isolate portions of the output signal due to optical Lhaldl,lelialil,s of the fleshy medium to provide a set of f,., ~, response values. The signal processor has a 1 i - module which linearkes the set of frequency response values and analyzes the linearized data to determine the sv of glucose within the patient's bl.,od~ .... In n one ' ' :, the ' . - module comprises a double logarithm 30 , di ~ In one embodiment, the light source comprises a plurality of emitters, each emitter ll, ~ light at a selected one of the plurality of ~ u..'l ~Ih~. In another ' - ' t, the light source comprises a L,.- " ' light source and the system further has an optical filter which ~c'~,li.s1~ passes ones of plurality of l~.a..' ~:hs present in the ' . ~ " ' light source.
In one b~ ' t, the detector comprises a single detector, , . to the ones of the plurality of v~.~' ":' to provide and output signal indicative of the sum of the : of the ones of the plurality of W O 96t41151 PCTAJS~6/~--OC
.4.
..,.~' Ih~. In another 6 ' ' t, the optical detector comprises a plurality of detectors, each detector ,_, iv~ to at least one of the plurality of ~ ~... ' ' to generate an output signal indicative of the intensity of the at least one ~
Another aspect of the present invention involves a _ ~ blood glucose I ~ system which 5 analyzes blood glucose within a fleshy medium - ~ blood. The system has a light source and an optical detector r., ~ to light from the light source to generate an output signal. A , ~ device is configured to provide for physical F,E. i L of the fleshy medium to express fluid from the fleshy medium. A signal processor is 1~, .., to a first output signal from the optical detector when fluid is -, L-7SOd from the fleshy medium and .., ~ to a second output signal from the optical detector when fluid is not; , LSSLd from the fleshy medium 10 to isolate ' relating to the : of glucose in the blood.
Yet another aspect of the present invention involves a method of r ....~ d~.; ' ' _ blood glucose conc~ dOon. The method involves a number of steps. A set of vaiues indicative of optical l,halal,l~,ii7liLs of I blood c is generated and a fleshy medium having blood is " ' with light of a plurality ~ of ~ ' Vlhs. The light is detected after di- '- of the light by the fleshy medium. A signal is generated 15from the detected light which is indicative of optical Ld~a~al~ 7 of the fleshy medium. C~, of the signal which are indicative of the r~ ~. of glucose in water within the blood are isolated in response to the detected light and the set of values indicative of optical Lha~al,ldd~lic~7 of the i~ C;~,a~l blood c,,..,lil -,U!; A value is then generated indicative of the glucose ~ . within the blood. In one - bL ' t, the si~t bbcd c comprise water, ' ' ' u~' ' ' and glucose dissolved in fluid.
20In one - bc ' t, the signal is linearized to provide an indication of the conc~ dliun of blood glucose.
Ad~ , the indication of the ~ ~r. does not vary - ', - lly with respect to the path length through the medium. In one bc ' t, the linearizing involves , a~ y a first set of values indicative of the optical cha,a~ .s of the medium, taking a first logarithm of the first set of values to generate a second set of values and taking a second logarithm of the second set of values to obtain a linearized set of values that are 25 indicative of the . ~., of the blood s :. In one: bc' :, the ' i~ài- further involves 1,_ c~ the first set of values using a F '~ ' equation to provide a 11 'u.l..~d first set of values.
These and other aspects are described herein in further detail.
Brief DU~; of the Drawinqs Figure 1 is a ' block diagram " t., the main ~u,..,i ' elements of the blood glucose 30i"g system of the present invention.
Figure 2 is a flow diagram which very generally depicts the method employed in accu"' -e with the present invention to obtain blood glucose ~:
Figure 3 is a flow diagram which illustrates in greater detail the method of the present invention used to determine blood glucose L ~- within a patient.
35Figures 4A-4C are graphs which depict the optical lla l,halal.l~D.7li~.s of a dichroic filter over different degrees of rotation and at three separate .. ' Ih~,.

W O 96/41151 PCT~US~6/~~'Q~ .5.
Figure 4D illustrates a matrix used to specify the optical LhalaLt~ ti~.~ of the dichroic filter used in - ' with the present inventiDn.
Figure 5 depicts the optical transmicivity of a fleshy medium, such as a finger, plotted against wavelength, the optical 1, :~ of a selected portion of the dichroic filter used in a ' with the present invention 5 plotted against ~ 3~ ,lb, and the combined optical 1~ it~ of the filter and the patient's finger plotted against ~ Jtll.
Figure 6A illustrates a plot of optical ll through a medium versus a product of the: b~ r ' c 'fi..;.,..l associated with the medium, the ~ dtion of a substance within the medium, and the path length through the medium.
Figure 6B is a plot of a "near-log" of the tl through an optical medium versus a product of the ~JIiUn cc~ r s ~ with the medium, the r ~ ~ of the ' s within the medium, and the path length through the medium.
Figure 6C illustrates a plot of the second log of the ll .;.y through an Opticâl medium versus the Iogarithm of the product of the ' - ~i & 'fiL:_..I of the medium, the &- of the medium, and the path 15 length through the medium.
Figure 7 is a flowchart which depicts the method used in accu,da"ce with the present invention to perform ": and c '' ~: functions.
Figure 8 is a flowchart which depicts in greater detail the method used in ? ~' ~ with the present invention to linearize values within the finger optical response matrix as indicated within a - ' . Ol~t block of Figure 20 3.
Figure 9 is a data flow diagram which ' 'I~ .b;~ the ~ steps ~ ~I, ' on an optical signal in order to determine glucose ~ , based upon the i" of a fleshy medium by the optical signal.
Figure 10 depicts the method employed in --- ,' with the present invention to caL~ the dichroic 25 filter depicted in Figures 1 and 9.
Figure 11 is a graph which plots glucose r t.~ s versus ~..,' 0I with average water ', ' . ~aLl~d.
Figure 12 . ' 'l~ depicts a hand-held glucose monitor such as may be ! ',I ' in aLLu. ' with the present invention.
Figure 13 depicts a dichroic filter as cu.. ll : ' by r . ' methods.
Detailed De.b.i: of the Invention OVERALL BLOOD GLUCOSE MONITORING SYSTEM
Figure 1 is a schematic block diagram which depicts the main functional and structural elements of a blood glucose 1 ~ system 100 in ~ . - with the present invention. As depicted in Figure 1, the glucose 35i ~, system 100 includes a broadband light source 110 for emitting optical radiation 115 over a broad band of optical . ~.. ' ~;h;.. In one ~ " t, the optical source 110 comprises a 3,000 kelvin, quartz, W O 96/41151 PCT~US96/08506 -6 halogen lamp. In addition to the halogen lamp, the light source 110 mav further comprise, in one embodiment, a first indium gallium arsenide (InGaAs) light emitting diode (LED) which emits light at a ~ of ,, I '~
1,300 : ~, a second InGaAs LED which emits light at ,, ~ 1~ 1,060 ~, and a third InGaAs LED which emits light at I ~ 940 ~.
In an ' Ilali.. , ' ' t, several LED's having ~ a.. 1~ Jth~ tuned to selected r. :) bands could be used in place of the lamp 110. Of course, such an ~ b~ is typically expensive due to the cost of multiple, ' ~ àHUi~.all3d LED's. Thus, the preferred - Scl : includes a device for producing multiple ~ _. ' "IhS in a cost effective manner as described below.
The optical radiation 115 emitted from the light source 110 is focused via a lens assembly 117 (which may comprise fiber optics or the like) and passes through a filtering element 120. The filter 120 may, in one - N ~ 9f - bc ' t, comprise a dichroic filter c L~.LDd in a d ~ with the teachings of the present invention, as described below with reference to Figure 10. The dichroic filter 120 comprises an optically ~
rotatable disk substrate which is layered with optical coatings having different i' ' ~s so as to modulate the b.uadbà,,d optical radiation 115 through a spectrum from the near infrared (NIR) (e.g., 700 nm) to the infrared (IR) (e.g., 1,400 nm). The filter 120 further includes an optically opaque strip 122 which may, for example, comprise brass or some other metal which is deposited radially outward from the center of the filter disk 120. The opaque strip provides a "O" location indicator. The filter disk 120 is driven in a circular motion (as described more fully below) by a stepper motor in one preferred e bcl Filtered optical radiation 125 passes from the filter 120 through a fleshy medium, perfused with blood such as a finger tip 130. In some rr'h . it may be desirable to provide a focusing lens between the filter 120 and the finger 130.
A pressure ~" " - device 129 is shown ' i '1) in Figure 1 to surround the finger tip 130. As will be discussed in greater detail below, the pressure 9"' i- device 129 is used to squeeze fluid out of the finger tip 130 for purposes of enhancing signal detection.
Light 135, which passes through the finger 130, is incident upon an optical detector 140. In one ~ '-. ~,2 _ bc ' t, the optical detector 140 c , i~d5 an InGaAs optical detector which has a well-defined optical response for . ~... k ~hs in the spectrum of interest in the preferred ' ~ ' (e.g., between 850 and 1,700 nm).
The output of the optical detector 140 connects to a prc. "i~, 150 which ~ L:~
a class A linear amplifier having an upper dynamic limit of between 1 and 5 volts. The output of the pre-amp 150 connects to a bandpass filter 160. The bandpass filter 160 N " 1~ comprises a linear RC filter having a lower cut-off ~ of 9~ O.S Hz and an upper cut-off frequency of ap~ 10 kHz. The lower and upper cut-off R~, have roll-offs of ~.~,..u,d...~ 6 decibels per octave.
The bandpass filter 160 connects to an analog-to-digital converter 170, which, in one preferred e ' - "
r , i~eS a /\-~: converter that converts the analog electrical signal output from the bandpass filter to a digital signal. The analog-todigital convertor 170, for example, may comprise a 16bit, 20 kHz ~ .. _ rate analog-to-digital convertor. One example is Model No. CS5317, available from Crystal S ' . The digitized CA 0222l384 l997-ll-l8 W O 96/41151 PCTAUS~6/Q~506 siynnal which is output from the convertor 170is then provided to the input of a digital signal processor 180 for signal, .c ~ In one ' ' t, the digital signal processor may be ,' ' in software within a computer. For example, an INTEL 486 MP, or an ANALOG DEYICES DSP chip, Model No. 21062, are two examples.
The digital signal processor 180 outputs a value indicative of the blood glucose levsl of the blood within the finger 5 130.
In ~ when light 115 is emitted from the I I " ' light source 110 over a ~ 0~ h rannye of 3" 1 ~ 1~700~ ~ to 1,700- ~, this ' ~ ' ' light 115 shines through the rotating dichroic filter 120. It should be noted that the light 115 is focused onto a portion of the filter 120 by means of fiber optics, a lens assembly (e.g., the lens 117), or the like. As the dichroic filter 120 rotates, the ' I " ' light 115 is filtered through a portion of the dichroic filter 120 producin~ the filtered optical radiation 125. As indicated above, the dichroic filter 120 is coated with optical layers of varying thickness so that different portions of the dichroic filter 120 pass different ~.u..'t ~,:' of light. Thus, as the filter 120 rotates, the optical radiation 125 output from the filter includes optical radiation of various ~ a.c'~..y;hs. In one r bc ' t, a fiber optic is used to couple the optical radiation 125 emitted from a portion of the filter 120 to the patient's finger 120. It shnuld be noted here, 15 that since the optical Lhalal,o,.i~lil,s of the filter 120 can be carefully measured and the rota~ional speed of the dichroic filter 120 is known, the pattern of optical radiation 125 emitted from the filter 120 to illuminate the finnyer 130 is well defined, and therefore, may be used during signal ~.,. " to determine the amtunt of -:
which is due to the optical filter 120.
The optical radiation 125 which is used to illuminate the finger 130 passes through the finger 130 to 20 produce the d-n ~ IP light 135. As is well known in the art, some of the optical radiation 125 passes ' ' through the finger 130, some of the optical radiation 125 is reflected within the finger 130 to ~produce a~.all~li The scattered radiation which is l, ' through the finger 130, together with the light which passes I ' ' through the finger 130, make up the light 135. Some of the optical radiation 125 is absorbed by c ~ within the finger 130.
The finger 130 is known to include a fingernail, skin, bones, flesh, and blood. The blood itself primarily c , i;.~.. water, o,~ ' ' . reduced ' ' ' lipids, protein and glucose. Each of these ~ within the finger (e.g., nerves, muscle tissue, etc.) . ': to the ~ i and scattering of the optical radiation 125 through the finger 130. As described above, the ' ~.i of optical radiation through a, ' , medium typically follows well defined laws in relation to the optical Lhala~ a of each of the ~o--l;l.----l,i taken 30 ., dtel~. These laws are expressed in the equations for Beer Lambert's law. The light 135 which passes through ~ the finger 130 is incident upon the optical detector 140. The optical detector 140 generates an electrical signal ", ' to the overall intensity of the light 135.
Although the light 135 typically has different ~ - at different ~ _.. ' "Ih~" the optical detector 140 , an electrical signal which is r ~ r i' ' to the area contained under the spectral response curve of the 35 light 135 within the optical band detected by the detector 140. That is, the optical detector 140 receives light having different at different ~ .. ' lh~. The detected ~ "lhJ are restricted over a band of CA 02221384 1997-ll-18 W O 96/41151 PCT~US9~'08506 approximately 850 nm to 1,700 nm due to the chalal,t~ of the detector 140, so that, if intensity is plotted âS a function of w... ' ' to obtain a spectral response curve, the area under the spectral response curve will be indicative of the average optical radiation intensity incident upon the detector 140. Thus, the electrical signal produced by the detector 140 is ! ~, ' to the overall (i.e., average) intensity of the light 135.
The electrical signal output by the optical detector 140 is amplified through the pre-amplifier 150 and is then passed for filtering to the bandpass filter 160. The bandpass filter 160 serves to eliminate high and low Gl, 1 noise signals which are eAll. to ': ~ the blood glucose level within the patient. An analog signal output from the bandpass filter 160 is ~ . t,,d to a digital signal within the analog-to-digital converter 170.
This digital signal is then ll ~x d to the digital signal processor 180 for I h ' _ The method employed by the digital signal processor 180 to determine blood glucose level from the digital signal provided by the analog-to digital converter 170 will be described in greater detail below with reference to Figures 3-9. Finally, a value indicative of the blood glucose level is output from the digital signal processor 180.
It should be noted here that the signal c( " ~ apparatus and method used to reduce noise in the signal provided to the digital signal processor (DSP) 180 ~ provides a very high signal-to-noise ratio lon the order of 90 100 dB, in one preferred ~ bcl ~, Such a p. I,ce~;"~ system is described in ' I ' ~ r ~ ~
No. WO 96112435 entitled SIGNAL PROCESSING APPARATUS. Particular attention is drawn to Figures 11, 11A, and 12, and the c I ~ , d i~.~ of the hllGI l ' r ~ ~ I;un, which detail the structure and method used to obtain a signal-to-noise ratio sufficient for the 9~ ' of the present invention. It should further be ' -d that, although the hltL.I ' r. ~ : iS directed to an apparatus and method for obtaining blood UA~, 20 levels, the same signal ,... steps are p_.~l ' in au~r,' with the present invention until the 20 KHz data signal is output to the DSP as depicted in Figure 12 of the hll~..llàliundl r, ~i.
Figure 2 is a data flow diagram which very generally l.~.b.~ i the overall method of signal analysis employed in a .' - with the present invention to obtain the blood glucose level of a patient through r optical ~( i"g. First, as indicated within an activity block 200, a patient's finger 130 is illuminated by light which is filtered in a known pattern by the filter 120. The optical signal which passes through the finger 130 is ~ . ~..lGd to an electrical signal by the detector 140, as indicated within the activity block 203.
In a first analysis step, electrical noise due to the electrical ~ , of the system 100 and ambient light incident upon the detector 140 are removed as,, b..~..laJ by an activity block 210. Thus, for example, 60 Hz electrical noise and room light are removed in this step. Since ambient light and electrical noise is linearly mixed 30 with the desired signal, this noise can be removed by a linear ' llabl at this stage in the signal ,ui, F I ~~ G.
Once electrical noise is removed from the electrical signal indicative of the light 135 passed through the filter 120 and the finger 130, a second analysis step is r ~1 I d, as ~I, G~ t~d by an activity block 220, whereby the electrical signal is processed to d( ' ' the, chdldl,l~.i~li..~ due to the filter 120. It should be 35 _ ~( : -d that the filter chàlal,l~ l;La are ' ' ' together with the finger chala~ G~.~ so that removal of the chal a~,t~ OI,;~ due to the filter 120 is ~ , " ' ' by means of ' ' ' t;.,.. rather than linear CA 0222l384 l997-ll-l8 W O 96/41151 PCT/U',.S/~rOf-.9.
X,,: In this way the portion of the electrical signal due simply to the: 't chd, i~ti~.~ of the finger 130 is isolated.
~ Once the electricai signal has been, . d to isolate the ~ solely relating to the optical attenuation chàla~ t;l,s of the patient's fin~er 130, those portions of the electrical signal wbich are due to the 5 ~ _ cha~aLIL.i;.~ ;> of the finger other than glucose and water are extracted in a third analysis step as indicated within an activity block 230. In the preferred embodime~t, a first sub-step olF this third analysis step involves ~ fluids from the fingertip to leave ' - 'I~ non-fluid elements of the fingertip whereby those finger r,~ : other than blood are extracted from the electrical signal. Another sub-step in the third analysis step involves extracting from the remaining signal ,~ ....tillg blood those portions of the electrical signal 10 which are due to the I _ cha~ of blood ~ other than glucose and water. Both sub-steps are advantageously accomplished by means of demodulation to remove the signal chalaLI.,.i~li..~ which are not due to the glucose or water within the blood.
Finally, as l,,,..l : d by an activity block 240, the ratio of the ~ of glucose to the of water within the patient's finger 130 is taken to extract signal chalacl~ s due to path length 15 through the finger 130. In this manner an absolute glucose u is obtained. This cùll~e~llai level can be output as a value indicative of the blood glucose level of the patient.
Figure 3 is a data flow diagram which details the method used during run-time to obtain the blood glucose level of â patient based upon detected optical signals. It should be . ' ~i -d, however, tha~ prior to run-time, initialization and - '' dOun routines other than the ' ' . " and salf testing procedures of block 305 are 20 F_.rull.._d.
PRE-RUN-TIME INITIALIZATION
The ~ ' ~ and - " dOull are ,Cel r~ at the factory or other time prior to use for patient monitor.
In general, a blood ~o~.til.----l matrix is cull~ ,d and a filter cllald~l~R~ matrix is also cu~ d, as described in greater detail below with reference to Figure 7. The blood c: :;IL ..I matrix l~.ll,..e,.l~ the -~ A r -25 chalal,l~ of each of the main s~ : within the blood over various ~ ' of light~ The filterchala~ lh~ .a matrix l~r b~._..l:i the b ~,: Lhalàl,l~ t;~.~ of the dichroic filter 120 at different portions of the filter 120 and for various .... ' .,lh~. of light. The filter chalal,l~li~lil.~ matrix is used in order to extract portions of the electrical signal generated by the detector 140 which are due simply to the optical c~ caused by the filter 120. The blood - : : matrix is used for cross s 1.' of light passed through tissue in order to 30 calculate the ~ , of the different co ~ including glucose, water, ' ~,' ' . etc~
~ The blood ~ : matrix and the filter l.halal~lL.i~ . matrix are both two dimensional matrices. The blood ~ matrix includes one column for each s : of blood ' ~d and one row for each ~ ~. ' ,,II~ of light which is measured~ In a preferred ' ' t, the blood matrix includes five columns and sixteen rows when five blood ~ : are analyzed and sixteen ~, ~...' Vlh~ are used. The filter 35 chala.,lL.i~li.. matrix includes one column for each ~ ~.. ",II, of light which is measured and one row for each rotational position of the filter 120, at which an analog to digital sampling is ~ tl ' Thus, in one embodiment, W O 96/41151 PCTrUS96~ 6 10.
the filter cl,dla~.t~.i~li~matrix includes 16 columns and 256 rows when 16 . u..................................... 'l ~ ' are used and 256 rotational positions of the filter 120 are defined. It should be ' -d here that it is not necessary that 16 different Ih;~ be used; however, the use of additional ~ ' "Ih~ is ~ia~ ad~ " for ;IIL~I 1' ,, the signal-to-noise ratio. Since about half of the incident light is transmitted through the filter at each position of the 5 filter, the same ~ , ' is detected multiple times lalthough in a unique - ' ~ with other ~ ' "lhs each time) so that the overall signal intensity is from 10 to 100 times the intensity of any single w ... ~,Ih and much higher than the noise floor. This is ~ referred to as Felgate's ~ , In this manner the spectral response of the entire filter 120 over the expected measured ~ a. ~,Ih~ is r ': '~ chalaLtt~ L The method employed to construct the blood constituent matrix and the filter ~halal,l~ lib~ matrix is described in detail below with reference to Figure 7.
RUN TIME PROCESSING
Once " and - "have been r r. d, the system is ready for run time use. As depicted in Figure 3, the start of run-time p,. is ,~". ' in a begin block 300. First, l ' , " and self-testing m ~ t;~ are, r. 1, as ,~y.l' in an activity block 305. Briefly, b ' !, " and self testing involves 15 boot ~r di' and ensuring proper operation of the blood glucose system 100, and to provide more accurate of the blood glucose COn-.L..IldliOm For example, the instrument first ': if there is a sufficient signal intensity to take an accurate reading, is there a ' i' ~ , is the patient following bl~l~. by pressing and releasing the finger 130, etc. After h- 'cl, " and self testing is c ,': d, the light source 110 is activated to transmit light 115 through the filter 120, as rl, t~ tod in an activity block 310. Initially, the light source 110 is activated while the patient's finger 130 is not ;,.l~ll o oo' between the filter 120 and the detector 140.
Thus, the light which is detected by the detector 140 r~ a baseline light intensity (I~) which can be used as a test to insure that a bulb which is too dim or too bright is not inserted as a r3,' ' bulb for example.
Once the initial baseline light intensity constant has been d: d, the patient inserts the finger 130, as indicated in an activity block 312, so that ",ed;..," of the blood glucose level within the patient's finger 130 may be taken. As described above, when the patient's finger 130 is inserted between the filter 120 and the detector 140, light 115 from the source 110 passes through the filter 120 and the finger 130 to be detected as light 135 incident upon the detector 140.
As indicated within an activity block 315, the light which is incident upon the detector 140 is c~.. II,d to an electrical signal and this signal is amplified in the pre-amp 150, filtered with the band pass filter 160, and sampled by the analog-to-digital converter 170. Since the filter 120 is rotating (at, " '~ 78.125 ",..... ' per second in one actual ~ , although other rotational rates could be ad~ as called for by the particular . , " ), samples of the electrical signal output by the detector 140 are indicative of the light intensity detected at various rotational positions of the filter 120. In one e '- ~, c ' ~ t, one complete rotation (i.e., 360~) of the filter 120 r ,. pr ' to 512 digital samples. That is, 512 samples are taken within the period o I-, " " to one revolution of the filter 120. Thus, for example, if the filter 120 rotates at 78.125 ~

per second, then 512 samples will be taken within 3~.- te'~ 1178th of a second, so that the sampling rate of the analog-to-digital converter 170 will be ., uAi,,,alul~ 40,000 samples per second.
~ As further described below, the filter 120 constructed in acco,. with the present invention includes , ' ' : regions within an entire ,. .. ' S~ ~' 'I~, the filter 120 is s~ layered so that the first 5 half-revolution of the filter provides a mirror of the signal of the second half-revolution of the filter t20. That is to say, as depicted in Figure 10, the filter is formed in a wedge shape so that the thickness in one direction is constant and the thickness in the r , direction increases linearly. Thus, the second half-revolution of the filter 1ZO is redundant. For this reason, digital samples taken for one-half of the revolution of the filter 120 could be discarded so that in each rotation of the filter 120 there are 256 samples used for purposes of digital signal 10 , . " rather than 512 samples in the Sc " described above. Alll"llal;. ,l~, all 512 samples can be used for . ~ 9 by averaging cu"l ' values. In yet an dll~lllali._; ' ' t, the, ' ' half of the filter may be used for filter and source ~ '' dliun. Each of the 256 sampies lif only half are used~ Il,~.~.,....l~ a different portion of the filter 120 having different optical ll chala~ lh,~.
Once the signal output by the detector 140 has been sampled, linearly added ~l , and S~J~r~l;
15 electrical noise inherent within the blood glucose ~, system is extracted (i.e., linearly 'I~LII,d), as indicated by an activity block 320. The method of subtracting this noise depends upon whether the noise is d~l~l ~ ~ or If the noise is dl' I ~ ~ '-, then this can be modeled with the ~" I, iale phase and ~aLllal,l~d. If the noise is ' : then the noise can be averaged towards zero. As described above, the filter 120 is specially designed to include an opaque strip (i.e., the brass strip 122). The digital signal processor 180 detects when the 20 opaque strip 122 of the filter 120 is ~ I d between the light 115 and the detector 140 by ~ ~, the intensity output from the detector. This intensity will be ~rr~"li.~ly zero when the light is blocked by the opaque strip 122. Since the opaque strip 122 blocks ,uL~lalli 'l~ all of the optical radiation ll ' from the source 110, any signal output from the optical detector 140 when the light is blocked (e.g., from ambient light, thermal effects, etc.), will be ...t~ d as electrical noise which is not due to either the spectral ' ~.l cllalal,t~,.i;~li Z5 of the finger 130 or the spectral P' ~,: chalal,~ lil,s of the filter 120. Thus, the digital si~nal processor 180 interprets the signal present at the output of the optical detector 140 when the brass strip 122 is i"l~
between the light source 110 and the optical detector 140 as ~ li noise which is . ' ~q :'~ subllal,led from all signals output from the optical detector 140. In one ~ bcd t, this is simply a , ' ' -' by ' Il the digital value cu,,, . " to the detected noise level from each of the digital values ,., ' ~ to the detected 30 si~qnal samples obtained within the activity block 315. All~.llali,l~, a shutter ' ~ could be ~ ~,cscd within the light path, or the lamp 110 could be turned off il~ to provide the same effect. In this manner, the electrical noise inherent within the blood glucose i ~, system 100 is removed so that those electrical signals due to the optical -' - r ' Cllala~,ll,.i:,li~.;~ of the filter 120 and of the finger 130 are c-- ' ~d in the further "., ~ ., steps.
Once the su- 1~ noise inherent within the blood glucose h~y system 100 has been extracted by averaging to zero and the d ~ :- noise is s bllaLILd by phase modeling the noise or averaging the d~t~

CA 0222l384 l997-ll-l8 W O 96/41151 PCTAJS96/08~06 noise to zero, control of the method passes from the activity block 320 to an activity block 323. Within the activrty block 323 the signal is divided by lO to normalize the signal. The I ~ ' ' signal is - ' ~ p,l ' within an activity block 325 to construct a signal intensity matrix, or vector, from the sample values obtained within the activity block 315 (taking into ~ the ~ of the electrical noise, ~I, ' within the activity block 320 and the signal normalization r r., in the activity block 323). The signal intensity matrix is a one column matrix ;~ referred to as a vector) including 256 signal intensity values li.e., one value for each sampled rotational position of the filter 120). Thus, the signal intensity vector is obtained by direct I of the optical signal which passes through both the filter 120 and the finger 130 and is detected by the optical detector 140. Of course, the values used to form the signal intensity vector are taken from the amplitude of the signals output from the detector 140 after subl, : of the noise from each sample. D~ 3 ~ each rotational position of a filter 120 which is sampled by the analog-to-digital converter 170 by the symbol ~, then ~I will ;-, to the first rotational position of the filter 120,~12 will c IG, ' to the second rotational position of the filter 120, to ~256, which Cu.l~ pr ' to the last rotational position of the filter 120 before ~l is taken again. Using this notation, I,j,l , ' to the intensity of light detected by the Dptical detector 140 when the filter 120is in the .15 first rotational position ~ ,2C~ , ' to the intensity of light detected by the detector 140 when the filter 120is in the second rotational position ~2. etc. Thus, the signal intensity matrix Sl i.~i. a single column matrix having 256 digital values from lS,l to 1~,256, which t~ , ' to the optical i.,l~ . detected at each of the rotational positions of the filter 120. In one ' -' t, the intensity values for several l~,' i are averaged to form the signal intensity matrix.
~' , : to the ~ ;OI1 of the signal intensity vector within the activity block 325, error checking, or testing, is ,r,~.flJ,l,.Ed within an activity block 327. This test is p~,i ' 'l~ pEI~urll~ed to insure that a valid finger sample is being ~: ed. Thus, at each stage of ll. 'u,, and filtering, the output is examined for an expected range of values to cu.l- r " to the expected range of ~..I, li~ for a human finger. If the output values are found to be off-scale, then a system error occurs to indicate that the ! " ~d sample is not valid.
Once error testing is r rull.. ~d on the signal intensity vector, ' ~,;.. ~.ildl d I ' as l(~), and the filter ..I,àla..l~liali.., matrix, ' I ~l d ~ ~ ' as F(~), has been obtained prior to run time, as indicated above and Il, t~,.ldd as a data input in a block 333, the signal intensity matrix together with the filter chala~ .;, matrix may be used to obtain a matrix indicative only of the optical ' ~,; chàlacl~l;ali~ of the finger 130, as llldd in activity blocks 330,331. That is, since the overall optical ' ~.0 Lhdl dbl~ .;. of both the filter 120 and the finger 130 are known as measured within the signal intensity matrix, I(~), and the optical absu,,ui chdla~ li,li.,s of the filter 120 are known as rt".~ ' by the filter cllalabld,i~l;.,, matrix, Fl~"l) the optical . of the light 115 due to the chala..ltl;~li..s of the finger 130 may be d~t~l ' by removing the optical ' ,~ ui due to the filter from the overall -' ~.i of the filter 120 and the finger 130. This is c ,' ' ' by first taking the inverse transform of the filter matrix, as r~,..t~ d in the activity block 331, and ' , ~l~
multiplying the signal intensity vector by the inverse filter matrix, as ll, b~ ~t~d in the activity block 330.

CA 0222l384 l997-ll-l8 WO 96/41151 PCT/US96,~-50C

Figure 5 C ,' 'l~ depicts the ~ ' ', between the optical ' l~i due to the finger 130, the optical ~ due to the filter 120, and the total optical ~' ~.i due to both the finger 130 and the filter 120. As depicted in Figure 5,ll ~ ~ of light over ~ ' from 850 ~ ~l ~ to 1,350 throu~oh the finger 130 results in a spectrum function desi,o,nated as TU~ wherein the transmission of light through 5 the finger 130 is plotted as a function of the ~ lh~ In similar fashion, the 1~ ~ ~ of light through a selected rotational position ~e.,o.., when ~ - 0, c I , l~ to 0~) of a filter 120 is plotted as a function of ~ _..' lh and is d ,, ' by the function F(<~) in Figure 5. Finally, the ' ~- . or . ' of the optical -b~ ~.i due to the finger 130 and the filter 120 is shown over the same ~ ~..' Ih~ and ,1.".l in Fi,qure 5 by the function l(~). To obtain l(~) from the finger transmission function T~l) and the filter transmission 10 function F(~"l), the optical l- ~ ~ p . at any ..,..' v:' along the functions T(A) and F(;b"l) is multiplied to obtain l(~). Thus, for example, at a ~ ~.. ~,lh of 1,050, the transmission of light through the finger 130 is 3,, 1 ~ ~ 0.24%, while the optical ll ~ ~ through the filter at the same ~.O~ ' for ~ - 0~
iS 3~ 80% SO that the total optical 1l_ through both the finger 130 and the filter 120 will be approximately 0.24% times 80%, or a total of 0.192%, as indicated at the wavelength 1,050 and the function l(~.
15 The functions II~) and F(~"l) may be l~ J by the signal intensity and filter .,I.a,d,.l~ matrices, ~ .,,.,.,ti.~ . Thus, since I(O = F(~,~) x T(~) (3) and l~p) ,., t....lt~ a v ' matrix (vector) - an intensity value for each rotational position value ~, while F(~ ,,.G.,.,.~t~ a two d ' matrix: _ a filter ll s ~ri..;...; value for each value 20 of ~ and each value of ,1 ~see Figure 4D), then the function T(,l), .., t~ tali._ of optical Ir through the fino~er 130, may be ~I,~..,~..,.It~,d as a one column matrix having values for each of the various ~ ... '~..~lh values, ,1.
In ? ~ - with one bc' L of the present invention, 16 ~u~ ~,Ih;, are selected over the range of 850 ~ to 1,400 r- : ~for purposes of chala~ the spectral chalaLI~.i~liL.:, of the finger 130 25 as well as the filter 120. Sr " ~I~, in the preferred ' ' :, the monitored ~, ~.. ' It.~ are 850, 880, 910, 940, 970, 1000, 1030, 1060, 1090, 1120, 1150, 1200, 1250, 1300, 1350, and 1400 - : ~.
The matrix form of equation (3) above is shown below:
.

As shown in Equation (4), the signal intensity matrix l(~) is equal to the product of the tWD ;" ' filter 30 I,hala~.ll,.i~lil. matrix, F(~"~), and the single column finger cllalaLI~ li.. matrix T(,l). In this equation, two of the matrices are given (i.e., I(~) and F(~.~l)). Thus, the third matrix, T(,~), which ~ c~ /t~ the optical ll uha-a..ll,.i~liL;. of the finger 130 for the 16 selected ~ Ih~ between 850 1 : ~ and 1,400 1 ~, may be obtained by simply, '~ the inverse of the filter chà,à"lL,i ,6,. matrix, cl~ ~, ' as F-'(~"l), by the WO 96/41151 PCTrUS96/08506 1(0 F(~ ) T(~) I~l f~lAl f~lA2 f~lAn tAl I~2 b~2Al ~ ~-- ~ tA2 (4) .
~-- : :

I~m f~m~l f~mAn tln signal intensity matrix, I(~), using c ~ ' matrix inversion and multiplication t~ ' as shown below.
T(~ ) F 1(~

tAl f~41Al f~2~2 f~lAm I~l tl2 fd)2~ 2 :
.

t~N f~mA 1 f~mAn I~m Thus, as indicated in an activity block 331, the inverse transform is taken of the filter Chala..L~ Lil, matrix, F '(~"1), and then this inverse matrix is multiplied by the signal intensity matrix, I(~), within the activity block 330 to obtain the r,l A~ response of the finger 130 as eA~ ..scd by the finger cha(aLl~lialiL matrix, or Ll. ,,, :-- o- vector T~).
Once the ll vector (or, as all~".dt~l~ referred to herein, the optical frequency response matrix), TU), has been obtained as indicated within the activity block 330, the digital values stored for each of the 16 selected ~ .... '~ Jths within the ll ~ vector are converted from a ' i 'I~ 1~ i ' function to a 10 linear function, as l~ .,t~d in a: ' . block 335. The method for - ~.Lillg the values stored within the l~ ,n vector, T(,l), to linear values will be described in greater detail below with reference to Figure 8.
LINEARIZATION OF THE TRA~Sr1'~SlON VECTOR
The ll vector, TU), is linearized to prevent cross ' ' i of illl~l rb. ~, signals. By linearizing 15 the ll ~ ~ vector, Chala~ Li~, ~ of the detected optical signal which are not due to the optical ~ of the glucose within the ' ' ~ a", can be linearly subtracted. Thus, the sequence of analysis and ' i~at- used W O96/41151 PCTAJS9~.'0~'0C

in ac- ' with the present invention is very important to prevent cross modulation of interfering signals with the desired ~qlucose signal.
The ultimate purpose of linearizing the ll vector, TU), is to eliminate path length ', '~
errors. As discussed above, the transmission intensity of light through a medium (e.g., such as the finger 130) has ~, 5 a ' ll~ c, ' " ' ', with the path length of the light ll ' through the medium. Thus, the different transmission ~.hala~.L,.i~l;L~ vary non linearly with the corresponding path lengths so that taking an average 1, ~ ~ intensity value over each of the possible 1,. ~ ~ path lengths will not result in a ll_ ~ -intensity value which is ' r ' ' of path lenyth. This is because, due to the non-linear character of the relation between the path lengths and the transmission intensity values, some path lengths are weighted more than others 10 so that errors along the heavily weighted paths become more, . - d in the final blood ~qlucose C~.l,;l -output value. By linearizing the optical ~I , ; response matrix, the path lengths become linearly related for each .... ' Lll so that the path lengths can be - ' :la~.L!d from the set of linear !, " - Once the values within the Il vector, T~,l), have been linearized, this linearized matrix is d ~ ' by the new name DU) herein, which .~ .... ...ls the optical density.
Once the values within the ~, vector have been linearized as indicated within the s.. ", : block 335 (Figure 3), control of the method passes to an activity block 340, wherein, in 7C 1' with one aspect of the present invention, ';La"l pressure ~i.e., hal above the patient's blood pressure) is applied to the finger tip 130 to express fluids from the finger tip 130. It should be ' -~ that this step in the ~ process is part of a lony loop method which is not, ~u"..~ ' along with the other signal r I "~ steps. This is because static . " ~ are required each time the - ' equations derived from the optical density vector are solved.
Thus, pressure is applied to the finger for a long enough duration to derive a solution to one or more ., and then the finger is released for another period of time sufficient to obtain accurate solutions for additional ' e~ i Thus, since the ~,, 'i~: of pressure to the finger tip 130 is p~lr, u.~d at a slower rate than the signal p,. steps, the finger tip is, for practicai purposes, in a static condition during each iteration of the signal p" r , method. Fluid is expressed from the finger tip 130 to obtain of the optical chal. ~ of the finger not due to blood. This allows removal of portions of tlle electrical signal ~, dl~.l by the detector 140 C.UII_, '- _ to optical ~halaLl~Bali~s of the finger not due to the blood within the finger tip 130. Briefly, the ~:AI~.l ~ of fluid within the finger tip 130 involves applying ~ h~ ' ' pressure or other means to the patient's finger 130 and releasing the pressure so as to cause ' 'I~ all of the blood to evacuate the finger 130. As iâ known in the art, the finger 130 contains not only blood, but also bone, flesh, skin ~ and nail, all of which absorb optical radiation, and therefore, contribute to the -' r ' ~.hala~ defined within the ll - : vector. However, for purposes of the present invention, it is not useful to include the optical - ' ~ i .. hala.. l~li;,liL~ of bone, flesh or otber non blood c of the finger 130 for - ' ' 9 blood glucose; these finger 'l : : may be r ' ~d as artifacts which should be extracted from the data used to 35 ' - I blood glucose levels.

CA 0222l384 l997-ll-l8 W O 96/41151 PcTAJs~ c ln order to remove these artKacts due to non-blood ~ ~ if of the finger 130, ' '" pressure is appiied to the finger tip 130. When the finger 130 is l h~ ' ' ' 'l~ altered so as to cause blood to flow in and out of the finger tip 130, the optical l, ' ~ LLIa~lL.i~lil.~ of the finger tip 130 when the blood is _..
from the finger tip 130 dKfers from the optical ll ' ' cllala~.l~li~lil.~ of the fingertip 130 when blood flows 5 into the fingertip 130. The ~ " is cr~..cli. !~ the optical ll ~ ' thalaLlL.i~l;l,~ of the blood, and other fluid, which was _._ ' Stated another way, the optical cllala..lL.i~li.,~ of the fingertip 130 which is full of blood is a - '~ of the optical ll. ~ ~ ~,halabl~ of the fingertip 130 when the blood is e.aLaat~d, and the optical ll_ ' ~ chalal.l~.i~li.,~ of the e.~ ' blood. Thus, by - ' t. a~.i' the signal produced by the optical detector 140 when blood is e.a~.L.Jt~d from the finger from the signal produced by the optical detector 140 10 when the finger is full of blood, the optical 1~ ' ' chàlacl~,;alil,s of the E.a.,ual~d blood may be obtained.
It will be c ' : ~d by this '~ ' that the blood need not be ~l , ' '~ e.acual~.d. In this manner, artifacts due to bone, flesh, nail and skin of the fingertip 130 are removed, or extracted, from the digitally p .~F d signa all,d by the optical detector 140. It should further be noted that other methods to express blood from the finger tip 130 may include, for example, raising the patient's arm in the air so that gravity causes a decrbase in the 15 flow of blood in the finger 130.
Figure 12 depicts an e , ' y: ' ~'' : of a device which could be used to provide for ~ ... ' of fluids from the fingertip 130 during spectral analysis in al,cul ' with the present invention. As shown in Figure 12, a hand held glucose monitor 1300 includes the light source 110, the lens 117, the dichroic filter 1Z0, and the detector 140 (all shown in phantom). The fingertip 130 is placed between a caliper pressure arm 1310 which may be used to apply pressure to the fingertip 130. A display 1320, which displays glucose conc~.. l, ' is also shown in Figure 12.
In a pal ' ' l~ a h, : : ' " I, a ,." d p~ ..ba~ of the fingertip 130 may be p .~- Illed in order to obtain additional ' ~ollllai' - regarding the optical chala~ .s of the fingertip 130 and the' blood therein. In such an . ' - " t, the fingertip could be placed within an inflatable jacket which is inflated to a relatively high pressure to squeeze most of the blood out of the fingertip 130. Thus, nearly 100% ' ' ' of the signal due to the blood flow in the fingertip 130 is achieved in this manner. This high ' ' increases the signal to noise ratio of the electrical signal due to the blood in the fingertip 130. Of course, it should be noted that such high ' ' " may change the path length through the fingertip medium ~since the thickness of the finger may be ' '~ altered). Such a change in path length may require . , ~ i' For this reason, in one b~ " I of the invention, a second water cG_~iL;~.. I value is included in the blood ~ " matrix, ~' 'I~. ' ' l~d as AU,r), where r :' ' t~ the number of blood which are used to form the blood ' matrix. This second water c~ '0.,;~..l value may be used to determine the change in the "depth" of water through the fingertip when there is ' '~;~.alll ~ d ' l' of the fingertip. Thus, a first water depth is - ' ': ' using a first ~ r ~ri~ at one time when the finger is not squeezed, while a second 35 water depth is - ' ' ' using the second cc_~ri..; ..t at another time when the finger is squeezed. The change of the depth of the water in the fingertip is ' :' 'I~ the same as the change in path length through the finger, W O 96/41151 PCT~US~6i'~D' so that the change in water depth may be taken to be .l, I - _ of the change in path iength. Thus, by using first and second i 'riL;~ for water in the ~in~qer 130 at seiected intervals, a ratio may be obtained which is indicative of the ratio between the path length throu~h the fin~er 130 with ' ' and the path iength through the finger 130 without modulation.
Once the signal features due to the non-blood artifacts within the fingertip 130 are extracted, those optical chala~.t~ h,~ which are due to the blood c - other than glucose and water are extracted from the optical si~qnal, as ..i, I ' in an activity block 345. The blood glucose ~~ level within the water of the blood stream is then d~ I, as further explained.
As indicated within the activity block 345, a ~ matrix is derived from the linearized fin~qer 10 rll, , response l.hala~ .;alil, matrix, DU), (after eAII: of non-blood signal features) and the blood : : matrix, AU,r). The blood : matrix is provided as a data input to the process r ~ ~ within the activity block 345, as indicated within the block 347. As is well known in the art, and discussed briefly above, when several ' are combined to form a medium with optical ' ,uliuil, the overall optical aL~u.~: and Il I chalaLI~iisli..~ of the combined medium are typically describable in terms of the optical ~.halal,t~ th,~
15 of each of the separate ~l of which the medium is: - ' Thus, for example, if a medium is: . - d of water, oil, and alcohol (which are only used here for illustrative purposes, since, in practice, these elements are ' ' ' ' the optical ll_ through the entire medium may be described, in acLu,d vuith Beer-Lambert's law, as I ., ; ' to the sum of the I, of the optical ' ~,liull ~haldl,~ of the first (i.e., water) times a ~ t. of the first c ~ : times the path length through the medium; plus the optical 20 ~ ,i c~ ~ril,; ..l of the second c (i.e., oil) times the c ~.~ of the second - times the path lengths through the medium plus the -' ~ c~ 'ri..i~..l of the third L .: :(i.e., alcohol) times the cuuc~..l. of the third c times the path length through the medium. That is, I e [aw Cw Pl +aO CO Pl +aA~A pl] (6) Since the optical ' ~,i c ~ri~ ,..l for each material typically varies as a function of wave len~qth, a 25 series of equations may be derived to describe the overall optical ~ b ~lliua chala'~ liL~ of the entire medium as a function of the - of that medium. Thus, Dividing both sides by IO and taking the logarithm of each side provides the following equation:

ln( A~ ) = a lW-Cw-Pl +a lo-Co-Pl +a lA -CA -Pl (8) W O 96/41151 PCTAJS~6/08COC

I e [~1W~W P1 +~IO~O P1+a~A-P1]
~1 o IA2 I e [~2W-C~W-P1+a2O~CO-P1+~ Ca-P1]

AN ~ e [~CNW-CW P1 +~NO Co Pl +aNA 'CA Pl ]

for the first ~ h~11, for example. If the optical ll. ~ through the blood Lor'l;u ~t~i perfectly followed Beer-Lambert's law, then a first logarithm would result in a set of linear equations as shown above. From these equations, simple matrix algebra could be used to obtain the c : of glucose in blood. However, as will be discussed in greater detail below, the optical l.halach,ialiL.s of multiple COliai Is within the ' ' ' - do 5 not result in a set of linear equations after taking a single logarithm. Thus, the present invention c , Ift a different approach which includes the steps of taking a near-log of the values within the ll ~ vector, followed by two true lùyaB-' or another near-log in an ' llaO.~ ! ' " t, to obtain more precise ' - i of the ll_ vector. It should be ' aluod that preliminary 1 i~di and other c~ liun steps are used in ; : with the above steps to insure that the r 'it -- of Beer-Lambert's law are satisfied before either 10 logarithm is taken. rc.i e, it may not be ncce~aaly to perform the second logarithm if a aurrii~"ll~ linear outcome is observed after the first logarithm is taken.
The double logarithm process is used in accGr.là"ce with the teachings of the present invention to arrive at a series of linear equations which may be described in matrix form. This i~ ai takes place within the ' ~,.JIi"e block 335 and is described in greater detail below with reference to Figure 8.
The ll ~ vector, TU), is now ' " ' as a single column linearized finger spectrum matrix, DU).
Thus, for each of the 16 selected ..a~ ,Iha of light, there exists a digital value cu"., ' " to the linearized optical t,~, r response matrix, DU). That is, DU) comprises a first value Dl~ indicating the linearized _' u rr 'ri~ ..l for the finger tip 130 at . O~ Ih~l, Dl2 indicating the linearized di' i' C ~r~ t for the finger tip 130 at ~ a~ 2~ etc-. to D~ln~
As discussed earlier, the above Beer-Lambert equations are typically not an accurate eA,ull a of the relation between the signal intensity values and the c lld: values. Thus, li..~ . of the ll vector involves a "~ ' " curve-fit process. The double IUyal"' ' process need only be carried out on the ll ~ ~ vector, however, and does not need to be carried out on the blood: : matrix and the cr dli.... vector to obtain a set of linear:, - Thus, each linearized signal strength value, D", is modeled 25 by the dLaul~Ji c ~ri~ t times the path length times the c- : _ of each of the c within the blood. The linearized signal strength matrix, DU), may be bA~.b~aLJ as a product of the blood c~ ~I I matrix, CA 0222l384 l997-ll-l8 A~l,r), and the ~ ~,, times path length matrix, h~ . ~t~ , ' as C(r)PL. This r " ~ ~i( is shown in matrix form as:
D(~) A(~,r) C(r)-PL

D~l all al2 alr C1PL
D~2 a21 a22 . . C2PL (9) D~ N anl - anr C,PL

where PL 1, b.. ~ the path length and c" C2, .. , C, represent the c di' n of the first blood c~ E, the 5 second bloDd c ,~ , ..., and the last blood c : L (d ~, ' by "r"). In ~ ,IO...,e with well known matrix algebra i ', the matrix IL~,.L~ thld the ~ times path length matrix, C(r)PL, may be d , - ' by taking the product of the linearized optical frequency response matrix, DU), and the inverse of the blood r " I matrix, A~l,r), as depicted below.
C(r)-PL D(~) A 1(~,r) ClPL D~l all al2 alr C2PL D~2 a21 a22 (lo C,PL D~N anl anr As indicated within the activity block 345, the inverse llallar A-iu~r)~ of the blood c~ : I matrix together with the linearized optical r,, ~ response matrix may be used to obtain a matrix indicative of the r ' dliun of the different c~ : : within the patient's blood.
Once a matrix indicative of the c ~dl;un of the different blood ~ has been derived as indicated within the activity block 345, the ratio of glucose . Il: times path length through the finger 130 15 to water: dliUn times path length through the finger 130 is calculatEd as indicated within an activity block 350. This ratio is the - of glucose in water which is the same as the glucose ~ within the blood stream. The above described method is F r. ' over several iterations so that several glucose . : .

CA 0222l384 l997-ll-l8 values are obtained. To obtain an average glucose c: value, least squares analysis, which is well known to those skilled in the art, is used to plot a line indicative of blood 91ucose - : . as indicated within an activity block 355. This value is output to a display (see Figure 12) which may be read by the patient or by the operator of the blood glucose monitor system 100. Control of the method then passes from the activity block 355 5 to an end block 360 wherein the method of the present invention is s , ': ' DERIVATIQN OF THE FILTER CHARACTERISTIC AND BLOOD CONSTITUENT MATRICES
Figures 4A-4D, together with Figure 7, illustrate in greater detail, the method employed in ~ ~ d ~ with the present invention to perform pre-run-time ~ ' and - "' , of the blood glucose i"~ system 100.
The initialkation and '' routine is " t~al~d in Figure 7 and starts with a begin block 800. Control passes from the begin block 800 to an activity block 810.
Activity blocks 810825 illustrate the method employed in acc...l' with the present invention to obtain the blood - : : matrix, AU,r), while activity blocks 830845 illustrate the method used to obtain the filter Lha~a..lGIial;~s matrix. It should be I ': -d, however, that although the method depicted in Figure 7 is IG~,.G:,~.llGd as though the blood : matrix is: :, ' followed by the filter chala~.ll..i;~l;l,s matrix, in practice these represent ', s ' I p..r ' G5 which may be run in parallel.
As discussed above, different ' ~ typically have different spectral chala~ .ialil,~. That is, different hn~es absorb light more or less across different . u. '~ 815 of light. Thus, when the optical tl_ ~ ~ of light is plotted versus ~ :' - for a given ~ a pattern, which is referred to as the spectral signature of a substance, is observed. This signature defines the spectral ~hdla~ liL~ of the - ': over the various ~ ' _LhS of light.
To form the blood; matrix, AU~r)~ the main Co",6~ of blood are separated into individual s ~h5~ res, and the spectral chalal~lGIi;~LiL;~ of these ' are plotted versus ~ ' Ll,. Separation of the individual ' may be p~ i by an actual physical ~ al. lr of the : in vitro or by means of an in vivo clinical test wherein the ~-~ ,, of the individual ': are carefully controlled to provide a reference for : the concGIl~ - of the various blood c In the present c bc ' t, since 161~.d~ h~ between 850~ and 1,400 : ~ are used to ~.I,a,a.,ll..i,G the optical ~.I,a,a~.lG,i~ ,s of the patient's finger 130, the same 16 ~ _.. ' Ih~ are used in order to construct the blood c~ matrix.
Thus, for each s ~h~l~ .e or main I within the blood, a I is taken to obtain the ' ~: at each of the 16 ~.~..' Ihs.
30Each of these 16 values, which are referred to as spectral ' ~,; c~ 6~ , is used to define a row of the blood : matrix, AU,r). Likewise, each column of the blood c- matrix will tG.I~, d to a particular blood c : so that there is the same number of columns in the blood ~ matrix as there are blood c~ : Thus, if there are "r" of blood and "n" ~, ....' ~ ' - over which the optical tha,a,.~G,i;,lics of these c~ are to be ,1~ , d, the blood: : matrix comprises an "n" row by "r" column matrix.

-W O 96/41151 PCT~US96/08506 As indicated within the activity biock 810, for the cases of blood ~ that can be physically separated, the first: I (e.g., waterl is illuminated as a separate r ~ '~ over the n ~ ' of light (i.e., in this - ' ' t, 16 . ~ Jth~) to obtain the spectral ~' r~ Cf' ~ of water over these end For example, a calibration sample could be used in this case. Then, as indicated in an activity block 5 815, the second - of blood is " ' with the n ~r ~.. ' ' of li~ht to obtain the spectral ' r'-'~iL;~ t:> of the second - (e.g., UA~ This is repeated for each of the main blood constituents until the "rth" blood constituent (e.g., rJlucose in water) is illuminated over the n wavelenoths to obtain the spectral r '' ~ of the last main blood -In a preferred embodiment, certain constituents, such as water and glucose dissolved in water, are measured 10 in vitro, while other blood I such as UA~ and d ~' ' ' ~ . are I ~ru. ' 1~ measured in vivo~ue to the difficulty of obtaining accurate in vitro measurements for these r 'il ~ Fur thermore, it shùuld be noted that in a non-imaging system, such as typically used to perform in vivo ~ ~ . the source and the detector should be diffuse. The in vivo and in vitro - I are used as an 3,,~ : . and an iterative process is used with the approximate values until the linearization process described above -o .. ~, In this 15 manner, n (e.g., 16 in the present ~ p~ ~ rOL;~ values are obtained for each of the r ~ In r ~ dalll,e with the present invention, one preferred ~ 5c ' the spectral cLdla~ ti"s for glucose, water, UA~h~ - and ~ ~1 ' ' as separate ' are '1 ~ ~ In addition, a fifth :- I" row within the blood : : matrix is defined by ' ~,; c '{ of blood due to SC~IIIL~
A further sixth L 'il : row may be added wbere the sixth L ~r is water again. This additional _r 20 may be used to double check the path lengths obtained at different times durin~, for example, ," of : ~h, ' ~ -' pressure to the fleshy medium.
As is well known in the art, the ll - of optical radiation through a medium typically involves the SCdll~li _ and reflection of light waves through the medium. This scattering ~rr.,~Ø~1~ increases the path length through the medium. Since some of the effects due to scaLI~.i are st- ~ 'Iy well defined, the extinction effects 25 due to Sl.dll~. " may be treated as an additional : within the bll - ' . The " ' r' 1~ c~.tri~ tS
for the s~atl~l g "t : may be d ' . i 'l~ by ll ~ light through an optical medium of which the r of each of the ~ : is already known.
Once all the main blood ~ : have been defined over the ~.. ' u ' of interest for each of the 16 1, u~ L~ from 850 nanometers to 1,400 a, the blood -~llnnt~ matrix is formed from the 30 spectral . b ~. c~, .ii"k,..t~ obtained for each blood r : Thus, the first column of the blood ~ matrix has each of the ' ~ ri.. ;.,.. l~ for ~.u~ Ihs ~1"12, to ~In, for the first blood c1r (e.g., water). The second column of the blood L~ matrix has the ' Jr~: c 'riL;.,..I values for each of the _Ih.. A 1, A2, to ,In, for the second blood : (e.g., OA~ b And so on until the last column of the blood ct l : matrix comprises the spectral ' ~:- r '~iL.;.,..;~ for the last main blood -I ~ at 35 the selected ~ ~.. ' _Ih~11"12, ... "n.

CA 0222l384 l997-ll-l8 W O 96/41151 PCTrUS96/08506 The activity blocks 830 845, together with Figures 4A4D, illustrate the method used in . with the present invention to construct the filter Lhala~te~;~liLv matrix. As discussed above, the filter 120 reflects and transmits optical radiation in different ~,-l, i for different ~ _. ' _lh~ at different places on the filter disk 120.
This is clearly illustrated in Figure 4A4C, wherein Figure 4A ~I, L...,.lt:i the optical ll_ of light at a . ~.. ' 0. of 850 : ~ plotted versus each of a possible 256 disk rotational posrtions. As shown in Figure 4A, when the disk 120 is in the initial starting position ~i.e., ~ - 0), the ll I ~ of light at 850 nanometers is 1,. u~illlalJ~ 10% through the filter 120, while when the disk 120 is rotated so that ~ - 32, the optical l-_ of light at 850 ~ : a through the filter 120 is 1", '~ 25%. Again, between the disk r. ~' positions of ~ - 128 to ~ - 160, the transmission of light at 850 ~ . _. ' _lh through the filter 120 is ~ 75%. Thus, the optical transmission for A - 850 - ~ is entirely cLalaLt,",,~d over 256 rotational positions of the disk filter 120, as depicted in Figure 4A.
Figure 4B depicts the optical t-_ ~ ~ thala~ liL~ of light at 1,150 : ~ over the same 256 rotational positions of the disk 120. Similarly, Figure 4C depicts a plot of the optical ll of light at 1,350 - ~ through the disk filter 120 at each of the 256 rotational positions of the disk 120. In one actual ; ' - " of the invention, the optical l, clla~al,l~ of the filter 120 are described for 256 rotational positions at each of 16 ..,.. ' 015 between 850 a and 1,400 rr 'E ~.
Thus, from these - . . a filter l.halal,tL.i~lil. matrix may be ~ ~l d, as shown in Figure 4D.
The filter ~ dla~.lu.i~ . matrix ' ~ : ' in Figure 4D as F~ t includes 256 rows and 16 columns. Each column of the filter cl.a.a~ 0" matrix cDmprises the spectral ' ~JIiUIl ~I,a.a~ lh.s of the disk 120 at each of the 256 rotational positions of the disk 120 for a given .. ,.. ' ~,0 In order to construct the filter l,hala..l~ li.. matrix depicted in Figure 4D, the filter 120 is " ' at a first rotational position over each of the 16 ~ ..ylh~ to obtain spectral aLsG-~i CGE.ii~.;....l~ for each of the 16 ~ ..ylll~, as indicated within an activity block 830. Once the spectral ' - ~J6UI1 ~ '6L;~ t:i have been dl: ~d for the first rotational position as indicated within the activity block 830, the filter is " ' at a 25 second rotational position (i.e., sb - 1) over the 16 selected ~ a~. ' v:hs to obtain spectral . ' ~.i Goeiri.,;L.~t~
for the second rotational position, as ,l, G..~.r,t~d in an activity block 835. This method is carried on for each of the possible rotational positions of the disk 120 until, as indicated within an activity block 840, the filter is illuminated at the "mth,n or last, rotational position (i.e., position 256) of the disk filter 120 over the 16 selected E ll.s to obtain the spectral ' p' - C06iiil.;~ t:i for the last rotational position. In one preferred 30 embodiment, where a stepper motor is used, the rotational positions will be precise from revolution to ~ L.~' ~' of the disk 120. Of course, a computer disc motor with salient poles and run at a constant speed could be used provided that phase dithers are minimized to less than one part in 256.
Once spectral b~ r - r 'i ~ have been ;': I -~ for all 16 ~ u..' "lh~ of all 256 rotational positions of the disk 120, the filter ~.halal,~ .s matrix is r ~.l d, as indicated within an activity block 845 by putting c 'ii~.;.,.il~ L 11, ' _ to various blood L - ' in rows and ~ ,lhv of the ll ' light WO 96/41151 PCTAUS9Gi~ CC

in columns. Ooce the filter chalaL~ il,s matrix and blood - matrix are ~ t,l : 1, the system has the y t~_ for routine, THE SOLVENT EFFECT
Although it is typically the case that when several ' are mixed to form c( I : elements of 5 an optical medium, the optical ll ~ and absorption ~.halal.t~ ti~ of the medium relate to the optical 1, and ' ,Jt;J.. tha~aL~ ,s of each of the c elements within the medium as described by Beer-Lambert's law, it has been found that in certain isolated cases, this is not the case. In parlicular, when glucose is mixed in water so âS to dissolve in water, the optical Lba,a.,tL,i~ti.~ of the ~ ' ~ of glucose and water do not correlate directly to the optical chala.,l~li,lb~ of glucose and water 9 ~ ~tu!~ in 1~ da"..d with the relation 10 described by Beer-Lambert's law. In fact, when glucose is dissolved into water, it is found thadt the glucose water solution has a lower ' ~,~i.ity li.e., a higher t. ~ ity) than water without glucose at certain ~ u~ ;hc In past systems which monitor blood glucose concentration.by means of opticai signal detection and this solvent effect, which was I ~ur~ ' in a, . ' - involving blood ~,.c.,ll )", was the cause of a s ~, ~iCal~l amount of ab~ and detection of the actual blood glucose :1 ~tiUII. Thus, in 15 a~ ~dàll~.e with the teachings of the present invention, special "~;Cai' are made to both the blood s matrix, AU,r), and the filter cha.ablL.;~Ii., matrix, Fl~) in order to ~ . for i al.i.,3 and I 1' iti~5 due to the solvent effect.
S, ~'i -'1~, the blood ~ : matrix includes lb ~ ib;~.~t~ for glucose as dissolved in water, rather than glucose as a separate ': The scale factor by which the ' ~ ~ 'riL.;~..ls for glucose dissolved in water differs from the ' r~ 5? ~rh.;.,~t of water is a~ equal to log(-log T~g) - log(-log Tw), where T~9 is the 11 ;;y of water having glucose dissolved therein, and Tw is the ll .;ly of pure water.
Fi ' ~ùle~ the ~ u~ ,lh~ which are to be monitored ~which defines the filter clldlal,l~ Li,,s matrix) are selected :''r~.L..;I~ than would be expected for a ~,' 'Y~at~l mixture. This is because the solvent effect 25 causes a small shifting in the r" of maximum bs ~-i of the blood cr IS. For example, glucose and water acting as separate agents would have IG~ O.~ ~ U~ ll C' ~ ~IJ; maxima at a" UAilllall.l~ 1070 ~ and 975 ~. However, due to the solvent effect, the water with glucose has one e' ~,i maximum at a~,ul oA;I~.alLl~ 960 r : ~.
Since the àL~ J' clldla~.ll,.i~lil.5 of pure water differ from the ' ~ ~: I,ha,a..l~ of water having 30 dissolved glucose as a function of ~ Ih, this ~ ,. in ' pt .,ha,a~ may be used to scale JI;U~IC~'riLi~...l values for glucose dissolved in water within the L'm~ ' ~,. Figure 11 I~lJIG~ the ratio between the optical ab~ ~ ~dlala~ H~liLs of water and water with glucose. As shown in Figure 11, water ~ " dissolved glucose has ' '1~ lower -' IJi than pure water at ,, . ly 960 ~ ~,1145 - ~, and 1380 - - ~, while pure water has - b~ lower ' ~:- - than water I - ., dissolved glucose at ., . tJ~ 1060 - ~, and 9 ~ the same .' ,Jt;.:~y is observed for both at ~ .. ~, ' of 1000 r : ~,1100~ ~ ~, and 1Z30 . ~. It should be noted that those values W O 96/41151 PCT~US9~/~QeO6 of the curve of Figure 11 within the shaded region are due to the solvent effect, and are therefore _ pr!~ ~ in normal ~. ,, a, . " The values of the d '~e.l between the ' ,uli.;I;~ of pure water and water with glucose at the peak ~ h~ (i.e., 960, 1060, 1145, etc.) vary as a function of glucose c~
By using the ~ hs for which the ' pli.;;y is ' 'l~ the same for both pure water and water with 5 glucose as a baseline, scaling factors for determining the bs ~,i 'Ih,; .Its of glucose dissolved in water may be ~ ~' Thus, the present invention provides for more accurate , of the blood glucose level due to recognition of the solvent effect, and increased sc,.~iti.ity if one uses 1060 nm ~ _..' "Ih (i.e., the minimum) to provide an increased ' '~c,l 'L value with the 960 nm and 1145 nm ~:~.. ",Ih~ (i.e., the maximum).
METHOD FOR LINEARIZING THE OPTICAL ~nt~u.,.C~ RESPONSE MATRIX
Figures 6A-6C together with Figure 8 depict the method used in ac~.u.. ' with the present invention to lineari~e the values within the t, vector, TU). As depicted in Figure 6A, the ,.~.u u ll of optical radiation at a particular ~ ~,Ih (each curve ~ y a different path length P1, P2, etc.) is plotted versus the product of the !' ~J" UErliC;0..1 and the cunc~ of the optical medium through which the optical radiation passes. Thus, the curve of Figure 6A indicates the relation between how much light is ll 15 through a given r ' 't ~Ih of a medium, and the ~ dP of a given ' ~ within the medium. A total of eight curves are shown in the graph of Figure 6A wherein each curve indicates a different path length (e.g., P1, P2, ... P8) of light through the medium.
As is well known in the art, some optical radiation which passes through an optical medium such as the finger 130 passes 1i; 1~ through the medium without ' ' Scall~ so that the path length observed 20 by that portion of optical radiation is a~, u~d~lldll,ly the same as the thickness of the finger. In other cases however, the optical radiation scatters within the medium so that the effective path length observed by these portions of the optical radiation is ' : 11~ longer than the thickness of the optical medium. It has been found that the average path length observed through an optical medium such as the finger 130 is typically three to five times the thickness of the finger 130. It is important that a linear relation exist among the several path lengths of optical radiation 25 through an optical medium such as the finger 130 in order to insure that any given path length results in the same relation between the s , of a ' within the medium and the optical Lhalacl~.;.li"~ of the medium.
If a linear relation does not exist for various path lengths, a set of linear equations is not c': ' ' and the solution derived from the matrix equations will not be accurate.
As shown in Figure 6A, a linear relation does not exist among the several path lengths of optical radiation 30 through an optical medium. Rather, each of the eight curves depicted in Figure 6A indicate a logarithmic ,l ' ', between the ll ~ through the optical medium, and the product of the; ' ,uP 'P and the c - l, of the ' within the medium. As a result of this r ~ ' ily, no ~ v,..ldli~ average spectrum (i.e., one which is . ~ path length i ' ' :: can be obtained. For example, a signal through a path twice as long will typically be ai ' d by a factor of 1000 times, so that in an average its sum would 35 be i , I.

WO 96/41151 PCT~US96/08506 lf the Lhalal,t~ iL~ of the medium were such that Beer-Lambert's law held precisely (i.e., the cuNes depicted in Figure 6A were perfect , 's), then a simple logarithm could be used to reduce the transmission vector TU) to a set of linear , However, due to Rayleigh SCa~ and multiple - ~I within the finger 130, the intensity of the optical radiation which passes throu~h the finger are not related according to a strict 5 application of Beer-Lambert's law. This is shown in Figure 6A by a curve 601 which represents an actual, non-exponential relation between the l,_ ~ ~ p _.. ~ and the product of: ' r' ~ m '- and path length. The d~r~,. between the curve 601 and the idealized , cuNe 1 is ll~r ~ d by the shaded area in Figure 6A.
In order to transform the relation between l- F--~ ,. and the prcduct of h~ ~
10 c : . and path length depicted in Figure 6A to a more linear relation, a cubic curve fit is performed on the actual l,_ ~ ~ ~ vectors to bring them into f t~ with the ~, ' curves depicted in Figure 6A, which is near to taking the logarithm of the ll. r ~ That is, a cubic fit equation is used to bring the curve 601, for example, in ~- ' . ~ with the ~ pr 'i~' curve 1. Thus, this cubic fit is described ~_.I' 'IL.
as a "near-log." Over selected regions, a cubic equation may be defined so as to resemble, or closely 3~,1 15 a ' g ' curve. Since the actual relation among the values in the transmission vector TU) is typically near logarithmic, this near-log (i.e., cubic) equation may be used in place of a logarrthm to more accu.dtely linearize the values within the l. vector T~,l). Of course, it will be ~f Jed that a fourth- or higher degree equation may be used to provide more accurate linearization of the values within the ll ~ vector TU). In one ' - " of the present invention, the near-log equation has been eA~,_. 'l~ d~,lLI I to be 20 for ll through an adult human finger.

OD - -2.8088*T3 + 4.7801~T2 - 3.4215~T + 1.1289 and is typically applied over the range from T-O to T-1. The precise values for the ~ '6b; .,t~ in the near-log 25 equation above may vary from ~" ' to 3~," ; . and, ~ , an empirical _._' : with a number of test samples is used to precisely define an equation for a given ,, ' Once a near-log has been taken of the 1, I ~ " as discussed above, the ll ' ', between the near-log of the ll and the ' ~,i r~ '~;L;_.It times the r ~ L ' of the ' within the optical medium is ': 11~ , ' Thus, by taking a logarithm of the resultant vector values, a linear 30 1~6i ', iS obseNed between the values on each curve as depicted in Figure 6B. Although the relationship shown in Figure 6B between the nearlog of the 1l ~ and the -' ~: - cG_f6L.i~ times the - ~, is a linear one, the slopes of the lines cr ,., s " ., to different path lengths is more or less steep depending on the path Ien~th (P1, P2, ... P8). Thus, a different ,.' ~; ' ~. exists between the near-log of the ll_ and the product of the aL;,u"J ~ 'fib;~..l and the ~,_ based upon the path length through the medium. Thus, the path 35 lengths will not result in a ~ identical output. That is, a set of path-length .al;d"l, linear equations has not yet been formed. In order to prevent any error due to a ' ' d 'f~., in path length (i.e., I " it~ of CA 0222l384 l997-ll-l8 W O 96/41151 PCTAJSg~'B3~06 the values within the l~TU) matrix), a second log (referred to ' . 'I~, after as the second log) is taken of both the corrected 1~' _ and the product of the ' ~ co 'ril.; .~t and the ç ~It should be, ' -d here that a near-log fit may also be used in place of the log, as required by the specific n~, ' to obtain the most linear values.
The effect of taking the second log is depicted in Figure 6C. As depicted in Figure 6C, each of the lines Il, . : _ differing path lengths is parallel so that each of these lines bears 5~ the same ..' to the log of the product of the ' ~: ~ c~ ri ~ and the c as to the double log of the l-_ ~
percentage. That is, the values within the resultant linearized optical r~ response matrix DU) are all linearly related so that a set of linear equations may be derived from each of the curves depicted in Figure 6C. Finally, all 10 ~ . relating to the p ' ' Ih is removed by rotating the curves of Figure 6CS0 that the curves are parallel to the x-axis (not shown).
The method of linearizing the optical f- , ~ response matrix is outlined briefly in the rh,. L.hal I of Figure 8. As depicted in Figure 8, control passes from a begin block 900 to an activity block 905, wherein the i - process initiates using fixed r,o~ri";_.. l~ that are , 'l~ before run time. Once enough data is gathered on the patient, to determine that a minimum ll variance is obtained Idue to the pressing motion of the finger 130), control of the method passes to an activity block 910 wherein the near-log cubic curve fit is ~-~ r~ , ' on the optical frequency response matrix. That is, each of the values within the ll . vector T(,l) is ': -' for "T" in the cubic equation:

OD - 2.8088*T3 + 4.7801 ~T2 3.4215~T + 1.1289 to form a new, i, ' matrix.
Once a near log is taken of the optical frequency response matrix, the first logarithm is taken of the Il h ' vector values as indicated in an activity block 920. The result of the first logarithm is a matrix having values related to one another as depicted in Figure 6B.
The second logarithm is taken of the optical r" ~ ~ response matrix, resulting in the relation depicted in Figure 6C, as ,., . ' in an activity block 930. That is, a logarithm is taken of each of the values within the matrix defining the values of Figure 6B to obtain a set of linearly related values.
A least-squares analysis is, h, I on the linearized values to determine if the lines defined by the matrix 30 (see Figure 6C) are maximally parallel as ,l t:;. ..lcd in an activity block 940. That is, the slopes of the lines are ' and the sum of the squares of the; r~.. between the ' ' 3~ slopes serves as a measure of how parallel the lines are. An iterative process which involves varying the cc 'ri.,;~ of the curve fit equation is used to d ~ ~ if the lines are maximally parallel, or at least parallel enough to obtain a viable glucose Thus, as ll,~ ed by a decision block 950, a test is I r~, ~ to determine if the lines are parallel.
35 If the lines are not parallel, then control passes to an activity block 955 wherein the values of the c 'li..; ..ls of the cubic fit equation are adjusted. In one ' - " :, the value of one r ~ri~ beginning, for example, with CA 0222l384 l997-ll-l8 W O 96/41151 PCTAUS96,'CF~UC

the added constant, is increased or d ~a~r,d by a fixed increment during iterations of the - ';
routine ~i.e., each time the block 955 is entered from the block 950). Control then returns to the activity block 910 to apply the modified cubic fit equation to the vector values. The process repeats and a new indication is obtained of the pa~ of the lines. If the lines have become more parallel, then a further adjustment of the same 5 r- 'ti~ .lt iS made in the same direction by the same or smaller ~ I This continues until the lines no longer become more parallel. If it is ':, ' that the lines become less parallel based upon an adjustment, then the same s 'i is adjusted in the opposite direction (i.e., if the adjusted r~ til~ t was first increased, it is d ~as~..). This is repeated until a relatively small change is observed in the pa- ~- ' of the lines upon ... adjustments of that c~ '' Adjustments are then made to the next c~ '~i..;~..t (e.g., the 10 s~ 'I ~ : on the first order term) in the same manner. Once this process has been repeated fcr each cc~r, ~ t, the decision block 950 ':, if the pal ~ll- ~ of the lines defined by the lineariied vector is sufficient to produce â viable blood ~lucose measurement.
In a preferred i ' - ' t, an indication is also given of the certitude of the glucose '~: based upon the measured pdl_" In~SS of the lines of Figure 6C. This measure of certitude may, for exampie, be displayed together with the glucose , on the display screen 1320 (see Figure 12). If it is ll:, ' within the decision block 950 that the - ~ is not viable, an error signal may be generated, for example, by means of a warning light and a new set of ~ is taken to replace the r .; '' values within the vector.
Once the optical density vector has been linearized, the linearized optical frequency response matrix, DU), iSS '~I 'El, as indicated in an activity block 960. Control then passes to an end block 970.

Figure 9 is a - ' 61~ diagram which pictorially l~ul~..la the overall data flow used in auu,dance with the teachings of the present invention to obtain blood glucose co.,..~,..t, by means of optical signal ~
As shown in Figure 9, the light source 110 ennits light 115 which passes through the lens 117 and the filter 120 to provide filtered optical radiation 125. The optical radiation 125 passes through the finger 130 to provide an 25 optical signal 135 used to generate a signal intensity matrix 1000.
The signal intensity matrix 1000 is multiplied by the inverse of a filter cllalà..l~.iali~. matrix 1010 as indicated within a block 1005. As shown in Figure 9, the filter l,haldl,leliali~. matrix 1010 is derived from an analysis of the filter 120, as described above with reference to Figures 4A-4D and Figure 7. The inverse transform of the filter chcla~ ,.ialiL. matrix 1010 is multiplied by the signal intensity vector 1000 to obtain the optical 30 tll, ~ response matrix, or ll_ vector, 1015. The optical frequency response matrix 1015 is then linearked as described above with reference to Figure 6A-6B and Figure 8, to obtain the linearized signal strength matrix, or linearized optical density vector, 1020.
~ The several spectral ~,.,. 1i~:5 of water, glucose, OA~ - '', and SCdll~li as Il, tv.,l,l~d within the table 1025, are used to construct a blood s- ~ : matrix 1030. The inverse transform 35 is taken of the blood cc matrix 1030 to obtain an inverse blood : matrix as indicated within the CA 0222l384 l997-ll-l8 W O 96/41151 PCT~US96/08506 block 1040. The inverse blood ~ matrix is multiplied with the signal strength matrix 1020 to obtain a blood c times path length vector 1045.
A ratio is taken between the element of the vector 1045 indicating glucose c times path length and the element in the vector 1045 indicating water ~ dlion times path length as indicated wrthin a block 1050. This ratio results in the :- of glucose in water, which is , ~ . ' : to blood glucose, and is output as a data value ~ glucose ,, METHOD OF PRODUCING THE OPTICAL FILTER
Figure 10 depicts ' ~l~ the general method used in a~cù~d with the present invention to 'd'll~.L the optical filter 120. It should be first noted that previous methods employed to fabricate such optical 10 filters typically involved laying out a circular substrate and then s ' ~tiv~l~ h~ d~ the coating i' ' on the surface of the circular substrate as the substrate is rotated with uniform speed.
Such a filter 1500is depicted in Figure 13 as having coating layers 1502,1504,1506, etc., of Ih~ to form a spiral r .:- . as the filter 1500 is rotated. Of course, it should be I ' ~ -d that the coating i' ' es depicted in Figure 1 are ~, dlLd for ease of i" : dOua. This method of optical coating is 15 carried around ' 'l~ the entire .;.~ F~.uce of the circular substrate so that as the coated substrate revolves, the thickness of the optical coating grows i' . "' the entire revolution and then suddenly drops back from the thickest coating to the thinnest coating at the end of one IU.~ ' It has been found, however, that such methods of optical coating require high precision and are extremely costly. Furthermore, mal '~ these filters is typicallv carried out one-by-one, since r . methods do not allow for laying out several disks on a single sheet for mass ~"- ' : purposes.
In ? 1. d '~ with the ' i~my method of the present invention, a flat substrate 1100 is coated with optical coating of increasing thickness to form . . 19 c' ' coated layers 1110. Of course, it should be noted that for purposes of clearly i" "di' ~ the present invention, the thickness of the optical coating 1110 has been )9~, dl~d, and in practical , ' m the thickness of the optical layer 1110 will vary from 1.66 . t~a to about 3.331 I : a, with an average thickness of about 2.35 ~ a. Of course, it will be i ' . Qd that these i' ~ ' ~ are 1" UAillldl~ and may vary depending upon the index of refraction of the layer materials.
Such a method allows for d pr of a minimal number of optical coating layers. In one preferred e ' -' t, only 17 layers are n~ ~ ~ to obtain the desired rL '~ In one: b~ " , .Jlt~,l layers of high (2.20) and low (1.48) indices of refraction are deposited onto the substrate. Although the F.: h~y method of the present invention may result in less perfect filters than other more expensive ~ a, such can be ~ ' 3d in digital signal ,uluce~ail,y steps as described above. For example, previous filters typically pass a single ll" ~ band at a time, while the filter of the preferred . bc " : may allow for multiple bands to pass, since this is e ~ d for by the signal I I - ~ of the invention.
Once the optical layer 1110 has been applied to the substrate 1110, a .~; ' iLdl portion 1120 is cut from the . .~ pLd slab formed by the optical layer 1110 together with the substrate 1110. A ~.t' ' i.al aperture W O 96/41151 PCT~USgfi~

is then formed in the center of the cylindrical portion 1120 to form the filter 120 and, ' . ~I~, an optically opaque strip such as the brass strip 122is formed over a portion of the optical filter disk 120.
~ The above d~ i, " provides ease of illustration for '~ : ding the essential aspects of the invention.
However, it should be understood that the method may, in practice, involve first cutting the substrate into a disk - 5 with a shaft. Thereafter, the optical coatings are applied onto the disk as though the disk were still square so that the excess falls onto the platform (not shown) supporting the disk within the vacuum tank. In this manner the wedge is formed on the surface of the disk 120 as shown in Figure 10.
In one ad~ 19 embodiment, the production ~ ns for the filter 120 are as follows:
10 SIZE: 20 mm wide x 20 mm \, ~.. ' " span, linear multilayer coating SUBSTRATE: 25 mm OD glass disc with 7.5 mm shaft hole in center WAVELENGTH: 700-1400- ~
112 BW: 50 to 200 nanometers, bands may repeat BLOCKING: none 15 ENVIRONMENT: Survive ~ ' " humidity, 0-70 c The pass band edges are produced so as to "'r~ ial~ a 20 r band edge.
The pass band may repeat within the window at as little as 400 cm' spacing, or 17-18 periods within the window. The pass band center 1, should approach 100%, and the region between pass bands should 20 approach 100q~ ,t r' Blocking ll, I outside of the window are not critical. They may be limited by ~L~ I or band edge materials such as RG660, RG700, or ~ , or O-H bands typically found in glass below 7100 cm'.
Only the ability to resolve wavenumber bands near 200 cm1 with one or more band edges should limit the cost. The system is cost sensitive to the filter disc for ~,..dl quantities of 1,000 to 100,000 per year.
25 CHARAL I thl~llCS FOR PRESENT EMBODIMENT
Preferably, the filter will not have a window narrower than 8,000 to 11,000 cm1 or about 910 to 1,250 nm. The t '~: Rll is - N. ~ '~ wider than 200 cm-', and the band edge is ld~. 9 '~ narrower than 200 cm-1. The transmission maximum of the primary band is - Nàlli g '~ above 80%, and the tl minimum is3(1~ t~e '~ below 20%. Any other bands should be ll r~ , unit to unit; but if they are not, a c "' a:-30 ROM could be used in acc".d with the DSP to perform initial calibration of individual filters.
MECHANICAL BOUNDARIES AND CHARAL I thl~ I ICS
The linear filter is N " 1~ rotated about its center at less than 4,800 RPM, with an aperturecentered at a radius of minimum 9 mm to maximum 45 mm, with a clear aperture diamet er of 1 mm to 3 mm and a numerical aperture of .12 to .40. The light path passes through an annular region of the rotating filter, causing 35 a - ' ' scan of the ~ ' although they are deposited linearly.

=

W O 96/41151 PCTrUS96/08506 For dynamic balance and low turbulence, the linear filter is deposited on a circular substrate. Since the center is not used optically, a standard diameter shaft mounting hole is preferred; most of the present hardware in the invention use either 0.5000-.000, +.0005" diameter, or 7.5 -0.0 m 1 mm. For a small filter, e.g., 20 mm diameter, bonding to the uncoated side would be r ' ~d. Note that the filter mount does not have spokes or 5 other ;.ll. dl ~ ". :- of the optical path.
Initial optical-mechanical alignment of the coating on the glass is not critical beyond 0.5 mm and will be ' ' ' ' ek,..b. ~ 'I~. Some marking of the deposit alignment at the edge or center is desired.
OVERALL SEQUENCE OF nATA PROCESSING
The following is a summary of the sequence of steps used to process the data signals detected by the detector 140. As described above, the sequence of signal I ~ is extremely important for obtaining an accurate I Ir~ : of the patient's blood glucose level. S1s " signal, .~l ~, and analysis is wasted in previous systems which did not employ the proper sequence.
THE INSTRUMENT SYSTEM
The signal ~ ~i begins with light source and finger ~, properties. For the glucose sensor, 15 start with the received l,. . and perform inverse r,: , - on the signal until only the desired pail ~l is left. Note that the inverse J~ _~i - are only i 'Iy aL, dl~"L, not the inverse functions of analytical D Each operation is some l,. S which rc CL the data without -I ..N: or otherwise mixing the signal with the noise. The bll-lll l;oa content of the siynal should not be d t..;.~d or corrupted at any stage. Many s, ~r, , ;,.~lll fail this criterion. Thus, this system design is intended 20 to be unique in this field by being in the a~J~JIupliaL~ order, with a clear objective that avoids r~ , signal D, 1. TEST TRANSDUCER SIGNALS
Perform Signal and Noise Çl~ r;, ~I;U S~ signals correlated in model bounds are ,~ "
(with the same autocorrelation features) and ~ LI~ ~; Ih (random added noise with no - ~L l~lalk,.. features beyond 25 r S ~ , each new received spectrum is added to the data matrix block until the classifier detects a new noise :" ~R' or any ~ " change (detected by means of a statistical test) in the r,..." ' liun matrix of the 15 nonzero ~ a..' ~,lhs. When a change occurs, it indicates that the finger 130 has changed to a new position, or that a new finger has been inserted, or that a new patient is in the optical path. If block-matrix algebra is being used, then at this change, the system stops adding to the old block and starts creating a new block, at least 15 samples long, to perform statistical tests on data set.
If a recursive method is used, the recursion time constant (sample number) is shortened when the classifier detects a sample change until noise increases to a set s~tistir~' limit, or the block receives the minimum number of samples (typically equal to the number of . a~,-' Lhs resolved, 15 or 16).
Each answer displayed by the glucose sensor and computer will be the result of one block (of variable length), or of a length set by any recursion process's time constant. Whenever the patient changes finger position, fin~qers, or even patients, the system observes the change, and starts a new calculation of iial_ ~, including glucose e Thus ths patient and doctor can have e ' ' in the readout by knowing every time the li~,., chan~qe, the instrument generates fresh data. At the same time, this feedback to the patients encourages the patient to remain steady durin~q the r y analysis time. For the present signal-to-noise ratios, and for the 5 amount of time a patient is expected to remain steady, the analysis time should be on the order of 3 to 30 seconds.
When used as a continuous monitor, longer stable periods will be available, and avera~qes on the order of 30 to 180 seconds might be typical.
2. CLEAN-UP ELECTRONIC NOISE
Perform the first linear, _ to remove electronic noises: subtract linear f~.. e and average 10 random noise.
D~h. noise may be adapted to, and ' Id~ d due to a r ~ , periodic form and phase;
t ' - noise, such as impulses, start-up 1, ~ . or white noise spectra, have no, - ' '' r ~ . Model signals of a priori information become first 3~, I ~ ' and are loaded into any adaptive or recursive filters at start-up, and are used to predict signals through any impulsive or transient noise. This is the general form of a low-15 pass filter, but with the feature of averaging the shape of entire ~ in a heartbeat vector (e.g., a heartbeatplethysmographic (equal to pressure, volume) ~ 'L I as a vector with one aver age beat-to-beat period of systolic, dicrotic, and diastolic shapes). The noise in this ~ fo ~ is, ~f~..' 1~ averaged to zero.
At the detector 140, ambient light, powerline noise at 60, 120, etc. Hz, el~.,b....lali,., and magnetic fields, and ~ 9 waves are added to the signal photons coming through the patient finger. These noises are filtered, c ' ~d~ d or a~ ' to-zero before any non-linear stage is I,d.
When ail linear l~ ' ' pr are removed, only non-linear r~'e ' remain. The next operation is a non-linear ll_ 'l 1". S~ , linear n, ~- are r R", ~ again. The last step is a linear output of some ~Jalalll~l~l, so the preceding steps are a sequence of I i followed by linear a~ to uncover signals in ~ ~ ~I. and noise.
The industry standards use '_ht - I.e carrier: ' ' to get away from 60, 120 Hz ~50, 100 outside USA), and narrow band filtering around the carrier. Such a frequency filter does not remove out of phase noise.
Direct ~uLIIaLlion of a best fit noise model with matched phase removes all ambient i h,.
3. PREPARE LINEAR SOLUTION
. Log Tl. ~ The object of the o~ .: is to ;' ~ make it '~ , make normal; d~- " ' ' the desired signal from a of added, multiplied, or distorting noises, until only the desired signal remains.
If the signal is not linear, then it may have multiple values for the inverse. Then, no operator or process can ' ~ I~ the answers.

4. LINEAR ALGEBRA
Perform matrix algebra on Linear data: Standard, '" , matrix algebra solutions are used after all of the a! , " for linear " ' - ' and for y signal and noise di.. ll ' are satisfied.

W O 96/41151 PCT~US~6/085CC

5. TEST PHYSIQI l'C"'~L SIGNALS
Perform error detection~ . ' ' frequency L ' i~a, and correlate these with any a priori i.~tvrfv.~ to continue signal and noise v6aaifi 6. FEEDBACK
Feedback to !~ . Iinearization or I ' ~ ' ' noises: Subtract any linear i.~t~. ~v.. and average the random noise.

7. FEEDFORWARD
Apply clinical algorithms and generate derived pdl~ ~ a.

8. TEST CLINICAL LIMITS, ALARMS
Define the finger 130 in the optical path. Test if it is mostly water with large avdlll~ v . (about 30db lossllO mm is typically observed), check the pulse, correlate with previous records to identify the patient and the chdldLI~.ialivs of the patient's finger. Display "Not a healthy finger" or e, ' t, but do not stop operating since - ' ' tests may be done with in vitro materials. If any unusual Cv~ltddi' or signals are present, display ~, number of auxiliary pdl ' a as a test, and for user credibility.
15 9. IIO, COMMANDS, CLINICAL DATA, DISPLAY, RECORD
This portion relates to input keys, input data and output format. Glucose h ~ - as well as other pvl_ ~ a and historical records are input. Other . ' " may also require input of diagnosis,, I ": .
;..~lll : ms per ,L.~G"VIilJi' , etc. All of the clinical data required to manage a diabetic diet, such as the lessons from the Diabetic C~ ," : and Control Tests (DCCT), can be L.. 'Iy put into the program. This data may be ,...._ ' into a personal computer to minimize the sensor computer.
Although the preferred - bc " of the present invention has been described and " t~dl~d above, those skilled in the art will 3~ r l v;dlv that various changes and - "io ~i to the present invention do not depart from the spirit of the invention. For example, the principles and method of the present invention could be used to detect trace molecules within the '' '~ ~.. (e.g., for drug testing, etc.). In addition, a single near-log equation of 25 sufficient accuracy may be used to linearize the optical frequency response matrix. A,~ , the scope of the present invention is limited only by the scope of the following appended claims.

Claims (23)

WHAT IS CLAIMED IS:

1. A system for non-invasively monitoring blood glucose concentration within a patient's bloodstream, said system comprising:
a light source which emits optical radiation at a plurality of wavelengths;
a receptacle for receiving a fleshy medium of said patient;
an optical detector positioned to receive light from said light source and attenuated by said fleshy medium, said optical detector responsive to optical radiation of at least said plurality of wavelengths to generate an output signal indicative of the intensity of said optical radiation; and a signal processor coupled to said detector to receive said output signal, said signal processor responsive to said output signal to isolate portions of said output signal due to optical characteristics of said fleshy medium to provide a set of frequency response values, said signal processor having a linearization module which linearizes said set of frequency response values and to analyze said linearized data to determine the concentration of glucose within said patient's bloodstream.

2. The system of Claim 1, wherein said light source comprises a plurality of emitters, each emitter transmitting light at a selected one of said plurality of wavelengths.

3. The system of Claim 1, wherein said light source comprises a broadband light source, said system further comprising an optical filter which selectively passes ones of said plurality of wavelengths.

4. The system of Claim 3, wherein said detector comprises a single detector responsive to said ones of said plurality of wavelengths to provide and output signal indicative of the sum of the intensities of said ones of said plurality of wavelengths.

5. The system of Claim 1, wherein said optical detector comprises a plurality of detectors, each detector responsive to at least one of said plurality of wavelengths to generate an output signal indicative of the intensity of said at least one wavelength.

6. The system of Claim 1, wherein the linearization module comprises a double logarithm operation.

7. A system for non-invasively monitoring blood glucose concentration within a patient's bloodstream via spectroscopic measurement of attenuation of a fleshy medium containing blood, said system comprising:
a light source which emits optical radiation at a plurality of wavelengths;
an optical filter which selectively passes ones of said plurality of wavelengths;
an optical detector positioned to receive light from said light source which has been filtered by said filter and attenuated by said fleshy medium, said optical detector responsive to light of at least said ones of said plurality of wavelengths to generate an output signal indicative of the intensity of the optical radiation filtered by said filter and attenuated by said fleshy medium; and a signal processor coupled to said detector to receive said output signal, said signal processor responsive to said output signal to isolate portions of said output signal due to optical characteristics of said fleshy medium to analyze said portions to determine the concentration of glucose within said patient's bloodstream.

8. A non-invasive blood glucose monitoring system which analyzes blood glucose within a fleshy medium containing blood comprising:
a light source;
an optical detector responsive to light from said light source to generate an output signal;
a compression device which is configured to provide for physical perturbation of said fleshy medium to express fluid from said fleshy medium; and a signal processor responsive to an output signal from said optical detector when fluid is expressed from said fleshy medium and responsive to an output signal from said optical detector when fluid is not expressed from said fleshy medium to isolate information relating to the concentration of glucose in said blood.

9. A method of non-invasively determining blood glucose concentration comprising the steps of:
generating a set of values indicative of optical characteristics of significant blood constituents;
illuminating a fleshy medium having blood with light of a plurality of wavelengths;
detecting said light after attenuation of said light by said fleshy medium;
generating a signal from said detected light, said signal indicative of optical characteristics of said fleshy medium;
isolating components of said signal which are indicative of the concentration of glucose in water within said blood in response to said detected light and said set of values indicative of optical characteristics of said significant blood constituents; and generating a value indicative of the glucose concentration within said blood.

10. The method of Claim 9, wherein said significant blood constituents comprise water, hemoglobin, oxyhemoglobin and glucose dissolved in fluid.

11. A method of non-invasively determining a concentration of a blood constituent comprising the steps of:
illuminating a medium having blood with optical radiation, wherein the blood has a concentration of the blood constituent and the optical radiation traverses a path length through the medium;
detecting the optical radiation to produce electrical signals indicative of optical characteristics of the medium; and linearizing the electrical signals to provide an indication of the concentration of the blood constituent.

12. The method of Claim 11, wherein said indication of the concentration does not vary significantly with respect to the path length through the medium.

13. A method as defined in Claim 11, wherein the blood constituent comprises blood glucose.

14. A method as defined in Claim 11, wherein the linearizing step comprises the steps of:
in response to said electrical signals, generating a first set of values indicative of the optical characteristics of the medium;

taking a logarithm of the first set of values to generate a second set of values; and taking a logarithm of the second set of values to obtain a linearized set of values that are indicative of the concentration of the blood constituent.

15. The method of Claim 14, wherein said concentration of the blood constituent does not vary significantly with respect to the path length through the medium.

16. A method as defined in Claim 11, wherein said linearizing step further comprises the step of transforming the first set of values using a polymonial equation to provide a transformed first set of values.

17. A method of analyzing a concentration of a blood constituent from an optical signal incident upon a fleshy medium having blood, the method comprising the steps of:
detecting the optical signal after attenuation from said fleshy medium to generate a signal having a indicative of optical characteristics of the fleshy medium;
linearly subtracting components of the signal quantity due to electrical noise;
removing components of the signal due to non-blood constituents of the fleshy medium;
linearizing the signal to generate a set of linearly related values indicative of the concentration of the blood constituent; and solving the linearly related values for the blood constituent concentration.

18. A method as defined in Claim 11, further comprising the steps of:
filtering the optical signal prior to the detecting step with a filter having known spectral characteristics; and deconvolving components of the electrical signal due to the filter prior to the step of solving for the blood constituent concentration.

19. A method as defined in Claim 18, wherein said deconvolving step further comprises the steps of:
generating a set of values indicative of the spectral characteristics of the filter to obtain a filter characteristics matrix;
taking an inverse transform of the filter characteristics matrix; and multiplying the inverse transform of the filter characteristics matrix with a matrix indicative of the optical characteristics of the fleshy medium.

20. A method as defined in Claim 17, wherein said solving step further comprises the steps of:
generating a spectral library containing values indicative of spectral characteristics of components of the blood including the blood constituent;
generating a set of linear equations having coefficients based upon said values indicative of the spectral characteristics of components of the blood; and solving the linear equations to obtain a value indicative of the concentration of the blood constituent.

21. A method as defined in Claim 20, wherein the blood constituent comprises glucose.

22. A method as defined in Claim 20, wherein the step of generating the spectral library includes the step of observing the spectral characteristics of a known concentration of glucose dissolved in water or blood to determine the spectral characteristics of glucose.

23. A method of manufacturing a rotating optical filter for use in blood glucometry, said method of manufacturing comprising the steps of:
producing an optical substrate having a top and a bottom; and depositing layers of optical coatings on said top such that said layers vary in thickness over said top of said substrate in a first direction and such that said layers remain substantially constant in thickness in a second direction perpendicular to said first direction.