CA2251919A1 - Improving radio frequency spectral analysis for in vitro or in vivo environments - Google Patents

Abstract

Concentration of a target chemical, glucose, in the presence of other substances, NaCl, in a specimen (4) is determined by subjecting (2) the specimen (4) to radio frequencies (6, 16) up to about 5 GHz. The real and imaginary components of the reflected and/or transmitted signal are examined (18) to identify the presence and/or concentration of the chemical of interest. The examination includes analysis of the effective complex impedence presented by the specimen (4) and/or the effective phase shift between the transmitted and reflected signals. The effects upon glucose concentration measurements of NaCl can be nulled-out by examining impedence magnitude at a cross-over frequency or measuring NaCl concentration in a first frequency range and subtracting from a combined glucose/NaCl concentration measurement in a second frequency range. This technique can be advantageously used by diabetics to measure blood glucose level.

Description

CA 0225l9l9 l998- lO- l5 WO 97/39341 PCTtIB97/00719 IMPROVING RADIO FREQUENCY SPECTRAL ANALYSIS
FOR IN-VITRO OR IN-V~VO ENVIRONMENTS

RELATIONSHIP TO PREVIOUSLY FILED PATENT APPLICATION
This is a cf~ntinl~fion-in-paft of patent application serial no. 08/103,410, filed 6 August 5 1993 entitled APPARATUS AND METHOD FOR RADIO FREQUENCY SPECTROS-COPY USING SPECTRAL ANALYSIS, now U.S. patent no. 5,508,203.
FIELD OF THE INVENTION
T'nis invention relates genera'ly to radio r~ u.,.l~ ~e~hos~yy7 and more particularly to ~ ylvvhlg srP~fif~ y and accuracy of such ana'lysis to ff"~t' ~ f the presence and/or 10 collc~ lalion of a desired chemical among other s~lbct-~7lres within a i~yer;... --~
BACKGROUND OF THE INVENTION
Many cvll~ iollal ana'ysis terhniql~ps measure the co~ " f ion of a chemical in a test ,.ye~ ;,... n or sample, even where the ~ye~ ;.. contains a complex rni~ture of rhPmir~lc Such tP~hniqlles include mass ~p~llvihOtc~ , nuclear I~S ~ ~ flame photometry, 15 c~ u~ re and lef., 1~ y. While these tef'hniqu~Ps work, lmfvllu~ t~ ly, their accuracy is too often directly related to their cost. Further, many such tefhniq~lPc alter or destroy the ~ye~h~ n under test, and require relatively ela'ovl~.t., e.luiyllle~

More recently attempts have been rnade to d~ ~e .-; .~ various ylvy~ ,s of n~'P~ s~ using sound, ele~,~lv ~-En. I;( waves, or single pulses as the basis for analysis. In contrast to 2 0 conventional chemical analysis, wave and pulse-based tPrhniqlles can provide a non-invasive m-vlvo analysls.

- For PY~mplP~ U.S. Patent no. 4,679,426 auly 1987) discloses a non-invasive in- vivo tech-nique for I~lF~ ;--g collcl..llld1ion of fhP-nir~ sodium chloride for e~ample. Periodic - ele-~;llv.. ~a~ waves having a repetition rate of about 10 MHz to 100 MHz were coupled 2 5 to a subject's finger, and sodium or chloride ions within the finger ~ ly distorted these waves. This distortion in the c~-. l-f~ e wa~,fullll was received from the finger, using the same ele.,liode-antenna pair used to couple the waves to the finger. The collq~fJ~;le ~a~.,rullll distortion was then eY~minPA~ and found to provide mP~ningfill data as to chemical c.,... ~..11.. ~ionc, Glucose is an especially hll~OIL~ll chemical, a knowledge of whose absolute conc.,lllldion 5 level can be vital to Ai~hetirc Several techniques for providing blood-sugar analysis are known, which permit subjects to Af t~ f their own glucose levels. U~lru~ t~_ly many such terhniques require invasive sampling of the subject.

One non-invasive tPrhn~ e for det~,.ll,h""g glucose levels in-vivo was Ai~losed in U.S.
Patent no. 4,765,179 (August 1988)in which a periodic train of ele~,l,u...,.E,,. lif~. energy, 10 preferably having a repetition rate of about 1 MHz to 1 GHz, was coupled to a subject's finger. The colllpo~ wa~.,rullll distortion was then analyzed and found to provide mean-ingful analysis of glucose levels in the range of about 50 to 150 mg percent. However, beyond about 110 mg percent, it was desirable to fine-tune the ele~llu,- ~g~ energy to maintain ll,ea~ulf .ll~.ll accuracy.

15 Ul-Ae~ A l~ly, blood is a complex solution. Mvnilulil.g the conce.~ ioll of glucose in blood presents ~uI,~alllial challenges to ~ e against other ~.~I,sl.~ ~~es in the blood that may mask or alter the analysis results.

U.S. patent no. 5,508,203 dPsrrihed a non-invasive in-vivo ai~lJalalu~ and method for deter-mining a chemical level in a subject, i~l~'.lu-l;ng the chemical glucose. The use of fre-2 0 quencies up to about 1 GHz was Aisclosed and the AicrloseA ~~ dlu~ p~ d even laypersons in~hlAing diabetics to d~ t~ r~ for ~ , '-, the level of glucose intheir blood system.

As useful as the in~_."ion Aicclosed in U.S. patent no. 5,508,203 is, applicants have since realized that electrolytes, e.g.l NaCl, KCl, Na2HPO4, and KH2PO4 of varying Cull~lll~-2 5 tions in human blood can affect the accuracy of glucose ,lle~u,~ using that hl~ iu,l.In the human system, glucose c. ll~ c typically range 60 mg/dl to about 150 mgMI
for a non-diabetic, and range from about 50 mg/dl to 500 mg/dl. In the human population, NaCI co~ nc can range from about 135 mM to about 145 mM. To err~lively and ccnfiA~ntly measure glucose and/or its collc~.llldlioll in blood, a rPcth-ti--n of about 10 3 0 mg/dl of glucose is desired. Non-invasive so~ ul~lo~y grade test C.lui~ lL can resolve glucose in-vitro to perhaps 1.5 mg/dl. Invasive co,'~ ~..l .-grade can resolve glucose to perhaps 5 mg/dl with an accuracy of perhaps :t10%. Ayylicalll~ are not aware of e~cisting non-invasive devices for resolving glucose to the desired 10 mg/dl level.

There is a need for a method and apparatus to reduce the varying COIlcf l~rdlion effects of -electrolytes, especially NaCI, when mPs~ rine glucose concf.ll,dions in human blood.
5 Such method and apparatus should be useable in-vitro and in-vivo, and should work in non-invasive in-vivo ~--. a~ul.,.nent c,~vhu..."~nL~. Further, such method and ~ y~la~ should be capable of use by lay persons. Such method and a~)aldtus should also have applicability in IllCd~ul~ unrelated to analysis of bodily fluid, inrhl~line applications in industry.

The present invention discloses such a method and al,y~ ..c SUMMARY OF THE INVENTION
A ~ye~ . cu~ls~ e a chemical of interest as well as other substr ~r-Ps is via probes s.ll,;e~,lcd to radio r,l,.l-.."l-;y ele~,L~ ag.~ Iic signals having high frequency co-"yonc~"~ ex-tending to perhaps 5 GHz. Preferably such LC~ PC are sequPn~ ly pl~,sen;~ using one Sinl,~.av~ r~c~u~,n~,~/ at a time, although Cim~ltr~Poucly yl~s~lted multiple r~ es may 15 also be useful. Reflected and/or ~ . d signal real and i,l,agin~y CO111lJO~ S at the syc~ are then spectrally f'~l5~ 1 as a function of r,~ ue.,e~ to identify the presence and/or co~ce.,L~dflon of the chemical of interest. Such e~5 ~-;-~ n includes analysis of the effective complex ;..~ e p~ h~d by the ~-pe(; - -, and/or effective phase shift between the 1-.~ and reflected signal at the ~-e~ ;~- - - . In this manner, greater speci-2 0 ficity can be attained with respect to detertine presence and/or conc~ flon of a desiredanalyte or chemical of interest.

For in-vitro ~ le.-t~, a probe is inserted into the ~l-e~ - ~ and is coupled to a net-work analyzer, or similar electronic system. In such in-vitro Ill~,a~lllc.ll.,~lS, the ~l,e~
may include blood or other bodily fluid, or may be a s~lbsts~ re unrelated to bodily fluid.
25 In in-vivo lll~,~ul~,....,.ll~, a network analyzer of similar cle~,l,unic system may be coupled to cle~,l,ùdc(s) on a probe. The probe is pressed against a subject's body, preferably a finger, and non-invasive analyses are made.

Applicants have discov".~,d that variable co~c~ innc of electrolytes, especially NaCI, affect accuracy and cp~Pcificity of glucose co~ a~i()n ~..C~u,~ . At r,~l..- ... ;-c 3 0 below about 1 GHz, increasing NaCl (or other small ions) conc-.,.,l,alion dec,cases imped-ance, whereas at higher fi~ s the imre ~ -e is increased. Applicants believe that at the lower r,- ~lu- ~~( if S, ions can respond to the rh ~leine ele~l~u ~, - field adjacent to the probe ends, whereas this is more difficult at higher r~ u~ - ;Ps, whereat water dipoles WO 97/39341 PCT/lb57~719 .

appear to largely dt~terrnin~ imre ~ ~e. In general, applicants have learned that over a wide Lc~lu.,ncy regime, higher glucose Col C.,.ll~dtiOIls h~ ses i",l.e-1~ " e, probably because the large glucose molecules hamper l,.u~ ...,ul of electrolyte ions and water dipoles in a solution ~l-e~ . Of special interest, applicants have disc(,~ d that increasing NaCl 5 conc~ -l-alions over a wide L~ ~lucllcy regime increase phase shift in a linear fashion, which phase shift is hl~e.l~iliv., to glucose conc~ dtions. Using these disco;~.ies, applicants can null-out or at least reduce or co",l, - ~ for electrolyte c~ m effects upon glucose con-;c~l~ldlion by using cross-over L~ s and by eY~ nining different Illca~ ,nl pa.a ll~..,.~ at different frequency regimes.

10 In a blood ~ I, electrolyte conc~"llldlion effects are erf~ ,ly "tuned out" by eY~nnining the m~ni~l~de of complex impe-l~nre using a cross-over L~u~,..~ of approx-imately 2.5 GHz. This use of a cross-over f ~ U.,.I~,~ and complex ~ A~ C ~.lea~ul~,..l~,..
provides low s~ ilivily to NaCI co~ n and thus more accurate and specific glucose cv..~ ;.." readings. Such analysis hll~ùvcll..,nt can be highly hll~ullalll, for e~ample 15 when the ~ comes from a diabetic or, ~l-e~l~d diabetic.

Differential analyses may be made by collll.ilullg ;",l~eA~ c g ~ and phase shift IllCd~ l data. For example, high L~ u~ ,y phase shift Ill~,a~ul~,.ll~,.ll~ taken between 2 GHz and perhaps 5 GHZ can provide data plol)t,lliol al to magr~ of ion co-u~- ~llld~;(lll, 2 0 particularly NaCI. On the other hand, i",l-cd ~ e q,, ~ ~ulcul~ s made usinglower rlc.lv ~.-~ , perhaps the 1 MHz to 400 MHz range, will provide a measure of cu...~hled con~.llldliol of glucose and ion cul~-,...l~lion, again primarily NaCI. The high L~;~u~ phase shift data may be used to subtract out the effective NaCI conc~.llldlion from the lower Lc~lucll~;y , .' ~~ total conce.llldli.)n data. The result is a lower 2 5 r~c~lu~, ~y measure of glucose conc,.ll.dlioll in the sl,e~; - -, a L. ~lue~lcy regime in which c~vl~ .ll e.luip u~ll is quite sensitive.

Analysis e~ ."-P-~ coupled to the i,..pc~ -c l..ca~u..,...~..l data and phase shift measure-ment data can include look-up tables or the like, correlating phase shift data to NaCI
co~ ..l.dlion levels. For in~ ctr~ rPI nc~ the look-up tables can store data correlat-3 0 ing in~reA ~nr~, phase shift and Lc4u~ y ....,a~u.~ to known ~ ;--,. es and con-c~ n levels. This h~rul~aliol can then be used to enhance nulling-out of NaCI in an i",l.cA-- rc l..c~u-c~ nl made at a cross-over L-~lu~ ,y.

_ Output inAir~tor~ coupled to such analysis e.lui~ ..l can enable even a lay user to readily un~l~r~t~n-l what chemical has been detected and at what collcelltld~ioll, or simply to confirm that a safe cvnc~ lation has been detected for the chemical of interest Ot_er features and advantages of the invention will appear from the following d~s( ~ iull in 5 which the preferred c..lbo~ have been set forth in detail, in co..ju...;Lion with the acco...l~lying drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
FIGURE 1 is a block diagram of a radio rl~lu,n~;y ~l~c,t~u~cv~y system;

FIGURE 2 is a block diagram of the ~ ,/receiver-signal p-uce~ lg system 14, 10 shown in Figure 1;

FIGURE 3A is a srh~m~tir of the calibration cell 66, depicted in Figure 2;

FIGURE 3B is a Smith chart i.~ re versus frequency ~ ion of the equivalent circuit depicted in Figure 3A;

FIGURES 4A, 4B, and 4C depict signal amplitudes provided by the system of Figure 1 fûr 15 different target rh~ lc in analyte test 50lutinn$;

FIGURE 5A depicts an in-vitro application of a radio L~ u~ Je~;l.ùsco~,y system with .on~ analysis s~,.silivi~y, according to the present invention;

FIGURE 5B depicts an in-vivo ~rplir~tinn of a radio f~ u. .I~;y ~Je~ llvs~y system with ~h~re~ analysis .~ ivily, accù-di..g to the present h~ io..;

20 FIGURE 6A cv...~ non-invasive and invasive ;~ re ~ a~~ 1c test data for a subject, using a test confi~llr~tion accvl.li--g to Figure 5B;

FIGURE 6B shows correction for electrolyte dilution for the sam data shown in Figure 6A;

FIGURE 7A depicts linear rel~tinnchip between ele~;llùlyl~ cullc~ lion and phase shift, in~ y of glucose c~ ,l . r~

2 5 FIGURE 7B depicts linear r~ between electrolyte co r~ m and phase shift in a PBS solution, ;, ~ ly of glucose and/or albumin c~

WO 97/39341 rCT/IB97100719 FIGURE 7C ~If .~ ! how h~lo~d ~ c;l;~ ly for a target analyte can be realized by.1i"g ull,aDul~lu~ D that are h~se.lDili~,., to a C~J~ l in the Sl,CC;... ~ for P~mplP, phase shift ~ Dul~e.ll~.llD at 1.5 GHz to null-out albumin COllC~ ld~iOll;

FIGURE 7D depicts a phase cross-over Ll~ u."lcy of about 20.1 MHz whereat phase shift 5 data is in~ of glucose conc~ ld~ion;

FIGURE 8A depicts the increase in i---perl~ ~e IllcaDuled from about 0.1 MHz to about 1 GHz with hlcl~a~illg glucose conc~,nlld~ion;

FIGURE 8B depicts a rlcyu"lcy regime in which increasing NaC1 and glucose CQ!~f'f~
10 tions increase i...l,c,~ e;

FIGURE 8C depicts a rlcu,u~ y regime in which in~,ltaDiug NaCl concclllld~ion does not suhct~nti~lly affect impetl~nre, but h~cledDing glucose conc~i,ll,alion illcl~ i...l.e~ e;

FIGURE 8D depicts a 2.0 GHz to 2.1 GHz rlc~lu~,l~ regime in which illClCa~illg NaCI
conc."l~lalion dccl.,a3es i...peA~ ~e, while h~ LDing glucose conc~i.llld~ion hlclc~cs 15 ;..I,e~--..c reasonably linearly;

FIGURE 8E depicts the non-linear behavior of i.--l-e.~ e ~ g~ P data over a 2.25 GHz to 2.75 GHz rlc.lu~,l.;y regime as NaCI conc.,lL.dlion is varied, accc.lding to the present invention;

FIGURE 8F depicts the ~ e of a cross-over rl~ ue.,l,y at about 2.5 GHz at which 20 NaCI co" ~ ;on effects upon lucaa~lcd i..~l.e~-~ are nulled-out;

FIGURE 8G depicts rlc.l.. .1~ versus ;~ e changes for a ~l~ec;~ cc l~;..;ng various D~l.DIh.-~es, and d~ l s a possible gamma globulin ~ region.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT
Figure 1 depicts a radio rlc~u.,n~ ("RF") D~Je~ D~lJy system 2 for d~....;..;.~g the 25 presence of one or more target chPm ~~l~ (depicted as x, y) in a cell lU,/~lal~c ",eci...- -- 4, e.g., a human finger. The D~,e~ -;.. - ~ finger 4 is pressed against a probe pair 6, plcf~,.dbly disposed within a concave d.,~ sion 8 formed in a lucite base 10. Probe pair 6 CC)I11~J1;SC
two con-lu.,li~., rods that protrude slightly from the depression 8, p ...;~ g elect~
contact to be made when finger 4 is pressed against the rods. Preferably the rods are 3 0 brass, perhaps 0.2" (5 mm) outer diameter and protrude outward from the concave surface . . .

about 0.05" (1.3 mm), and into the lucite base about 0.5" (12 mm). Of course other tissue could be probed, e.g., an ear, and the spcc;.~ ~ need not be a human.

A pair of llA~ ,ion lines 12 elecllically couples the ele.,lludc pair to a system 14 that includes a ~ unit 16 and receiver-signal plocessol unit 18. Briefly, unit 16 5 Lla~ s a high rl~,4uen~,.y signal via ll~ Cm lines 12 to probes 6, which couple the signal to the ~l,ec; - fmger 4. Although the precise mPrh~nicm is not fully lln~lerstood, it appears that the presence of target rh~ c, e.g., x and/or y, within the ~l~cci.", may cause energy transfer of certain spectra of the source signal from 11~ - 16. The result is that a return signal from the ~l.c~ ;.... n. present at probe pair 6 and coupled via transmis-sion lines 12 to unit 18, differs from the sûurce signal. Of course separate probe units 6 could be used to couple the lr~ unit 16 to the ~ c;~ and to couple the return signal from the ,l,e-~ . to unit 18.

Unit 18 receives and plOC.,sSeS the return signal such that spectral signatures ~Csori- ~ with the presence and con~,.,.llld1ion of various target rhemir~ within the ?l,e~ can be 15 I-,co~l~d. The plu~,css~ data is then coupled to a display system 20 that conveys the detected inrolllldlioll to a user. Operation of the receiver-signal IJlUC~o;~OI unit 18 can be tailored, manually or ~ lu~ lly by a neural network, to I~Oglli~ specific target chemi-cals, for e~ample glucose within the blood stream within finger i.~CC;..~ ~. 4. In such instance, the various output devices within display system 20 might provide a user with 2 0 calibrated data as to his or her glucose concclllldtion.

Display system 20 may include a monitor that can display a S~)C~;Ilulll analyzer output (22A), and/or alpha-numeric/gla~hical output (22B). Display system 20 may also include a bar graph or alpha-numeric ~ 24 inflir~ing, for e~ample, the co~ nlr. ~ion level of the target chemir~l, for example, glucose. A calibrated output meter 26 could provide the 25 user with cn~ ;o~l data. All~lldi~,ly, a simple "GO/NO GO" output hldi~.~tcl 28 could alert the user that e~cess glucose cc~ on has been detected. A diabetic user would thus be a1erted to take insulin immf~

Figure 2 is a block diagram of the l ~ ,ce;~_.-signal plo~sDillg system 14.
- Oscillator 50 g~ r~ a high r,c~lu~ cy stimulus signal that will be llA~ lled via probes 3 0 6 to cl.c~ . 4. In the pl.,f.,.lcd c.,.l)odi"l~ t, osci~ or 50 provides a 30 MHz funda-- mental square wave having a 50% duty cycle, and l.,~ ;lin,~ times of a few n~ eC~ c As such, the osrill~ ~r output rl~lu~.lc,y ~ .,IIulu will be rich in ~ s, the odd-I,u,~.,.~d h~rmfmirs p,ctlf~ g. In the Llc.~ ~ domain, a perfect square-wave CA 02251919 1998- lo- 15 WO 97/39341 PCT/IB97tO0719 source signal would have h -mt)nirS with a sin(x)/x c..~,lope, where x lcl~lcs~ s a h~rml-nir frequency.

The spectral output of such an osc~ t--r 50 is commonly referred to as a comb spectra, as the various spectra are uniformly spaced similar to the teeth on a comb. The power output 5 level at the oscillator output is preferably about 1 mW, which is O dBm, although other power levels may also be used.

In the p.,_f .-I,d embodiment, the various source signal spectra are h~lllùnically related since generation of a pulse train provides the h~ ;r r~ u~- ~ ;FS ~ ;c~lly However the source rr~lul --- ;F C need not be harmonically related, and a single oscillator 50 may be 10 rapidly changed between discrete frequPnripc (e.g., in the manner of spread ~,.,e~;l,u.,.
L~ ). A~ lllalively~ oscillator 50 could CO~ ;SC. a plurality of signal genc.dlo~
whose separate frequency outputs may or may not be h~ lly related. If h~ ically related, one such generator could provide a sinusoidal output at a filn-l~ nPntql rl._.lu~"~;y, e.g., 30 MHz. A second ~;.,.I.,Idl~l could provide a 60 MHz sinl~coi~l~' output, a third 15 gell.,.a~ol could provide a 90 MHz c;mlcoi~lq~ output, and so forth. In a different e.l.bodh..~ L, one such 6~ alOI might provide an output at Lc~tu~ ;y fl, a second genera-tor might provide an output at ~f2, not h- ,..nn.f~lly related to fl, and so forth.

As used herein, oscill.~tor 50 is lm~lPrstood to be a source of ele.;L.~ ir signal that contains a plurality of high L~ u~ ;y c4...l.ù~ regardless of whether such co...l.ol.l .
2 0 Ic~,.es~,.L h -- ~ .irS of a single source L~ u~ ;y, or le~ ,se.ll many source L~ tv n- ;~ s, that need not be h-- ~.. i~ ~lly spaced-apart.

Unit 52 preferably includes an qrnpljfi~r stage and a power splitter, and co,..~ ;s a MAR-3 ~~r~ifiPr and a Cougar amplifier stage and a power splitter in the pn~f~ cd c .-l~.odh.l.,nl.
These Cul~ ul~ ,;ally available cu. .~ ,t~ boost the oscillator signal provided to divider 54 2 5 to about 15 dBm, and provided to power splitters 62, 64 to about 3 dBm. In turn, each power splitter 62, 64 divides the thus ampliLed signal into two signals at nodes A and B, each having O dBm power output. Splitters 62, 64 are pl~f~dbly wid~alld, e.g., about 10 MHz to 1,000 MHz (or 1 GHz).

The i--l- ."1~.1;, L~,~tu~,-lcy ("IFn) for system 14 is preferably 21.4 MHz, an i~lt~ ' L~ u~,.lcy co-...... ~ly used in co.. ,.. ,ial e tui~ ,ll, for which rrc~ ;y many stan-dardized t.,-.~r~,.. ~ and circuits are readily available. High-side mixing injection plcf~,ldlJly is used. Thus, to generate a local oscillator r~ u.,.l._y that is 21.4 MHz higher , . . . ~ .

_g_ than a center rr~u.,.l~;y, it is necessary to develop a s~ l.r~ d leÇ,l~ e 6.4 MHz signal.
Unit 54 divides the f~ln-l~m~nt~l L~,~ue.l~ of the oscill~tor signal by 6, to yield a nominal 5.0 MHz ~ciç~ ce signal.

This 5.0 MHz l~f~,..,.~ce signal and a 6.4 MHz phase-locked crystal controlled osrillr~nr signal 58 are yloccsscd by offset module 56. Offset module 56 outputs on line 60 a signal having a frequency of 6.4 MHz that is phase locked to the 30 MHz frequency of osri~ or 50. Because phase lock loop systems are well known in the art of digital signal pluce~mg design, further details of the gen~ ion of the L~.lu~,n~ locked 6.4 MHz signal on line 60 are not yl~ ,d here.

In Figure 2, calibrator unit 66 is an electronic model of a typical human finger, essPnti-1ly the electronic equivalent circuit of a finger .l.e~ 4. While calibration unit 66~lyylO~ at~s the .~l.c~ i...l.cd~ .e, unit 66 will not include the target chemical.

Figure 3A details the circuitry within calibration unit 66, namely two seE;."...,l~i of ,,,;c :o~ line having 50 n i. ~.eA ~ e at 400 MHz, and assorted resistors and capacitors.
15 The ~ ",;.~jon lines, resistors and c~r7~il0, ~ were selected c ~,p;.i~lly by co...y~
L~ u.,.l~,.y versus i...pe~ e data from human fingers with data from the equivalent circuit of Figure 3A. Figure 3B is a Smith chart i...l~e~ e versus L~ U.~.J~ l~ 7~ of the e~luiv ' circuit of Figure 3A. Point A in Figure 3B ~ ,e.lls an i.. l-eA ~ e of about 192 n/-201 n at 10 MHz, point B is 39.5 Ql11.5 n at 300 MHz, C is 52 n at 400 MHz, 2 0 and point D is about 57 S2/-2.6 n at 500 MHz.

With further l~,f~.~nce to Figure 2, as will now be d~o~ ~ ;btd, various c~-. l-- n. l!~ are r~li-; ~d to provide a pluce~hlg path for the tl .C...:ll~ d source signal, and to provide a plocessing path for what will be termed the sarnpled return (or received) signal. The 2 5 sampled return signal adv-~lac~u~ y permits c~ the system of Figure 2 for Colll~o~ lll variations and drift between what will be termed the received and the transrnit-ted signal plucejsing paths.

More specifir lly, the response of s~,e~ - . 4 to the source signal (e.g., the return signal at the probe pair 6) is switchably sampled by switch Sl with the response of the calibration 3 0 unit 66 to the source signal. H:~n~-ni- L~lu.,-lcy-by-L~-~u~,.l~, the output from probe pair 6 and from calibration unit 66 are sarnpled, the output of Sl providing a sampled return signal at node C to the l~ of system 18. Of course, if source oscillator 50 provided discrete rl~l.. ~-- i~ that were not h~ lly related, it is understood that L~lu~,.. cy-by-r~ u~ ;y, the output from probe pair 6 and from calibration unit 66 would be sampled. In the l,l.,f~ d embodiment, the r~ u~,.lcy bands of interest begin with about the si~th or seventh h:~rrnt-nir of source oscillator 50, e.g., about 195 MHz, and extend to about 1 GHz, or higher, which range is the bandwidth of system 18. Within that ban.lwid~l, 5 h~livill~al r~ FS are sampled between probe pair 6 and calibration unit 66.
Switch S1 pl~,f~,~ably is a co.~ ;ally available monolithic Illi~;lUWaVe ill~ '(' circuit ("MMIC"), a relay, or other s..il~Lhlg ' Sl switches between the probe 6 output and the calibrator under control of a l~ i-;lul)lucesso. 74 within system 14. In the ~cr~ d embodiment, llhclu~.oc_ssûr 74 was a Motorola 68HCl l, although other 10 mic-u~-ùccssol~ could be used instead.

S1 may sample the output of probe 6 for a time period ranging from perhaps 30 rns to perhaps 7 seconds, and then may sample the output of the calibration unit 66 for a time period also within that range, the duty cycle typically being areri- Air. For example, during the time Sl is coupled to probe 6, the probe output signal is sampled for one or more 1 5 r,c~ s that are h-- ---n--;- ~ of the funA~ nFnt~l rl~lu.,.l~;y of osc~ or 50 (or for one or more discrete ~l~qu- .r;~Fs provided by an oscill~or 50 that does not provide h-..nonil~) During the time Sl is coupled to the calibration unit 66, the response of calibration unit 66 to one or more r,~lu. -- ;FS that are l ....~ of the r.l~ A ~ l ' osr~ r 50 r~e~u~,, are sampled.

UnA~ A~l)ly~ if cn~ ol~ ~lc 76T and 76R, 78T and 78R, 80T, 80R, 90T and 90R (to be ~1~F~ ed) were identical and e~hibited no drift, CàlibldtiOn unit 66 could be ~ d with, and Sl replaced by a wire making a p~ co~ ~F~-l;O,~ in the probe 6 Sl position. Such an ideal system would require no ~F. 1~ for CO...~ i..g for drift and other dirr~ .nc-es in the signal plu.,es~ g paths for the h~ of the oscillator signal 50, and for the 2 5 h ~ in the return signal obtained from probe 6.

In practice, variations in Lc~ cldul~ and/or pressure between probe pair 6 and the tissue in the sl,e ;.~ ~ 4 may collllib~ some error to the Illc~ul~,.ll~,.ll process. To perrnit lUiClUIJlu~;.:~Ul 74 to COl r for such error, in addition to providing the ~ u~lo-cessor with phase and ~-nrlit~ F h~r ~ for } - ....~ , phase and ~n~r!itl)de informa-3 0 tion is also provided for the osrill~tor fUn~l~nlFnt~l r,~lu~l.~. This r,~ " has beenfound e,~- ;".. .~lly to be sensitive to such le.~ ,.dlul~ and/or pressure variations. It is understood that suitable ~ .dlUI~; andlor pressure ~., c-l~ and analog-to-digital cull~ iun c...~ v--~ that are not shown in Figure 1 are used.

As shown in Figure 2, within the ll~ . d source signal prûcessil~g path, a bandpass filter 68T has a center rr~ n-:~r equal to that of oscill~ r 50, e.g., 30 MHz, and a bandwidth of about 1 KHz to perhaps 1 MHz. Other bandwidths could be used and in fact, a 30 MHz lowpass filter might instead be used. The ~ d signal from node A is5 coupled to b~dl ~Cc filter 68T, and the 30 MHz center r~ u~ ;y cc,lllpoll.,lll of this signal passes from filter 68T and is amplitude limited by lirniter 70T. The thus b~-~dp~g filtered and amplitude limited signal is coupled to an input of a phase detector 72.

In a parallel path, the sampled return signal from switch S1, present at node C, passes through a sirnilar 30 MHz bandpass filter 68R,~rnplinl(1e limiter 70R to provide a second 10 input to phase detector 72. (The letter T or R attached to a ,~,f.,..,..ce element herein denotes that the element is used in the ~ llcd source path, e.g., 68T, or is used in the sampled return signal path, e.g., 68R.) Phase detector 72 culll~ es the dirf~ nce in phase between the ll ~ 30 MHz r".~ Al rlc~lu.,n~y and the sampled return 30 MHz fnnf~m~nt~l L.,~ue..~iy signal. The 15 phase detector 72 output signal voltage will be plupolliùllal to such phase shift, e.g., a number of mV per each degree of phase shift. As shown in Figure 2, the phase output illfiul~ion from detector 72 is coupled to llf~clul~rucessor 74 for analysis.

~u~elillg h--.;"~ Ally across the top of Figure 2, parallel paths are also depicted for ~lucessh.g the i itt~d source signal h- ",n~ (available at node B) and the sampled 2 0 return signal ~ ~s from switch S1 (available at node C). These two h-,. i,....lrl paths use S~ a; ~ y identical co..~l.or~ (as denoted by the n~ c) to provide d and sampled return signals at an i"ll . ",. .I;- ~e L ~.lu.,.l.,~ (IF) that is about 21.4 MHz in the p-~,f~ d l ...ho~

Briefly, the CG--~I~u~ now to be de,~ ed resolve the h~ -nir L ~ U~,Il~ co..~ of 25 the signals at node B and node C into preferably four bands of discrete L~u. r- rc, de-pending upon what h- ,..nl-ifs of the source oscillator signals are desired to be e~minrd Much of the l ~~ of the signal l,loceisor r . ~ as a scanner-receiver, that under icluplucessor control scans discrete h~ ir L~ u~ 5 of interest. The l..- .~...;i~. d source signal path cc,...pul~ will first be dc~ ,d, it being u~ ood that identical 3 0 cu~ ul~ ; are used in the parallel sampled return signal path, as in~ir~ed by the nu--.- --rl ~ , e.g., 76T, 76R, 78T, 78R, etc. R~n~1p~ s hlter 76T (and thus also 76R) preferably is a filter bank that includes an internal MMIC ~ illg m~orh~ni~m o~ dlillg CA 02251919 1998- lO- 15 WO 97/:~9341 PCT/IB97/00719 under control of Im.,lul)roceiOol 74. The input port of filter 76T passes the ~ ...;u~ d sig-nal from node B through an internal switch into two banks of pre-shaping three-pole filters. These first two internal filter banks have b~n-lp~cc~s of 195 MHz to 395 MHz, and 395 MHz to 805 MHz. Still within filter bank 76T, the outputs from the 195-395 MHz and 395-805 MHz filters pass through ~ ition~l internal MMIC switches and b~ filters. These ~ litinnPl filters pass 195-295 MHz, 295-405 MHz, 405-610 MHz,and 605-815 MHz. Still within 76T, the variously filtered colll~c._llls are colll~illed into a single signal that is ~mrlifiPd by amplifier 78T.

In similar fashion, the sampled return signal at node C is passed through s..ilcl.hlg 1 0 b ~ filters within bandpass filter bank 76R, and the variously filtered copo~ are cc,llll,hled and amplified by amplifier 78R. While the operation of b~n-lpacs filter banks 76T, 76R has been ~Pcrrihed with lcfe~ ce to specific frequency bands, those skilled in the art will recognize that the frequencies COlll,~ lg the signals at nodes B and C may be filtered using b~ filters having different ranges of b ~p~ c. Because the design of 1 5 units 76T, 76R is known to those skilled in the relevant art, 5~h.om ~irS are not here provided.

For eY~mrlP-, if the target chemical of interest is best resolved by ~ss~-njnjng say the seventh h_. ~ ;r of the 30 MHz l" ~c ";ll~ d source signal (or a given discrete rlG~ .J~,y of a source signal ~lo~idillg a plurality of fi~ s not nC~SSalily h=-,...~ir~lly related), 2 0 llli-,loplU~ss-)l 74 is caused to control the Owil~hing within units 76T, 76R to pass 210 MHz fi~luell~ Cnl~-p~ ;7 e.g., to select the 195 MHz-295 MHz b~ p~C Amplifiers 78T, 78R pl.,f.,lably have s~lffi~i~-nt gain to co...~ r for ~nPnllPfion caused by filters 76T, 76R, and have a bandwidth of at least 195 MHz to 815 MHz.

Of course, if P~rlifi~s 78T, 78R were ideal and not subject to front-end overload, it would 2 5 be possible to delete the bandpass filter systems 76T, 76R, and rely upon the op~ of mi~ers 80T, 80R, and narrow band IF units 90T, 90R (to be ~ . ;hed), to separate the various h~-mnnir CG~ of the oscillPtor signal and of the return signal.

As shown by Figure 2, the output signals from - nrlifi~nc. 78T, 78R are provided as an input signal to rnixers 80T, 80R. Fl~luell~r synth~ci7~l local oscillPtorc LOl or LO2 3 0 provide ll~o~ second input signals to mi~ers 80T, 80R, via a MMIC switch S2 (or sirnilar device) that switches between the two S~ d nsc~ r.r signals under control of i~,loplucessor 74.

CA 02251919 1998- lO- 15 The synthP~i7Prl LO1 or LO2 signals are then r~,4u.~ll~ mixed against the selective spectral CollllJùll.,.-ls of the ~ . d source signal and sampled return signal that have been switchably selected to pass through filter banks 76T, 76R. The LOI or LO2 output signals are 21.4 MHz above the h~rmlmir rl~u~ l~y of interest. Because of the difficulty associ-5 ated with imrle.~ e a synthPci7Pd local oscill.q~- r whose output r~ u~,.-~ can range from about 231.4 MHZ (e.g., 7x30 MHz + 21.4 MHz) to perhaps 800 Mhz ~e.g., aboutthe twenty-sixth h~Vmnnir 26x30 MHz + 21.4 MHz), the pl~r, .l~d c.llbodil..e.ll employed two local osrill~t~rs, LO1, LO2. If, however a suitable ~ . d oscillator having a two-octave rl~ u~ ;y output could be rl ~, such oscillator would replace LO1, 1 0 LO2 and the necessily for S2.

Stages 90T, 90R are nallu~.l alld; ~ .. ",~ t.~ rlG lu~ ;y circuits that pass a 21.4 MHz center frequency with a bandwidth of about 25 KHz. Of course by suitably offsetting mi~ing r,~ s~ an IF of other than 21.4 MHz could be used. In the p.~f~ d c,.~o-liln~.-l, IF units 90T, 90R are similar to IF units commonly found in cu-lu..~.L,ially 15 available cellular ~ h~ $

The h~rm~nir rl~ uelley h.ru...~lion passing through IF units 90T and 90R are input to phase detector 92. Phase detector 92 Cu~ u~ . ~;lt-~ source and sampled return signals at each h~ m~nir rl~lu~l;y of interest. The dirr~ nce in phase between these signals is then provided by phase detector 92 to nuelo~ùc,~or 74. At the same time, the 2 0 relative voltage levels from the IF units 90T, 90R at node D are also provided (after suit-able analog to digital co..~ io,~, cu..~ not shown) to ll~ oplu.,essor 74.

To recapitulate"-~lul,locessor 74 receives phase hlrulllldlion from detector 92 that is relative to the various h~ 5 of the source signal (or discrete rl~que ,~ of interest if a non-h~mr~ni~ g~ .alor 50 is employed), and that is relative to the various h~ (or 2 5 discrete rl~.lu~ s) of the source signal as altered by the target s~ c and received at the probe pair 6. Similarly, ~lP~ u~lùc~s~or 74 receives :~rlin-(le hlrolllldlion of IF units 90T and 90R relative to the various h~ ni~s (or discrete rl~ S of interest) of the source signal, and that is relative to the various ~ , (or discrete frequPnriPs) of the source signal as altered by the target s-~ e and received at probe pair 6. Further, to 3 0 perrnit co~ ion for probe te..l~e.~tule and/or ~lubc-s~ pressure v~ri ~tions, limiters 70T, 70R provide uplucessor 74 with amplitude of the source frequency, and with ~ de of the source rl~.lu.,n~;y as altered by the target s ~ re= and received at probe pair 6, while detector 72 provides sirnilar phase information for the source fre-quency.

MiLlu~lù.,eaaol 74 operates under program control, g~,..e.diug data for further ~luCf ;.aih~g by a so-called neural network, look-up table"qlgon~hm or other method of signal process-ing, shown symbolically in Figure 2 as element 100. In a manner known to those skilled in the relevant art, a neural network 100 can be "trained" to lcco~,lli~ a spectral signature 5 qcsociqted with a given target chPmirq-l, glucose for example. To ease this recognition, neural network 100 can optimize the manner of signal prù~,c~aillg within unit 14.

For example, the ol)l.dLiol1 of filter banks 76T, 76R can be altered under control of llucluplùccsâûr 74. In a more ge ~f ~ d PmhotlimPnt, the number and bandwidth of indi-vidual kqn~lp~c hlters within units 76T, 76R could be ~ylldllucdlly m~ ified by suitable 10 MMlC-selection, all under ll~ClUplUC~,aaUl control. However, unit 100 may simply be a look-up table, correlating relative qmrl ~ ~~ changes in a return signal with harmonie frequency against presence or COIlC.,ll~ldliOn of a target chemical in the ~e~ ... Further, a suitable neural network 100 might control llnc~ lùcess.)l 74 to optimize the genPrq~ion of discrete rl~ f 5, based upon plu-,essaed signature data. For e~ample, if a certain set of r,~q~, . f~ ~ from oscillator 50 provided a slight spectral signature, network 100 might direct oscillator S0 to provide slightly different r~ :os until the signature was more recogniz-able.

Mi~ilulJluccssûl 74 in turn provides output signals to output ' (s) 20. As has been d~ e~ output hl~licdcl(s) 20 can, in a variety of formats, display hlru~ ';Ol~ enabling a 2 0 user to d~ the presence and c~ - of a desired target chemical (e.g., ~) in a In the plcf.,..ed ~ lJodi~ l, the ~ is in fact a finger of the individual using the disclosed system. Although the system shown in Figures 1 and 2 was ;...pll ..- ~.l~d in breadboard fashion, those skilled in the art will aypl~ that it may in fact be r;,~.. ;( ~t~'~d in a h ~ hPI~l, battery op~" d, portable unit. In such elllbodill,~,.ll, 2 5 output i " ~ (s) 20 would preferably include liquid crystal displays (LCDs) or simple GO/NO GO intlic~Or~, to preserve power and space. Preferably base 10 would be attaehed to the case housing the l~ of the system for ease of portability.

Figures 4A and 4B ~~I,.c..~ multiple averaged in-vitro data obtained with the system of Figures 1 and 2, using as a test ~l~e~ whole blood (e.g., red blood cells) to which glu-3 0 cose or lactose or sucrose or urea or NaCl was added as a test chemical. The test cellswere CC~ J~ed to a calibrated cell that c~ only red blood cells. Figure 4C repre-sents similar data for whole sheep's blood (e.g., no glueose), and for sheep's blood with various conr~ "c of glucose, where the nome"cl~tu-.; "Blood 102'' denotes 102 mg-%
or 102 mg per dL glucose. Typically, a healthy hurnan has perhaps 80-120 mg% glucose, while a diabetic has 200-400 mg-% glucose. The vertical axis in Figure 4C l~)~ G~ the vector amplitude the return signal, talcing into account mae~ e and phase. The horizontal axis 1e~ ;IGIIl j h ~ --i- 5 of a 30 MHz source fi~.lu~ y, the first harmonic being at 210 MHz.

5 To ...;..;,.,;,- probe-related variables, the ~ c in Figures 4A, 4B and 4C were tested using parallel plate ca~ a ;ilivG cells. These cells colul -ised two ~licle.,llic ~ubDll_ ~ having a relative p-"ll~-llivily a~ e that of water (~ 80), with an electrode surface baked onto each substrate. The test subst~~ce was placed in a chamber between the sul,;,l...s;..

The varying degree of signal ~n~rlitll-~P shown in Figures 4A, 4B, and 4B are termed 10 "spectral signatures". What is depicted is the di~ ,.,cc. in amplitude between the cali-brated cell (analogous to the use of the calibration unit 66 in Figure 2) and the test eç;~" ~ (analogous to the use of probes 6 and sl,e-; - 4 in Figure 1). These data indicate that the system of Figures 1 and 2 may be used to discern the presence of a target chemical within a test ~ or sample.

15 A p~ef"~,d aMIirP~ion is the ~letectisn of excess glucose in a user's blood, e.g., within the ~,c~ ;...- ~ Because the present i..~.",LiOl operates non-i--v~i~ly, it suffices for the user to press his or her finger against the probe pair 6, as shown in Figure 1. In response to the high Ll~ cy, high k~ k content signal from 1"..~...;11. l 16, rhPrnir~lc within the sl.cc;...- -- can ~.,co~,ui~dbly cause energy transfer of certain spectral cc....l.o..~ of the 2 0 1. ~ source signal. lt is hypothesized that within the ~l~e~ the target chemical glucose interacts with the lipid bilayer and/or red blood cell .. ~1." .~

Thus, in the presence of fi~ u~ ,y CC~ yOII~ from the signal l~.~ .c. ..;l~ d via probes 6, the grucose seems to bring about non-linear ~ ~ n~ lPtin~ or mi~cing of ~ u~,...;y compo-nents, possibly due to a non-linear diclc.,l.ic r~ involving r- ~ e ~Qsc~iP~Pd 2 5 with glucose. Using the system of Figures 1 and 2, a diabetic user may rapidly obtain glucose c~-"c~ n level h~ n Signal p--)CC66il-g by unit 18 would, ~ccentislly in real time, provide glucose level h~u... alion on display unit 20.

Of course other target chPmir~l~ may also be detected, i~,~h~ g for example fructose, g~ o.se, alcohol. For ~~ nrle, a system accc..di~ to what is ~liccloced herein may be 3 0 used to sense alcohol in a motorist's system, either by a motorist before ~1- ..1.l;..~ to drive, or by a police of ficer attemrting to ~ r~ -r whether an individual is under the ;~.11. - --e of alcohol.

, Because the disclosed system of Figures 1 and 2 appears to be sensitive to buul~daly con~1itiong at a lipid bilayer ~ f, disruptions to such boundary conditions may be detected by a spectral signature. Thus, the presence of glucose in varying amounts at a .llhl~le may be detected.
In a different utility, however, trauma to a ~yPuc;.. that hlltlr~.~,s with such boundary coll~itionc may also be detected, primarily for the purpose of providing medical ll~dl~ ,n~.
For example victims of Cle~ U~;uliOl may received localized injuries, for example on an arm. Unless the injury sites are ~lulllytly treated by the injection of certain mPrlir~tir~n that is potentially rather toxic, the victim will lose the injured limb or die. Use of the invention 10 ~ligclosP~~d herein would permit 11i~nociC of such injury sites, and qll~nti7ing the injury to facilitate prompt and accurate medical ll~d~

Subsequent to the invention ~hPsrrihed with ,~L,~ncc to Figures I~B, applicants came to a~ idle the role that rhq~ging electrolyte ccnce.llld~ions can have upon glucosecoll~.llldion lll.,a~ul~lll~,lll~ in blood ~ . Applicants further discu~_..,d that it is 15 possible to improve analysis for a desired chemical by reducing the effects upon such analy-sis of varying conc.,.lllalions of other ~ .s~ C in the ~

Figures 5A and 5B l~ ly show in-vitro and in-vivo applir~tion.c of hll~lu~_d analysis using a system 200, aCC(jl-li~ to the present i,.~.,liom In Figure 5A, }.l~f~,.ahly two probes 202A, 202B are coupled by short lengths of coa~ial eable 12 to ports A and B of a 20 L~lu~,.lcy gen.,.dlul and analyzer system 250. In general, the ll --~ t~d signal is sent from port A or port B, and a portion of the ll ' signal is reflected by the ~I,c~
back into the t~ - g port. In ll~ mode (e.g., Figure SB), port B returns the fraetion of the signal ll~ t. ;I via port A through the subject's finger.

In the l....l)o~ of Figures SA and 5B, cables 12 ~/l.,fi,ld)ly are 20 cm or less lengths of coaxial cable, and probes 202A, 202B pl~f~,ldlJl~ are Hewlett Packard HP 85075B di-electrie probes. These probes are eoa~ial in col~hu.;lion, having an outer diameter of perhaps 2 cm and a probe length of perhaps 3.8 cm. The probes have a center cr~nrhlclor that is sulluulldcd by a groundpl~ lP sheath at the probe tip. However, other cable couplings and probes could also be used.

3 0 As will be dc~clibcd, system 250 includes a ll --~ unit 260 that can output discrete shlusoiddl ~ ,ful~ that are spaced-apart in r.~lu~ es linearly or lo~;~ilhl.~dlly in user-s o'e ' '- steps. Further, the output rle.~ f c are stepped between user s~1P~ lr lu~ l and ~y~ l rl~ , - ;r~ f~ and fu~ ,y~li~_ly. In the ~ f~ -o~

f~ was about 300 KHz, ~f, was about 3 GHz, with d~ c- ~ Iy 801 linearly-spaced r.t 4~ jec output between f~ and f". Applicants believe, however, that an bf of about 5 GHz would also be useful to the present h~ ioll. In the plcr~lcd embodiment, system 250 was implemPnt~d using a cvlllll..,l.ially available Hewlett Packard HP 8753A network 5 analyzer with an HP 85046A S-pdldlll~t-,. test set. However, other systems ;~ nl;"g similar functions could be used instead.

System 270 further contains a receiver and signal prucessor unit 270 tnat analyzes wa~crulllls iqCcociq~d with signals l~ d by and/or at least partially reflected back to system 270. The ~d~erc,lll;, under analysis are q-csc~ -~ with discrete user-plu~,la.lllllable 10 fre~lpnriec. The analysis can examine real and hllagill~y ~o,.,l.ont~l~ of these ~.d~fc.",~"
inrh1Aing complex (e.g., having real and hllzgindly col,l~,vl~.,ls) reflection coçffiri~nt data.
These various data are signal plvcessed by unit 270 to provide hlrv~ inn inrlll~in complex impefl~nre mq~ni~ (Z), phase shift, and/or p.,.ll~illivily.

Among the electrolytes, NaCI has the most .cignifir~lt i,.n~ ~re on llleas.~l.,..,e~ " in that its normal concc.ll~dlion range in the human body is 135-145 mM (millim~lq.), whereas KCI, by example, is omy abut 4-10 mM. S~ - c s such as urea were c~.--fi.. ~ to not i,.n"... ~ glucose ",e~~ r~....,.ll~" probably because urea has a mo!c lP- size that is one-third that of glucose, and has a physiologically controlled c~ " ranging from 5-40 mg/dl. The range of glucose in a human normally is about 50 mg/dl (or mg%) to 150 2 0 mg/dl, and can reach about 500 mg/dl in a diabetic.

In Figure SA, probe 202 contacts a cl.c.-;.. of interest 204, perhaps about 40 ml, retained within a beaker or recep~q~ 206 whose volume is perhaps 100 ml. Specimen 204 includes a chemical of interest denoted X, as well as one or more other s~bsl~ c, denotedcollectively Y. In a preferred elllbodi~ spcc;...~ ~ 204 is a bodily fluid, for example 2 5 blood, X is glucose (whose presence and/or CO'~f ~ is to be dc~ ;1), and Y may include varying col~c.,l,lldlions of blood electrolytes such as NaCI, Na2HPO4, KCI, and KH2PO4, as well as proteins and lipids.

Although large concci.,l,dlions of proteins and lipids are also found in blood, the human body ",,;,.l~i".c relatively tight control over ~i~uialivlls in such s~hs~qqr~s and thus their 3 0 presence appears not to s..~ iqlly affect ~..casu~ ,..ls accoldillg to the present in-vention.

WO 97/39341 rCT/IB97100719 In an inrlllctri~l application, c~ c~i,... - 204 may be a solution in which X and Y l~l~oc~
different rhPmir~lc, in which the presence and/or conc~ ldLion of X is to be ~icc~."rd, for example to confirm quality control of the production of solution 204.

A second co..l-:.... 210 into which probe 202B is inserted contains a test or control solution 208 that intPntil n~lly lacks at least one chemical found in ~l~c~ 204. Both ~l,e. ;.. ~
preferably are retained at a same le~ eldlLIle by partially hlllJ~ Ding COl.l;~ D 206, 210 in a preferably constant l~ ~p ~ bath 212 ~ ;llPd within a larger beaker or COIIIh;ll...
214.

In Figure 5A, analyzer unit 250 is operated with signals at ports A and B in a reflP~trlre 10 mode, e.g., in which signals ~ d out of each port are at least partially reflected back into the ports by the l~;o~Jc~ c ~pC'f;~ , From the real and hll&~inly COlllIJOIICIllD of the reflected signal data, useful hlrulllldtion as to the presence and co,~ -, I;on of at least one chemical in solution 204 may be d~ t~ I III;--P~l, according to the present invention.

Applicallls have disco.~,l"d that the real and hll&gin~uy cc.lllpon~,..ls of the reflected signals 15 can be affected by the nature and content of the ~l.e~;.... -- sohlti~n~ in the _ "
vicinity of the tips of the probes. What is believed to occur is that fringing fields extend from the center con~ rtnr of the pl~f~ ~ably dielectric probes to the sulluul~ding ground plane. As the properties of the ~l,e~ solutions change, e.g., due to the presence and con~- ..n~ion of one or more ~h~mir~l~ or other ~llhst~ es therein, the fringing field is 2 0 affected. The alterations to the fringing field in turn affect the reflected signals being returned to ports A andlor B of the analyzer unit 250.

The comple~ data gathered and plu~,coo~d by unit 250 is coupled as input to a Cvlll~ut~,. unit 280 for further plOCe~ g. If desired, c~ . unit 280 may include any or all of the output i"~ 22A, 22B, 24, 26, 28 described earlier with respect to Figure 1, as well 25 as any other output hldicdtvl(s) that may be desired.

Computer unit 280 may be a personal cOIll~/~.~, e~Pcvting a software routine p..,..;/l;"g coll~,.Di.)n of the real and ill a~;illa.y data it receives into forms ;,~.'l,..1i,,g the m~itll~e of the effective complex i...~ -re Z pl~o~t~d by the slJe~ phase shift between signals d and at least partially reflected back by the ~ , effective p.,.lllillivily, and~ 0 the like.

ln the preferred embodiment, co,.,~ P~ 280 executed Excel s~lcadsl~e~l software to convert the incoming complex data into more useful form. A m- ~ifiPd Bao procedure was adopted, in which complex i,..l.e~ re (Z) is dtlelllulled from the complex reflection coefficiPn~ (r) at the interface between the flat end of a probe, e.g., 202A, and a ~,c~;hll~ll solution, e.g., 5 204.
Z=zo 1 - r ( 1 ) The char~r~ritr i",l.e~ e Z0 of coaxial line 12 may be c~lrul~tPd from the rPI~tion~hir ln b Z 377 )1R a (2) '\ /E R 2 II

in which 377 l~ s~nls i",l.c~ - e of air, b is the outside diameter of the probe, a is the diameter of the inner lead on the probe, ~R iS the p.,.-,lcability of air, and fR is the permit-tivity of Teflon.
~0 However, ll.c~.ll~d reflection coPffiriPnt from analyzer 250 is not n~C~s~ y an accurate ion of r, due to errors caused by the col.l;.i.", 206, the coaxial line 12, and col-l-P.~"~ at port A, for example. The Bao l"ucedu,~ reduces these errors, using a cali-bration ploccdul~ based on a linear assumption. This assumption and the values collPct~Pd from the calibration ploce.lul~ give rise to a matrix derivation ~= AlPm A
A - P~5 in which A" is a frequency d~PpPn~Pnt complex constant related to a sc~tt~nng matrix.

During the course of e~l~- ;",- ~ , applicants realized that if analyzer 250 were calibrated with port co~ c~,L~ and coaxial cables 12 attached, the analyzer output would be r, whc.~ on use of the Bao matrix ~lucedu~ would be ~ U ~cecc~y.

Thus, while equation (1) is valid, its real and iu.agill~y COlll~oll.,lllS should be s~ua~_d to 2 0 be effectively used by culllpùlel 280 during eY~PCuti~n of a data plucessillg routine, e.g., an Excel sl,,ca~l~L~l program.

Consider then equations (4) and (5), in which p is the complex refleclion coeffi~iPnt output from analyzer 250: ~
z ZO ( 1 PRea1 PImag) ( 4 ) ( 1 PReal ) PImag Z ~ 2 p ~ag l/zReal ZImag ( 6 ) Euler's formula is used as shown in eqll~tiong (6) and (7) to convert c.~ (4) and (S) to the more commonly encuu..~ ,d i",ped~n~e m~gnitllr~r~ and phase q~ntitir~

e z Rea 1 5 Referring back to Figure SA, the various analytes in a blood ~I.e~ - especiatly smatl ion electrolytes (atso ca'.led blood electrolytes), can .~ .. hl~ affect the ~ e~1 ~n. e and phase angle. In an ~Mli.~if.n in which glucose co~ .......... is to be det~ ;. r~ what actuatty may be Illc~ ,d with system 200 is the effect of glucose, e.g., X in Figure 5A, upon ions or water dipoles in the ~pe -;~ solution 204. Applicants have discu~ d at certain cross-10 over freqn~ n~ies output by system 250, the effects of other ..~ s Y in the ~.c~i.-....
204, including electrolytes, can be reduced or nulled-out. For r Y~mrle, at a cross-over rl~4u.~.l~ of about 2.5 GHz, the conc~ iu" effects of NaCI and most probably other blood electrolytes in a blood 'IJ'' ;l~- ~ are nulled-out, without dc~:la.lillg glucose cunc~ ion Ill~,~7uJ~ J7,nls. In an analyticat scheme in which N eqU~inn~ would have to 15 be solved for N U~OWIJS, the ability to null-out electrolyte cullc~ ions ~rf~ ly reduces the number of vOlia~'~ and thus the number of equations that must be solved. The end result is that glucose concellLI~lion can be clc~ ...;nr,d with higher gpecifirity and confi-dence. Further, as de~.~libed later herein, phase shift I~ .Ul.,...~ S (e.g., co~p~-;cûl~
between ~ le~ and reflected signals) over a wide rl~4~ regime provide a surpris-2 0 ingly linear response to electrolyte col-r ~~ ul~ The phase shift data can then be used to c.,~ for NaCl con~ ';o-~ cr~ntrihutir7nc to total ;~ .ed~.~r.e IllC~-ul-,.llC ll~ made at fl~ - ;PC lower than the 2.5 GHz cross-over fl~ ue.l~.

When g, 260 outputs r,~ 5 greater than perhaps 1 GHz or so, the ~I.er;.-~ n i.,.l.ed ~ e mqgnih-~ appears to be primarily a funetion of the ability of water dipoles to respond in the presence of the resultant osrillq'ing field in the vieinity of the probe(s~. At output freql-~nrips less than perhaps 500 MHz, i,.~peil~ ~~e magnit~l(le seems to be more a 5 function of ionie response to the osrillqting field in the vieinity of the probe(s). Within a blood sl,eA ;..,~n, NaCI is an h~ l source of sueh ions. At h~bch.-~n rl~lu~llc;cS, the ;---l~c1~ e function llal~ iOI~.

Below a~ lllal~ly 500 MHz, glucose in the ~l~e-; - - solution appears to impede ionic mobility in ~nJiilg to the osrill~ting field, and thus the effeetive iL r-~ ~.e in,lti&se~s.
For example, between about 10 MHz and 100 MHz, imreA~lre change due to NaCI
con~ inn changes in the s~e~;~ - - are s~lbstq~ti-lly stronger than i...l.cA~ -~e changes due to eol~cG.ll,àlion ehanges in glucose.

Applieants have disco~ ,d that at test r~ s below about 1 GHz, inereasing COIlC~ d~iOlls of NaCI deerease i.. pe~ .~e ~ de (nZ"), and that at a eross-over fre-15 queney of about 2.5 GHz, i",l.~A~-..e ."~u~.,."e."S are sensitive to glueose col-~ irn but i"s~ to eleetrolyte col ~f ~ ion Further, applieants have learned that over a wide r~ u~ r regime, phase shift h~cl-ascs linearly with hl~l~illg NaCl conccnlldion, with little or no effect due to ch~ ~ging glucose and/or albumin co~e .~ ion Thus, it appears that at higher r~.lv~ (e.g., above 1.5 GHz or so), larger lllGI~ 1PS simply do 20 not respond sl~ffir~ ly rapidly to m~ ingfillly i..n"~,~rc phase shift ~C~u~ ls. By eontrast, eleetrolytes, ;~ lv~ .g NaCl, have small ions that ean respond luca~ulably with respeet to phase shift ",ca~u,~.".,nls. As desllil,cd herein, eolleetively, these di~ ,.i.,s provide ulca~ul~ -ll protoeols to reliably and with ~e~,irl.;ily d~ -f glueose eoneen-tration, despite the presenee of eleetrolytes of varying collc~ ldlions.

In Figure SB, a non-invasive system for in-vivo testing is depieted. In this ~.. ho~li.. l network analyzer or system 250, and co".~,ul~,. system 280 may be identical to what was ;I,ed with respeet to Figure 5A. However, an cle~ .Jdc assembly 310 COlll~ i~ two metallic probes 320 spaeed-apart perhaps 2.5 em on a substrate 300 is used. S~lbct-~q~ 300 may be a sheet of single-sided eopper elad printed eireuit board ~r~ g perhaps 5 cm x 3 0 7.5 em. Eleetrodes 320 pr~f~.. dllly are made from brass and are about 0.6 em tall, 0.6 em wide, and about 1.2 em in length. Spaeed-apart faees of the probes define a surfaee slanted at about 45~. Eaeh COIl~lu~,ti~., eleetrode 320 is c~"~nP- ~d to one eoaxial eable 12. The finger 4 of a subjeet to be tested for glueose eoll~.lllàlion, for s . '~, is pressed against the slanted surfaees of the probes, thus c~.".l,lc.;ng an electric~' eireuit with eoaxial eables CA 022~1919 1998- lo- l~

12, and thus ports A and B of analyzer system 250. It is understood in Figure SB, that port A will receive back a portion of the l.~ ''...;llrd signal as a reflected signal. Port B
will receive that portion of the tr~n~mit~d signal that propagates through the .sl,cc;.,...
tissue.

5 In practice, probe assembly 310 provides ~ ed signal to noise ratio, and hl~lu~,d ~ Jcd~àbility relative to other probe designs, inrluAing the probe assembly depicted in Figure 1. Reliable data have been obtained with probe assembly 310, typically in the r ~lu~,nl,y range of about 1 MHz to about 3 GHz. It will be ~l~;ialed that the config-uration of Figure 5B is especially useful to laypersons, inrlnAing s..~ d and actual diabetics, who wish to monitor their own blood ch.,.~ ly, especially glucose Con~ n levels.

Figures 6A and 6B plot predicted and actual glucose cullc~lllalion against time, for non-hlvaDively obtained test data (shown by "plus signs") and for hlvaDi~,ly obtained data (shown by "boxes"). Both figures depict the same e,.l,~.i.l,.,nl in which a human subject drank water at 14:00hours (2:00 P.M.) and ate food at 15:15 hours (3:15 P.M.) The non-invasive test data were obtained using finger probes 320 such as shown in Figure SB, whereas invasive test data were obtained from actual blood samples from the subject.

Applu~i,,ldltly 101 separate ~lc4u ~ s were used to obtain raw data during the experi-ment. Figure 6A depicts non-invasive ~I~Jict~,d glucose c-.n.~..l., ~i.ll~ based upon imped-2 0 ance and phase data taken at about 17 MHz. The i~ A~ e and phase data were then Cull~,.t~d into pl~ t.,d glucose CO.~f l., ~;on data using an algorithm.

In Figure 6A, pl~ di.,~ed glucose col-~e ~ . shows an increase at about 14:20 hours, a~pal~ ly coll~ unding to the subject's intake of water. In essence, the water has diluted electrolyte cun~ dlion in the subject, which has caused predicted glucose conccnl~alion to 25 offset vertically, e.loneouDly, by some S0 units. After lS:lS hours, the pl.,li-,t.,d glucose level rises, which l~ lD the subject's intake of food. Note, however, that the same S0 unit vertical offset is still present.

Using m ~hPm~ir~1 regression analysis to examine data for the ~JI.)x;... ~ 'y 101 r~ P~ used, applicants realized that non-invasive phase shift data taken at 103 MHz 3 0 would provide a cull~lioll for the S0 unit error offset in non-invasive glucose ~l~A;~
taken at 17 MHz.

Figure 6B shows the same ~ I. now plotted with correction data taken at 103 MHz, in which "plus signs" depict predicted non-invasive glucose cu..c~,.l.dlion data from the subject using 17 MHz tr~ncmicci~ln-mode i,.,pP~ c ..~=g~ data as CCu-u~ ,.i by the 103 MHz phase shift data. Clearly the use of the higher r~ u~,...;y phase shift co..~lion has 5 largely co~ nc~Pd for the 50 unit offset (present in Figure 6A but not in Figure 6B), re-sulting from water dilution of electrolytes.

In general, Figure 6B shows close agre.,l..."-l between actual h.v~iv~ly ll.~,a~ul~d glucose conc.,.lllalion, and non-invasively predicted glucose co~-~e ~ on. Although not fully appreciated by applicants at the time the subject t~ was co-~h~ 1 it appears that 10 the 103 MHz phase shift data provides a good measure of cle~ .lyle COllC.~.IIldliOll including the effects of electrolyte dilution. At 103 MHz, small ion electrolytes i"rh~ g NaCI could respond to the o5~ ting field, whereas larger glucose -'~ IPS could not, and thus would not c~lbst~n~ ly inflmPnr~p the nlC~t~ . By contrast, the 17 MHz data provided a measure of glucose and electrolyte conc~,.-l-alion, which data when co...~
15 for by the 103 MHz electrolyte conc~nl~dlion data provided a truer measure of predicted glucose conc.,..lldlion.

Collectively, Figures 6A and 6B suggest the wisdom of using data obtained at different L~lv- ~ ~ ies or fl~ .lu~.~cy regimes (e.g., 17 MHz and 103 MHz in this e~ample), to measure different palalll~,t~ (e.g., total i~ .c~ e, and phase shift), to provide a measure of 2 0 co",l"."c ~ jon to more ac~u~ahly arrive at the desired data (e.g., glucose COl C~ ali with a greater ",~. ;1;~ ;,y confi~lPnrP level.

Figure 7A depicts the startlingly linear rPl~tionc1lir observed by all~licallts between NaCI
collc~i.llldlion and phase shift between tl - .~...;1l. d and reflected signal at a ;",~ . In Figure 7A, various fl~ u~,nc;es between 2.25 GHz and 2.75 GHz were used, phase 2 5 .lirf~ ce was between two probes, e.g., probes 202A/B in Figure SA. The ~1.... ;.,....l began with distilled water, which at shown at the bottom of the graph had 0 radian phase shift. Adding i..c.~....,-.l~ of 20 mM NaCI to the distilled water showed a very linear rela-tionship: higher NaCI co~ ;. - hlw~_z6cd the lll~,a~u~d phase shift rather linearly. At the very top of the graph, data were obtained first for 300 nM NaCI, after which two 100 3 0 mg/dI, of glucose powder was added to the salt water solution. As seen, in the 2.25 GHz to 2.75 GHz fl~ u~ ,y regime displayed, ch ~ ging glucose conc.,..lldlion (indeed a rather s~ctr~ change in glucose concc..llalion) did not affect phase shift lllC5~
whereas fh~ ging NaCl conce.llldtioll pluduced a linear change in l"ea~u,able phase shift.

Figure 7B is averaged phase shih data obtained with two probes, using r~ s ranging from 2.0 GHz to 2.5 GHz, in which varying concc.ll-aLions of NaCI, glucose, and albumin were added to a baseline solution of phosphate buffered saline ("PBS"). PBS was used in that it rnimics the electrolyte ~llvhu~ t of blood well, without proteins or other 5 ~ es being present in the solution.

Concict~nt with the findings of Figures 6A and 6B, increasing NaCl col.c~,,ll.alion inc.~ascd phase shih in a linear fashion in Figure 7B. Of special ~;g ir, - ~re, however, is the bot-tommost portion of the graph, which cu..~llds to a phase shih of about 0.11 radians for a 246 mg NaCI solution. This data line ~ aillf d constant, even when 40 mg (100 mg/dl) and then 80 mg (200 mg/dl) glucose were added, and even when 100 mg (250 mg/dl) albumin was further added. The data of Figure 7B d~,.llon~ll..t~,s that the linear phase shih lable for varying electrolyte conc~..L,aLion is not - gfillly ;--n~ ~ed by glucose collc~ LlaLion and/or albumin c~l.. ~.. -~i Figure 7C is a co~ c,Oile graph that d~ that a cross-over L ~ ue-l~ of about 1.5 15 GHz renders phase shih lllca~ul~ highly h~cll~ ,., to varying albumin col~-f..l,.~ir,nc in a PBS solution. In Figure 7C, the bollulll-l-oOL trace at about 0.017 radians l~ ,O~ O
phase shih caused by ch~ lging concf nl-alion of gamma globulin by S g/dl, and the trace at 0.005 radians ~C~ ,o~ltS phase shih caused by rh~ging cvnc.,.llldlioll of gamma globulin by 2.5 g/dl. The u~ oL trace in Figure 7C l~ ,o~ phase shih due to intralipids at 2 0 1.4 g/dl con~.-l-alion, the trace at -0.02 radians .~.eO~,.. lO a different analyte with glucose, not herein relevant, and the -0.015 phase shift l~pl~;OelllO h~ J;ds at 0.7 g/dl conrrntr~
tion. Of special interest are the three llacclilles centered about 0 radian phase shih. The trace at -0.005 radians .e~-.,s~..L~ albumin at 5 g/dl c~n~ ion~ the trace at about -0.0025 radians I~JI~,o~illto albumin co..~.,l.Lldlion of about 2.5 g/dl, and the trace at 0 phase 2 5 shih is the baseline PBS. The various conc~ l.l.alions above noted are s ~l.,l ~li I jy greater in mag~it~lAr than variations that would ever occur in a hurnan being. Note that at a fre-quency of about 1.5 GHz, phase shih is _b. :~lly hl~ lloiLi~_ to albumin concenllalioll level. Thus, by ~ ;I.g different ~ r 1~ 1;l 5 s~cco ' 1 with a ~l~cc;~ n at different L~ ucncies or over different L~lu~ ,y regimes, the effects of various co~ i can be 3 0 nulled-out. In the example of Figure 7C, greater ..leasu.~,ll.~,.ll cperifirity is attained for a desired analyte, e.g., glucose, in the presence of other oub~ res e.g., albumin.
Figure 7D depicts phase shih data between about 300 KHz and 100 MHz for cl~-.8;1-~con~.,llL~dtions of glucose, the glucose being added to sheep blood in h.~,..,....,l.l~ of 250 mg%. It is seen that at about 20.1 MHz, phase shift data is insensilive to glucose COllC~

Figure 8A depicts magnitude i~ eA --~ e data measured in a sheep blood baseline solution, for various glucose con~,..lralions, using frequencies ranging from 0.3 MHz to 1.0 GHz.
5 Over this extremely wide L.,.lu.,.l~ regime, hll IG~hlg collce~ dtions of glucose increase impeA~nre. The relative change of glucose conc~.llldion upon impeA~-Ite is greater at rl~ " 5 lower than about 0.5 GHz, no doubt because at lower frequencies the large glucose m-~lec~ s exert greater hind~,.~,ce upon ion IlW~re~

In general, applicants have learned to appreciate that i..~l.eA --.~e ,..ea~u~G~ l accuracy is 10 higher at low L..~ 5 than at higher L~lu~ c Thus, as will be seen, i",peA_.,. e ...e~u.."..~,..ls at 2.5 GHz can provide a measure of glucose collcG,lllalion nulling-out NaCI
and other electrolyte conc~ alions, the e~lui~ lll Ill,a;~ se~-;,ilivily is suhst~mi~l less than at say 100 MHz. For example, a ~...,a~u~".~".l sG..silivily of 0.1 Q is a good design goal. However, at 2.5 GHz, i~ A ~ e ...~g~ SG..~ilivily will be about 1/25th the s.,n~ilivily at 100 MHz. Thus, as AP~rrjhed herein, a .~co.. A~d protocol will involve i,..l.eA--.re and/or phase Ill~ lls in the GHz range, as well as ",ea~u.
at much lower LGq~,.... :~s.

Figures 8B depicts ;,-.~.o1~- ~e change when a ~ of sheep's blood has glucose added, but relatively little change when co"ce,ll.dtions of NaCI are added. The bott- mmr.st plot 2 0 (with "bo~es") is baseline sheep blood with a declotting agent. One addition of NaCI was then added (e~luiv ' {o change in con~.llldtiu.. of 10 mM), and data taken at five minute intervals for the ne~t five runs. During the last (U~ IUOSI) four runs, glucose was added.
Glucose ~rld;~ C clearly increase the ~,.ea~u,~d . - ' -e. Note that, contrary to behavior at lower L~lu- -- :~sl adding NaCI in the 2.94 to 3 GHz regime actually hl.,l.,dsed im-25 pedance, probably due to an a~,liull of ions with water -'e l~es.

In Figure 8C, ;.. peA--~e data were obtaining using r.~.,.,.. : i ranging from about 2.42 GHz to about 2.48 GHz. Again, a baseline solution of sheep blood (drawn with "bo~ces") was used, into which one addition of NaCI was made, followed by four a~Mition~ of glucose. For the NaCI ~AAition~, essenti-'ly no ~ .ed~ e change results in this r~ u~.luy 3 0 regime. However, the u~,."~OSl four runs, which rG~ G.ll addition of increasing c~ n~ of glucose, clearly increase i,..~ e in this r~ u~ y regime. Thus, l ed - - e ",e~u,~"".,.~s in a rl~lu~,.l~ regime of about 2.42 GHz to about 2.48 GHz are sensitive to glucose concG.,lldlion, and are i,.~Gn~ , to NaCI and other small ion elec-trolyte concc.,l~alions. While sheep blood was used as the s~;c;..,~ n, similar results are obtainable with human blood. Further, as noted earlier, the human body ...-;"~ c tight ho..l-Pos~ control over co,lcen~ld~ions of most electrolytes, proteins and lipids within the blood.

5 Figure 8D depicts i",peA~... f data for a frequency regime of about 2 GHz to about 2.1 GHz for a baseline of sheep blood (drawn with "boxes"). In the bulio ~Dl runs, the addition of NaCI (10 mM cunc~ l.dlions hl~,lc.ll~,.lls) caused a decrease in imperl~nre However, in the ~ DI four runs, ~ itil n~ of glucose clearly h~ ccD~,d ;"~l~e~, e in a linear fashion.

10 Figure 8E depicts i"~pf ~ re q, d-P, ,u.,aDul~ ,..LD made using r,~,q~ ... iP5 ranging from about 2.25 GHz to 2.75 GHz, with a D~e'';~ll' 'I of distilled water into which increasing concentrations of NaCl were added. The buLL~.lll~llGDL traces l.,~,~,s~,..l distilled water baseline data, and the lu ~ :"i"g traces reflect hlclcaDillg cl-nc. .""..i,.,.c of NaCl, with the U~ IIIOD~ trace .~ li"g highest con~.,l.alion (200 mM NaCl). I~lt._.~,;,Lil.~ly, the 15 effect of in~,lcaDill~ NaCl conccl.l.diùl. upon hllr-' -e varies non-linearly with Lf ~u~ y.
The right portion of Figure 8E ~P IIO~ at~ that imped~ re increases with h~ caDillg NaCI
col1c~ ,aLion (a result opposite to what is e ncuul.t~,~,d below about 1 GHz). By contrast, the left portion of Figure 8E shows first an increase and then a decrease in ;llq)~~ as NaCl co~ ion i~..;.casPs (e.g., as more Na or Cl ions are added to the test solution).

2 0 Figure 8F d~ t~,s that use of a Lc~lu~ ,y of about 2.5 GHz can null-out ~$QPnti~1y all changes in NaCI ~ ion upon i...peA~ ~e ll~a~ulc~ llS. The data shown in Figure 8F were gathered using a distilled water spc~ ;. P-~ into which hlcle~ g COI Cf .l~ iolls of NaCl were added. At the al)~lv~ 1y 2.5 GHz cross-over Lcqu~ ,y, a11 curves inter-sected, inflf p".~ 1y of NaCl col1c~.,l.dliun. Note that the NaCl c l~n~ used inFigure 8F included the human physiological range of about 135 mM to 145 mM NaCl.
Figure 8G depicts average i...l.ef~ as a function of L~ ~lu~ ~ ranging from about 1 MHz to about 0.4 GHz. Note that between about 0.1 and 0.2 GHz, gamma globulin appears to saturate.

In other c~l...;....1~;, applicants lll.~aD~ ,d i...pf d~ .~e magnit---1f~ using PBS at various 1~ CD to dF~.. ;.. F tC~ dtUlC s~,nD;livily. These ~,l.~ ;.. l~ f~icc10sed that use of frequencies ranging from about 800 MHz to about 900 MHz provided ;1--l~ 1f.~" I..~g.~ d~F
data that was t~ Ulc in~ . The ~ aDu.~ were made using reflective mode, but the same result would apply to l~ d mode data. When using a non-invasive in-vivo IllCa.5Lllc~ configuration such as shown in Figure 5B, skin t~ t~c at a subject's fimgers can range from about 24~C to about 37~C. In practice, it is l~CO...,.,...A-Fd that in addition to other data, that data also be taken in the 800 MHz to 900 MHz 5 t~ p~.alulc hl~e,~iLi~, regime, to provide a measure of correction as needed for the other data.

To recapitulate, the present invention r~CGg~ S that electrolyte ion inl~lr~.~,.~c, especially NaCI, with glucose ll..,~ul.~ "ls can be reduced. In one application, the intc.r~ ,..cc is crrt~,lively nulled out, using impe-l~nre q~ de ll.ca~urc..~.ll~ at a cross-over L~ u~,u~iy.
10 In another app1ir~tion, c~ on for electrolyte ion effects upon glucose lll~,~ul~ ...
are made. The configuration of Figure 5A and likely that of Figure 5B can predict total glucose cn~ with acc~l~blc specifir~ity and error tn] ~ e.

As noted, it is adv-- l ~geouC to make high frequency and low Lc~lu~...,y ll.ca~u.~,....,.-L~ of various p~ t~ ~ to provide a good glucose col~ or~ prediction (with good specifici-15 ty) in a s~ Preferably, low L.,~lu~,,lcy regime data is taken over 21 or moreL~UF ~ es, and high rlc.lu.,ucy regime data is taken over 81 or more L.4~ ;es While the plef~,.lcd e .~ho~ used a network analyzer that provided discrete L.,~ c, one-at-a-time, the various Lc~l~ .r ~ c could instead have been p-~ ' en masse, or as groups of freq~l-FnriFs, rather than as discrete separate L~ u- ~ FS

2 0 High Lc~u~,u~, e.g., 1 GHz to perhaps 5 GHz, ~ clu~,.lL~ of phase provide a good measure of electrolyte conc~ Ll~lion, in which Lc~u~,~luy regime the phase u.e~u.e..l.,~
are ill~nsi~ , to glucose con,~ .LldLion. On the other hand, use of a 2.5 GHz cross-over L~quc,l.;y permits i...l.~ e q, ~ ~e inAir~tinn of glucose co~ r,~ ns, with little c--~ oi, from electrolyte col-~- ..1" ~innC, But the most sensitive lll. ~ul.,~ of glucose 25 cn~ n are obtained at lower L-~ -s, at which ;"~ A~ "I~,~g";~ F is a measure of glucose conct..Lldioll plus electrolyte con~

High L~u.,.l~y phase response was used to predict changcs in NaCI co~r~ inn Thispl~ Lcd NaCl cu~ inl~ change was then used to predict the i..~peA~ ~'F~ ..;I,.A,~
change at low L~yu~,~l~ due to electrolyte cnl~....I.a~inn change. The ~lcdicLcd low 3 0 L.,~lu~ ,y electrolyte cnntrihlltinn was then ~ubLI~ t~d from low L~lu.,,~;y tot~ pF;A - -=e magnihlAF The ~ A., was i~peA~ change due to glucose c~ n In mathe-matical terms:

~[NaCl] = CAL_CURVE_NaCl_PHASE_HI * APhase ~ high r~ u~ y ~ZNaCl =
CAL_CURVE_NaCl_MAG_LO * ~[NaCI]
~Zglucos~ CI
~[Glucose] = CAL_CURVE_GLU_MAG_LO * ~Zglu,~e 5 The "CAL_CURVE" eApression is derived from calibration eql~otif~nc When ~~~~nl~ing c.,nccl-l-dLion changes from phase or i~ c~ e change, it is necessary to solve an appropriate calibration equation for an unknown, e.g., IlA" in terms of a know, e.g., "y".
NaCl calibration was made using a PBS baseline solution into which NaCl was added in 2 mM in~l.,lll.,llls, up to 12 mM above normal PBS. A second NaCI calibration involved di-10 luting a PBS solution with distilled water in 2 mM h.~ ""s, to -12 mM from normal PBS, during which time the solution volume changed from about 588 ~1 to about 685 ~1.
However, the resultant calibration curve provided a linear response with an eAcellent fit, e.g., R2 >0.999. Glucose calibration involved three separate cAy.,lhll.,l~ls using -10 mM
PBS, normal PBS, and + 10 mM PBS baseline solutions, into which glucose was added in 100 mgldL in~ ,lu~ s to 500 mg/dL. The glucose response was quite linear with good correlation for the calibration curve.

In making r~y. ~ runs, error was defined as 100*(Predicted value - Actual value) I
Actual value. On a run-to-run basis, NaCI COI~C~"~,dliO.l pre~irtif~nC were <3% and overall NaCI concG"I,dion predictions have <0.2% error. Overall, glucose cu-,~,-l-dion 20 prefiiçtionc had <13% error, and run-to-run glucose predicfif)nc had <23% error. These results are gratifying, although future cl.,bo~ will no doubt return even more accurate glucose yr~fl;l ~iu,-c.

The prediction method has the advantage of being fairly sensitive to NaCl, whose low r~ u~,ll~ response is stronger than that of glucose. Although NaCl changes may be 2 5 prcdict~d with accuracy using high r ~4~ .~ phase data, any error in such Ill.,a~
tends to be "mog~ifiçd" by the leveraging effect of NaCl at low r.~.lu.,l.c,cs relative to glucose. Ideally, co,l",~ dion would occur at some r,~4u~,l,~ whereat the NaCI response and glucose response were closer in g it-~de. Applicants are also e~ .i,.g use of m~th~ tiro1 d~livdi~ . of the i.,.l.ed~-,rc and phase data obtained with the present inven-3 0 tion.Moflifirotions and variations may be made to the disclosed cl.~odi~ ,ls without departing from the subject and spirit of the invention as defined by the following claims.

Claims (22)

WHAT IS CLAIMED IS:

1. A method, capable of in vivo operation, for determining concentration of a first chemical in the presence of a second substance in a specimen, the method including the following steps:
(a) subjecting said specimen to radio frequency signals having a frequency regime ranging from about 0.1 MHz to about 5 GHz;
(b) at a first frequency regime, using at least some of said radio frequency signals to obtain data proportional to magnitude of concentration of said second substance in said specimen;
(c) at a second frequency regime, using at least some of said radio frequency signals to obtain data proportional to combined concentration in said specimen of said first chemical and said second substance; and (d) using data from said first frequency regime and data from said second frequency regime to obtain a measure of concentration of said first chemical in said specimen.

2. The method of claim 1, wherein said specimen includes blood.

3. The method of claim 1, wherein said specimen includes blood, said first chemical includes glucose, and said second substance includes NaCI.

4. The method of claim 1, wherein at step (a), at least some of said frequencies are presented sequentially.

5. The method of claim 1, wherein at step (a), at least some of said frequencies are presented simultaneously.

6. The method of claim 1, wherein at step (b), said first frequency regime ranges from about 1 GHz to about 3 GHz.

7. The method of claim 1, wherein at step (b), said data proportional to magnitude is obtained by measuring phase shift between radio frequency signals input to said specimen and radio frequency signals returned from said specimen.

8. The method of claim 1, wherein at step (c), said second frequency regime ranges from about 0.11 MHz to about 3 GHz.

9. The method of claim 1, wherein at step (c), said second frequency regime ranges from about 800 MHz to about 900 MHz, in which regime temperature effects upon data are minimized.

10. The method of claim 1, wherein at step (c), said data proportional to combined concentration is obtained by measuring magnitude of independance at said specimen.

11. The method of claim 1, wherein at step (d) a concentration value determined in step (b) is subtracted from a combined concentration determined in step (c) to provide said measure of concentration of said first chemical.

12. The method of claim 1, in which said method is carried out non-invasively on a human subject, and wherein step (a) includes coupling said radio frequency signals via at least one probe that contacts a distal portion of said subject's body.

13. A method, capable of in vivo operation, for determining concentration of a first chemical in the presence of a second substance in a specimen, the method including the following steps:
(a) subjecting said specimen to radio frequency signals at a cross-over frequency at which frequency concentration effects of said second substance are essentially nulled-out; and (b) determining from data taken at said cross-over frequency concentration of said first chemical.

14. The method of claim 13, wherein said specimen includes blood, said first chemical includes glucose, and said second substance includes NaCl.

15. The method of claim 14, wherein said cross-over frequency is about 2.5 GHz.

16. The method of claim 13, wherein at step (b), said data is impedance data.

17. The method of claim 13, wherein step (a) is carried out non-invasively on a human subject by coupling said cross-over frequency via at least one probe that contacts a distal portion of said subject's body.

18. A system, capable of in vivo operation, for determining concentration of a first chemical in the presence of a second substance in a specimen, including:
a transmitter outputting radio frequency signals having a frequency regime ranging from about 0.1 MHz to about 5 GHz;
at least one probe, coupling to said transmitter, contacting a portion of said specimen; and a receiver-signal processor system, coupled to said at least one probe, that analyzes at least some of said radio frequency signals present at said probe;
said receiver-signal processor system providing data including at least impedance and/or phase shift present at an interface between said specimen and said at least one probe;wherein data provided by said receiver-signal processor system is used to determine said concentration of said first chemical in said specimen.

19. The system of claim 18, wherein said transmitter transmits signals appropriate for a specimen that is human blood, for a said first chemical that is glucose, and for a said second chemical that includes NaCl, and wherein said transmitter includes a network analyzer.

20. The system of claim 18, wherein:
said specimen is a human subject including said subject's blood;
said first chemical is glucose;
said second chemical includes NaCl;
said transmitter transmits signals appropriate for discerning concentration of glucose present in NaCl within human blood; and said at least one probe contacts an exterior portion of a finger of said subject such that non-invasive data is provided by said system.

21. A non-invasive method for determining concentration of at least a first constituent in the presence of a second constituent in a specimen, the method including the following steps:
(a) subjecting said specimen to radio frequency signals having a frequency regime ranging from about 0.1 MHz to about 5 GHz;
(b) at a first frequency regime, using at least some of said radio frequency signals to obtain data proportional to magnitude of concentration of said second constituent in said specimen;

(c) at a second frequency regime, using at least some of said radio frequency signals to obtain data proportional to combined concentration in said specimen of said first constituent and said second constituent; and (d) using data from said first frequency regime and data from said second frequency regime to obtain a measure of concentration of said first constituent in said specimen.

22. The method of claim 21, wherein said sample includes in vivo blood, wherein said first constituent is hematocrit in said in vivo blood, and wherein said second constituent includes at least one chemical whose concentration in said in vivo blood can affect a measured concentration of said first constituent in said specimen.