1 Introduction

There are various approaches to brainresearch. Among these, measurement ofhuman brain activity and analysis of resultantimages has developed rapidly in the lastdecade. Brain-activity measurement instru-ments used in these cases are also used inmedical practice, and were not originallydesigned by researchers. Researchers mustthus ensure that these systems have passed rig-orous manufacturer examinations if these sys-tems are to be used in research.

Determining methods of measuring thebrain and the types of conclusions that are tobe drawn from measured values represent sig-nificant tasks. Before drawing such conclu-sions, it may first be necessary to examineseriously whether the obtained data can sup-port detailed analysis. Our group establishedas one of its research tasks the development ofspecific tools for brain research, as detailedanalysis of brain research cannot be carriedout if we simply accept the current tools, andconsequently we must devise tools specificallytailored to the tasks at hand.

In the main part of this paper, we will firstpresent an overview of the objects of non-invasive brain-activity measurement with theinstruments currently in use. We will thenprovide a description of the principles andapplications of brain-activity measurementusing light.

2 Visualization of tissue

Physical phenomena involving ultrasonicphenomena, electromagnetic waves, X-rays,light, and so on all involve waves. Throughthe use of these waves, a living body can bemeasured and visualized. Technology to visu-alize the interior of the living body has longbeen sought after in the field of medicine, withX-ray tomography (X-ray CT) and magneticresonance imaging (MRI) now indispensablein medical practice. Measuring tissue withoutincision is referred to as “non-invasive” inEnglish, and the JIS has stipulated use of theJapanese term “mu-shinshu” (non-invasive) todenote this medical practice. (Currently, theterm “hi-shinshu (also “non-invasive”) is oftenused in Japanese.)

EDA Hideo 73

3-7 Measurement of Brain Activity by NearInfrared Light

EDA Hideo

Non invasive brain measurement systems are widely used for the brain research. Thereare two kind of equipment. Ones measure electro-magnetic signal from the brain by EEG orMEG. Others measure hemodynamic response by fMRI or NIRS. Development of analysisand research tool is also important for the brain research. NIRS calculates changes inhemoglobin parameters. We compared images by fMRI with images by NIRS. Theseagreed quite well.

Keywords Brain activity, Non-invasive measurement, Hemodynamic response, Near infrared light,Spectroscopy,

74

X-ray CT began with a paper byHounsfield published in the British Journal ofRadiology in 1973. Subsequently, an X-rayCT image of the head was presented byAmbrose et al. For MRI, Lauterbur publisheda paper in Nature in 1973, the same year as theinitial report on X-ray CT. In 1978, the com-pany EMI presented a number of MRI imagesof the brain. Both X-ray CT and MRI havebrought Nobel prizes to the researchersinvolved.

It is very interesting that in each case ofthese initial reports the brain was the measure-ment target. In terms of obtaining a structuralimage, the brain is not a particularly difficultorgan to image relative to other internalorgans. Organs involving intense movement,such as the heart, are far more difficult to view(in fact the technological challenges of heartimaging have formed a core target of signifi-cant past research). In terms of radiology, theorigin of human body imaging even todayremains the heart. Accordingly, images of thebrain are from below (roughly from the posi-tion of the heart), and, consequently, theresultant brain image must be mirror-reversedto correspond to the actual position of thebrain relative to the exterior of the subject. Ifthe brain is displayed as viewed from above,not from below, the right/left orientation of theimage coincides with that of the actual struc-ture, rendering the brain image easier tounderstand. However, we must note thatimages taken from above are different fromconventional images, with the heart located inthe origin.

Any device that measures human charac-teristics through interior imaging irradiateswaves to the living body and produces imageinformation based on the degree to whichthese waves are absorbed based on signalreflections from the body. The waves may beclassified as sonic waves or electromagneticwaves, depending on wavelength. One char-acteristic of the waves used in somatometry isminimal absorption by water. Since it cansafely be said that most of the living bodyconsists of water, this characteristic can be

phrased conversely: waves at wavelengths atwhich absorption by water is minimal are usedin somatometry because such waves can easilypenetrate the living body. On the other hand,devices exist to measure the waves emittedspontaneously by the living body, includingequipment used in electroencephalography(EEG) and magnetoencephalography (MEG).These devices operate at wavelengths resistantto absorption by water. Additionally, positronemission tomography (PET) provides impor-tant metabolic data through the injection of aradioisotope and subsequent detection of thegamma rays issued from the interior of thebody.

The diagnostic imaging equipmentdescribed above can be divided roughly intodevices that display structural images of theliving body and those that display functionalimages. Images used in brain research areappropriately referred to as functional images,which differ significantly from structuralimages in a number of respects. The presenta-tion of functional MRI (fMRI) by Ogawa in1992 [1] is now viewed as a monumentalachievement in the imaging/measurement ofbrain activity. Figure 1 shows an explanatorydiagram of Blood oxygenation level depend-ent (BOLD) signal. Ogawa et al. gave a pure-ly physical explanation (the right-hand part ofFig. 1), while the other explanation (the left-hand part of the Fig. 1) was added separately.

This fMRI technology plays an enormousrole in current brain research. However, itmust be notes that an fMRI image is usually

Journal of the National Institute of Information and Communications Technology Vol.51 Nos.3/4 2004

Explanation of Blood OxygenationLevel Dependent signal

Fig.1

comprised of statistical values. In otherwords, it does not show a general state ofcerebral activity, but instead indicates the spe-cific response of a given area to a stimulus. Athreshold response is set to enable statisticalcalculation of the correspondence betweenstimulus and the area in question. The selec-tion of this threshold value is very important:if the threshold is too low, the image may indi-cate that every site is active. The type of stim-ulus also affects the results, as with the selec-tion of the threshold. Therefore, in brainimaging, it is extremely important to generatea precise stimulus that activates only the targetarea.

3 What should be measured in brainresearch? At what resolution?

If brain activity is considered from amicroscopic viewpoint, a neuron fires, a meta-bolic motion is initiated, and there is a hemo-dynamic response change. Figure 2 shows adiagram illustrating the principles of fMRI.Firing of the neurons results in activitybetween various areas in the brain, finallyresulting in consciousness and qualia—a rangeof higher brain functions. When studyingthese higher brain functions, it may be saidthat the issues involved may only be addressedafter images of all neurons and their hemody-namic responses are obtained with high tem-poral resolution. However, since measure-ment of the head and analysis of the resultantimages remains a topic of discussion, brainresearch must proceed based on incompleteimages with bold conjecture. Further, even ifthe activity of all neurons could be measured,we would quickly lose our way among theinnumerable resultant graphs, estimated tonumber roughly twelve billion.

The primary object of brain measurementis considered to be the state of neural activity.Simply speaking, measurement of such neuralactivity would have to be on the order of mil-limeters in terms of spatial resolution and onthe order of milliseconds in terms of time.Further, when measuring the hemodynamic

response following neural activity target spa-tial resolution is on the order of less than acentimeter, at a temporal resolution of lessthan a second. However, as shown in Fig. 2,the brain is active even without external stim-uli. In this context we must emphasize onceagain the fundamental importance of distin-guishing spontaneous activity from activityevoked by a stimulus.

4 Theory of measurement of lightabsorption

Equipment for non-invasive measurementof brain activity using light can be dividedinto two categories: devices that translate val-ues calculated by the modified Lambert-Beer’s law into images; and devices that con-vert absorption coefficients into imagesthrough an inverse calculation of a light-diffu-sion equation. Included among the formercategory are optical topography and near-infrared spectroscopic (NIRS) imaging sys-tems, while the latter category consists prima-rily of optical CT (computed tomography) ordiffused tomography. Devices of the formertype currently play an active role in brainresearch. However, reliable results will not beobtained unless the special features of thesedevices are adequately understood. Here wewill introduce a brief history of the theoryinvolved, beginning with Lambert-Beer’s law(which has come to form the underlying basisof hemoglobin-related calculations) and lead-ing up to the modified Lambert-Beer’s lawcurrently in use.

EDA Hideo 75

From Neural activity to BOLDFig.2

76

4.1 History of Lambert-Beer’s lawLambert-Beer’s law forms the basis of

measurement by light. This indicates, asshown in Fig. 3, that absorbance, expressedlogarithmically as an intensity ratio of incidentlight to detected light, is proportional to theproduct of the distance between the site ofincidence and the site of detection and theconcentration of the target material. This lawis occasionally simply referred to as Beer’slaw. However, if one were to list the names ofthe researchers contributing to this law inorder, it would properly be referred to asBouguer-Lambert-Beer’s law. We describethis law throughout as Lambert-Beer’s law.

A brief history of this law begins with ref-erence to the writings of S. F. Johnston [2] andother writings. In the 18th to 19th centuries,when this law was formulated, science wasnot as specialized as it is today. The mostwell-known figures in this era producedachievements in various fields. Thus it isinteresting to note that none of these figuresmay be referred to as a researcher in a specificfield.

Pierre Bouguer, a French physicist andgeodesist (1698–1758) published his “Essaid’optique sur la generation de la lumière” onthe radiation of light in 1729. Bouguerassumed that light radiation obeys a particularlaw (the inverse-square law) stipulating thatthe radiation of light is in inverse proportionto the square of the distance traveled. Hedeveloped original observation equipment,established a method of determining bright-ness with the naked eye, illustrating the con-

cept with a comparison of brightness betweenthe moon and a candle. Further, Bouguerreported in his essay that a change in bright-ness shows exponential dependence on thedistance from the light source, based on theresults of these observation experiments.

Johann Heinrich Lambert (1728–1777),born in France, was a philosopher, a physicist,a mathematician, and an astronomer; he pub-lished his “Photometria” in 1760, presenting aformulation of the phenomenon in which theintensity of light decreases as an exponentialfunction of distance from the light source, asfollows:

In this formula, I denotes observed lightintensity, I0 irradiated light intensity, d the dis-tance from the light source, and k a propor-tional constant.

In brief, the phenomenon in which theintensity ratio of detected light decreasesexponentially with an increase in distancebetween detected and incident light was firstobserved by Bouguer and later formulated byLambert.

August Beer (1825–1863), a Germanmathematician, chemist, and physicist, pub-lished a paper in 1852 in which he discussedthe phenomenon in which light decreasesexponentially according to the concentrationof the target material. He published his workon optics entitled “Einleitung in die höhereOptik” in 1854.

As described above, Lambert-Beer’s law,the foundation of all current spectroscopy the-ory, was established through the achievementsof Bouguer, Lambert, and Beer.

The formula is expressed as follows:

where C denotes the concentration of thematerial. Since absorbance itself is a non-dimensional figure, the proportionality coeffi-cient k of this formula features the followingdimension: {inverse of (product of distance d

Journal of the National Institute of Information and Communications Technology Vol.51 Nos.3/4 2004

Lambert-Beer's lawFig.3

EDA Hideo 77

and concentration C)}; this is referred to as themolecular extinction coefficient. The princi-ple of analysis applied by the spectrophotome-ter entails the measurement of I/I0 and compu-tation of concentration C through prior deter-mination of the coefficient k and the opticalpath length of distance d.

4.2 Practice of spectroscopic analysisThe fundamental structure of the spectro-

scopic analyzer includes a light source, amonochromatizing device, a detector, and acell in which the sample is housed, amongother components. Although Lambert-Beer’slaw is a simple and convenient method ofquantitatively measuring concentration, cer-tain limitations to the application of this lawhave been indicated. For example, the lawassumes a sample target featuring a low densi-ty concentration; also, if the sample is local-ized, flattening occurs; additionally, the lawapplies only to one-dimensional quantitativemeasurement of a sample containing onlyabsorbing materials.

If the concentration is dense, the problemof molecular interaction arises, and conse-quently the relation between absorbance andconcentration deviates from a straight line. Asshown in Fig. 4, when the material is non-uni-form i.e., when it is “localized,” a phenome-non referred to as “flattening” may occur.This can easily be understood by thinking ofthe relationship between hemolyzed hemoglo-bin and red blood cells. If red blood cells arehemolyzed, hemoglobin is distributed uni-formly and absorption measurement becomespossible. In contrast, if hemoglobin existsonly in red blood cells, the measured concen-tration often appears very low, even when theactual concentration is the same as in the caseof even, uniform distribution. This is because,in the case of non-hemolyzed red blood cells,light passes through the red blood cells withno absorption; thus the peak of the absorptionspectrum appears smaller (i.e., flattened). Theterm “flattening” is derived from this phenom-enon. Finally, if a tissue contains non-absorb-ing substances, light may be diverted from the

detector by scattering. Since determination asto whether attenuation observed by the detec-tor is caused by absorption or by scattering isimpossible, the presence of scattering mayrepresent a source of error.

Each device has its own dynamic range ofmeasurement, and these devices often featurea sample holder approximately 1 cm length.Based on this hardware-specific restriction,light transmitted through a particularly thicksample may be too weak to measure. In thiscase, the sample is diluted until it is measura-ble. Furthermore, two samples, a referenceand a sample, are normally prepared. Themeasured amount of light for the reference isexpressed by I0 and the measured amount oflight for the sample is expressed by I. Con-centration is then calculated based on the dif-ference between I0 and I. In other words, anappropriate reference is chosen depending onthe object of measurement. However, thisprocedure is not applicable in non-invasivehuman measurement. This is because whenmeasuring human hemoglobin, the use of asample holder is not applicable and living tis-sue cannot be diluted. Moreover, it is difficultto prepare a living reference sample in whichthe amount of hemoglobin is zero.

4.3 Treatment of light scatteringIn spectroscopic analysis, the effects of

phenomena other than absorption are roughlyexpressed as scattering. As shown in Fig. 5,scattering consists of light traveling in direc-tions other than a straight line between thelight source and the detector. However, to be

Flattering by localizationFig.4

78

more accurate, if we imagine light collidingwith particles in the sample one after another,collisions that decrease energy are collectivelyreferred to as absorption and collisions that donot affect energy levels are collectivelyreferred to as scattering.

Particles below a certain size in an absorb-ing material will cause light to be scattered.Theoretical analyses assuming such non-homogeneity had been proposed from the1960s through the 1970s, and a number ofrelated reports are available. One avenue ofresearch involved measurement of hemoglo-bin concentration without hemolyzation [3].Extension of studies from single scatteringcaused by a single particle to multiple scatter-ing caused by multiple particles is also beingconsidered [4].

Optical scattering includes both scatteringthat changes the wavelength and scatteringthat does not change the wavelength. Insomatometry, Rayleigh scattering and Miescattering [5], which do not change the wave-length, must be taken into consideration. Ifscattering is handled as an interaction betweena photon and particles, it is first necessary todetermine the wavelength and the size of therelevant particles. If the scattering particlesare smaller than the wavelength of light, theRayleigh scattering formula is used. (Therelationship between particle size and wave-length is often used to explain why the sky isblue and why sunsets are red.) Moreover,when the scattering particles are of approxi-mately the same size as the wavelength of

light, the Mie scattering formula (which ismore complex than the Rayleigh scatteringformula) is instead applied in analysis.Assuming a measurement wavelength of 800nm, the size of a single cell will be equivalentto or smaller than the wavelength, whereas thered blood cell is on the order of a few micronsand hence is comparable to the wavelength insize. However, the living body is inhomoge-nous at all levels, from red blood cells to ves-sels to tissues and internal organs. It is in anycase awkward to apply a single scattering the-ory based on a given physical treatment and toperform analysis accordingly; scattering there-fore must be treated in a macroscopic manner.One procedure to treat scattering macroscopi-cally involves an optical diffusion equationand the modified Lambert-Beer’s law.

4.4 From a transport equation to anoptical diffusion equation

Generally, irradiating light to the livingbody and generating an image based onabsorption is referred to as quantitative meas-urement within a “turbid” system. An exam-ple of a turbid system can be seen in the use ofautomobile headlights to view the road in adense fog. Monitoring the ground from asatellite through cloud cover is a similarexample.

When analyzing how light is transmittedwhile being absorbed and scattered on amacroscopic level, the transport equation [6][7], or the radiation equation [8] can be applied.The transport equation has been used todescribe the movement of electrons and holesin a semiconductor and to illustrate the move-ment of thermal neutrons in a nuclear reactor.The radiation equation has been used foranalysis of optical multiple scattering in theatmosphere. If both equations are moderatelyapproximated in simplified forms, one canobtain an optical diffusion equation with threeparameters: an absorption coefficient (1/mm),a scattering coefficient (1/mm), and an angulardistribution of scattering. We have adoptedthis optical diffusion equation as the govern-ing equation. Figure 6 summarizes a method

Journal of the National Institute of Information and Communications Technology Vol.51 Nos.3/4 2004

ScatteringFig.5

of deriving the diffusion equation from thetransport equation according to an approachproposed by Kaltenbach et al. [9]. By diffu-sion approximation, the two parameters of thescattering coefficient and the angular distribu-tion of scattering are now approximated byisotropic scattering, with the product of thetwo applied as a reduced scattering coeffi-cient. Conventionally the diffusion constant Dhas been expressed using the formulas for theabsorption coefficient and for the reduced-scattering coefficient, but this is in fact unrea-sonable, as this assumes that energy is lost dueto absorption in the course of diffusion.Furutsu and Yamada focused attention on thisproblem and proposed a strict solution for thedefinition of D [10]. Subsequently, a formulacontaining only the reduced scattering coeffi-cient has come to be used. As a result,research on the diffusion equation progressedsignificantly in the course of research intooptical CT in the late 1980’s and through the1990’s. However, imaging equipment basedon the diffusion equation has yet to enter com-mon use in the experimental field, largelybecause the calculations involved are extreme-ly time-consuming.

4.5 Modified Lambert-Beer’s lawLambert-Beer’s law is simple and conven-

ient, and consequently its use in non-invasivehuman measurement has been subject toextensive investigation; the law is in fact cur-rently under application in optical measure-ment [11]. This investigation has given rise towhat is referred to as the modified Lambert-Beer’s law, whose characteristics have givenus the disadvantages of current topographicoptical brain functional imaging. Therefore,an understanding of the modified law is essen-tial.

When one intends to measure the livingbody, including scattering elements, using thesites of reflection and the like, Lambert-Beer’slaw can no longer be applied simply. The fol-lowing equation is applied in a formulationsimilar to Lambert-Beer’s law.

For a structure such as that shown in Fig.7, a linear relation is assumed between thechange in concentration and the change inabsorbance over a very short interval; theeffect of the change in scattering is then

EDA Hideo 79

Diffusion equation from the transport equationFig.6

80

added. In order to calculate the concentrationfrom measured values, it is sufficient to deter-mine the three parameters appearing in thisformula. Here, denotes the molecular extinc-tion coefficient, C the change in concentra-tion, <d> the mean optical path length, and Sthe scattering change. The appropriatemethod of determining these coefficients haslong been the subject of investigation. For acase in which the concentration of hemoglobinis measured, three methods are available foreach of the three parameters.

1 Molecular extinction coefficient(1) Using the spectrum of pure hemoglobin;

i.e., a reference spectrum or an experimen-tally obtained spectrum for hemolyzedblood(Note that since there are two spectra forhemoglobin—one for the hemoglobin mol-ecule and the other for the heme protein—care must be taken to select the appropri-ate spectrum.)

(2) Using a spectrum measured in vitro forhemoglobin into which a scatterer, such asmilk, is mixed

(3) Using a spectrum measured in vivo for arat head or other animal subject

2 Mean optical path length(1) Setting to unity by ignoring the actual opti-

cal length(2) Determining the coefficient for the optical

path and multiplying the irradiation-to-detection distance by the coefficient

(3) Actual measurement of optical path length

for each wavelength involved in measure-ment

3 Scattering change(1) Setting to zero by ignoring the actual scat-

tering change(2) Setting to a constant featuring no depend-

ence on the wavelengths used in measure-ment

(3) Setting to a different value for each meas-urement wavelength. With respect to the molecular extinction

coefficient 1 (1), Figure 8 shows the spectrafor two types of hemoglobin. The left-handvertical axis represents the spectrum of ahemoglobin molecule obtained by Matcher[12], and the right-hand vertical axis representsthe spectrum of a heme protein by Zijlstra [13].The hemoglobin molecule features four hemeproteins. The four-fold quantitative differencebetween the right-and left-hand side verticalaxes reflects the differences between thehemoglobin molecule and heme protein.When the change in hemoglobin is calculatedusing this spectrum, the same measured valuesresult in an apparently four-fold difference,and care must be exercised to take this differ-ence into account. In the hemoglobin calcula-tion, the inverse matrix of the spectrum is usedin multiplication, and consequently the hemo-globin calculated using the spectrum of thehemoglobin molecule becomes 1/4 times thevalue calculated using the spectrum of theheme protein.

As is the case with ordinary calculations

Journal of the National Institute of Information and Communications Technology Vol.51 Nos.3/4 2004

Modified Lambert-Beer's lawFig.7

Hemoglobin spectraFig.8

using Lambert-Beer’s law, when the materialto be measured consists of two substances,absorbance is calculated as the sum of theconcentrations of the materials. Therefore, ina system in which the two components of oxy-genated hemoglobin (oxyHb) and deoxygenat-ed hemoglobin (deoxyHb) coexist, all that isrequired to calculate absorbance values for thetwo is to perform measurement using two ormore wavelengths. Figure 9 shows a methodof calculating hemoglobin based onabsorbance when the molecular extinctioncoefficient is that of the hemoglobin molecule(method 1 (1) above corresponding to thisparameter), the mean optical path length isunity (method 2 (1) for this parameter), andthe scattering change is zero (method 3 (1) forthis parameter). Since measured values atthree wavelengths are obtained for twounknown parameters—oxyHb and deoxyHb—the concentration changes are calculated afterdetermining a generalized inverse matrix.This procedure is equivalent to calculationusing the least-squares method.

4.6 Problems with the modified Lam-bert-Beer’s law

Generally, since linear approximationworks only for very small sections, when the

section becomes large error will increase.That is, when the change in concentrationgrows large in the modified Lambert-Beer’slaw, the measured value may not fully reflectthis change. For reference, Fig. 10 shows asolution of the optical diffusion equation in aninfinite medium. Actual optical propagation ismore accurately shown by the solution of thisdiffusion equation for an actual medium. Ascan be seen in the graph, it is understood thatthe modified Lambert-Beer’s law, whichassumes a linear relation between concentra-tion change and absorbance change, postulatesa tangent to the actual curve. The gradient ofthis tangent is equivalent to the product of themolecular extinction coefficient and the meanoptical path length. In our approach, in whichthe mean optical path is ignored, the value forhemoglobin is not calculated in units of con-centration, but instead in a strange compoundunit: the product of concentration and dis-tance.

It is also significant that the gradient of thetangent varies depending on the position ofcontact. When calculating absorbance change,it is often the case that the original point ofabsorbance is set to a value corresponding tothe time at the start of measurement. Thisoriginal point of absorbance corresponds to

EDA Hideo 81

Hb parameters calculationFig.9

82

the position of contact. Without a change inabsorbance, no change in concentration willoccur as far as the modified Lambert-Beer’slaw is concerned. In other words, the meas-urement start time that provides the point ofreference is always zero. This means that caremust be taken when comparing measured databetween different subjects or data measured atdifferent times.

5 Brain activity measurement bynear infrared light

The near infrared light used for somatome-try is of wavelengths in the vicinity of 800nm. Light in this wavelength region easilypenetrates the living body for two reasons: itis subject to particularly low absorption inwater (at wavelengths longer than 1,000 nm,light is sharply absorbed by water) and to lowabsorption in the blood (at wavelengths short-er than 600 nm, light is sharply absorbed byhemoglobin in the blood). Therefore, thesewavelengths can be used to acquire informa-tion on the interior of the living body in thesame fashion as X-rays. Figure 11 shows an

optical absorption spectrum of water at wave-lengths from 200 nm to 100,000 nm [14].From the figure, it can be seen that absorptionis low in the vicinity of 800 nm.

Research into somatometry systems usingnear infrared light was initiated with a report[15] by Jobsis published in 1977. Thisresearch began with the development of a sin-gle-channel oxygenation monitor [16] andapplication of this device [17], progressed tooptical CT [18], and is now focused on opticaltopography [19] and near-infrared spectroscop-

Journal of the National Institute of Information and Communications Technology Vol.51 Nos.3/4 2004

Analytical solution of the Diffusion equationFig.10

Water spectraFig.11

ic (NIRS) imaging systems [20]. These sys-tems calculate changes in oxyHb and deoxy-Hb as indexes of oxygen of the living bodyand display the Hb parameters together withthe totalHb change, represented as the sum ofthese changes. It may be said that it was thedevelopment of optical topography that beganto attract attention to brain-function research.Since these systems have spread so rapidly inbrain research, it may be said that the accumu-lated background research described above inthis article cannot have been subject to appro-priate follow-up verification. While a genera-tion of new researchers are beginning to famil-iarize themselves with near infrared imagingequipment, it seems to us that a growing cho-rus of dissenting voices are indicating the dis-advantages of using this equipment in brainresearch. Optical measurement does offerseveral advantages relative to other brain-activity measurement instruments. However,we must avoid foregoing reliability simply forthe sake of these advantages.

Here we will compare near-infrared andX-ray measurement. Since X-rays areabsorbed strongly by bones, these appear asshadows in X-ray images. Moreover, since X-rays travel in nearly a straight line through theliving body, bones will appear in images with-out blurring, as when an X-ray image is takenof the palm. X-rays are also absorbed byblood and tissues, in addition to bones. Sincethe relevant absorption coefficients are allknown, if the scale of the displayed image isadjusted, a given target can be observed with-in the living body. Thus, since differences inthe absorption coefficient correspond to tissuedifferences in the case of X-rays, visualization(i.e., imaging) of the living body becomespossible. On the other hand, when nearinfrared light is irradiated to the palm (innerside) and the palm is viewed from the outsidewith a dedicated camera, vessels, as opposedto bones, may be seen. This is because in theliving body, near infrared light is stronglyabsorbed by the hemoglobin contained in thered blood cells in the blood, enabling an imageof the vessel to be formed. (However, since

these rays are simultaneously scattered by thetissues, the image suffers a certain degree ofblurring.) At the same time, since the hemo-globin contained in the blood serves as anoxygen-transporting material, and changescolor depending on its state of oxidation, theoxygenated state of hemoglobin can be calcu-lated from the absorption coefficient measuredwith light at multiple wavelengths. In brief,unlike the structural image obtained using anX-ray, the use of multiple-wavelength nearinfrared light enables us to obtain a functionalimage of oxidation states.

6 Equipment for imaging opticalabsorption

Figure 12 shows the equipment we havemanufactured for imaging light absorption,showing probes attached to the head of thesubject. This consists of a light source (con-sisting of near-infrared semiconductor lasersat three wavelengths: 780 nm, 805 nm, and830 nm), light guides (compound glass fibers),photodetectors (photomultiplier tubes), a con-trol unit, a display unit, as well as additionalcomponents. This prototype device includessix sets of light-source and irradiation lightguides, photodetectors, and photodetectorlight guides, enabling measurement of irradia-tion and reception using six device pairs. Atpresent these can be extended to up to 16 sets.The device was designed to enable externalanalog input. The 12 strands of optical fiber(six strands for irradiation and six strands forlight reception) are arranged in a predeter-

EDA Hideo 83

Optical systemsFig.12

84

mined manner and used in measurement,enabling image display. In order to performthis imaging, however, it is in principle firstnecessary to solve an inverse problem of lightpropagation. However, to avoid this initialstep, variation of the optical signal is assumedto result only from hemoglobin change; signalvalues are then arranged and subjected tointerpolation to form an image.

7 Measurement of occipital areaduring experimental visual stim-ulation

During visual stimulation using a checker-board pattern, data is measured with the endsof optical fibers disposed on the scalp over theoccipital area of the subject, as shown in Fig.13. The optical measurement data is convert-ed to an image of hemoglobin change and dis-played superimposed on an MRI image to cre-ate the image shown in Fig. 14. An MRIimage processed as an X-ray image was usedas a template. This display enables us tomeasure hemodynamic response variation inthe visual area for a given stimulus.

Figure 15 is an illustration of a process ofsignal analysis. The upper-right figure showsthe timing of the stimulus. In this case, thetask is given three times. The second figure inthe right-hand column shows the measureddata as absorbance (Abs) after logarithmictransformation, which serves as raw data forthis process. Absorbance was subjected tobaseline correction. Here, “baseline” refers toa straight line connecting the respective Absvalues corresponding to the start of each task.The change in absorbance can be calculatedby subtracting this baseline signal from theraw data. As can be understood from theupper right figure, it was determined that thelight was able to capture a state of brain activi-ty even without the addition of other dataitems. The change in the Hb parameter wascalculated based on this change in absorbanceto obtain the second figure in the right-handcolumn. The three data items are then addedto obtain the figure at bottom right. From the

figure, it was found that oxyHb increases anddeoxyHb decreases with a visual stimulus.These peak values are arranged, interpolated,and subsequently placed within an image tocreate the image shown in Fig. 14. In thisexperiment, temporal resolution was set to 0.5seconds. In short, a single image can be dis-played every 0.5 seconds; this means that thedevice is capable of capturing the dynamicactivity of the brain with sufficient accuracy.

The data obtained is initially no better thana tile or check array, where the diagonals ofeach tile are formed by illumination and detec-tion fibers because the two types of opticalfibers are arranged alternately. A simplemethod of obtaining a topographic imageinvolves two-dimensional interpolation of twoor more items of initial data. Various mathe-matical interpolation procedures are available,such as linear interpolation and spline interpo-lation, and each procedure gives a slightly dif-

Journal of the National Institute of Information and Communications Technology Vol.51 Nos.3/4 2004

CheckerboardFig.13

NIRS image and fMRIFig.14

ferent image. Though interpolation among tiledata is likely to improve the spatial resolutionof images, we must keep in mind that the orig-inal data is in the form of tiles, not completeimages. Discussion of the resolution of topog-raphy is hardly meaningful; if we consider thematter carefully, we must recognize that reso-lution merely consists of differences betweeninterpolation procedures. Interpolation itselfis an operation in which data is converted intoan easy-to-view format, and cannot be viewedas a method of improving spatial resolution.

If spatial resolution is defined as the dis-tance at which two elements can be distin-guished one-dimensionally, the spatial resolu-tion of a topographic image will, strictlyspeaking, be determined as three tiles, wherethe diagonals of each are formed by illumina-tion and detection fibers. This is becauseempty spaces between image spaces must alsobe measured. Assuming that each tile side is2.5 cm, three tiles will measure 7.5 cm.

In measuring the brain, it is safest todevise a task such that two or more activationsites are not located in an area as large as threetiles. In this experiment, the right and leftbrain areas corresponding to visual processeswere successfully imaged.

8 Features of optical measure-ment equipment using nearinfrared light

As described above, a device displaying atopographic image using light is capable ofdisplaying the extent of change in oxyHb,deoxyHb, and totalHb as an image in near realtime with spatial resolution of a few centime-ters and temporal resolution of 1 second orless.The following can be proposed as advantagesof optical measurement equipment:- Ease of measurement- Enables measurement near the subject- Extremely low device maintenance and run-

ning costs, comparable to costs of EEGequipment

- High temporal resolution; capable of dis-playing Hb images at a rate of one or moreframes per second in real time

- Excellent S/N ratio, thus images can beobtained without being subjected to statisti-cal operation

- Subject may move slightly insofar as fibersare affixed to the head

- The equipment enables observation of hemo-dynamic response changes based on optical-absorption changes in hemoglobin

EDA Hideo 85

Signal analysisFig.15

86

On the other hand, the following disadvan-tages may also be indicated:- Concentration of hemoglobin not linearly

related to hemoglobin amount as determinedby light-based measurement

- In applying the diffusion equation, opticalproperties (absorption coefficient, scatteringcoefficient, phase function) of the humansubject have yet to be determined

- Limited spatial resolution- Difficult to resolve space among tissues in

the direction of depth- Despite ease of measurement, successful

measurement is questionable, entailing veri-fication of numerous factors, e.g., S/N ratio

The first drawback mentioned above rep-resents a fundamental problem. As shown inFig. 15, when the hemoglobin parameters arecalculated based on absorbance, a generalizedinverse matrix of Hb absorption spectrum val-ues is used as the coefficients. The procedureis originally predicated on the linear relation-ship between the change in hemoglobin con-centration and the change in the measuredvalue. However, the solution of the diffusionequation reveals that this relationship is notlinear. In other words, the gradient of thecoefficient obtained by linear approximation

between the hemoglobin concentration and themeasured value varies depending on the initialvalues of the reduced scattering coefficientand the absorption coefficient (assumed as ref-erence values in calculating the absorbancechange). This gradient corresponds to what isknown as the “optical path length factor.” Thevalue of this gradient must be verified todetermine whether it remains the same fromthe initial time of measurement through theremainder of measurement. Various structuressuch as the cranial bone determine scatteringand absorption patterns in the human head,which indicates that the functional imageshowing hemoglobin change is affected by thestructural image of the head, as shown in Fig.16.

9 Conclusion — What type ofbrain research makes mosteffective use of optical meas-urement? —

The advantages and disadvantages of opti-cal measurement by near infrared light weredescribed above. Essentially two points mustbe examined when applying this type of meas-urement to brain research. These points are

Journal of the National Institute of Information and Communications Technology Vol.51 Nos.3/4 2004

Structural imageFig.16

described below.- If hemodynamic response information is to

be obtained, fMRI is also available. Howev-er, quantitative values of oxyHb, deoxyHb,and totalHb cannot be measured by fMRI;whether these values are essential in thebrain research in question should be consid-ered.

- If mapping is the main purpose of the brainresearch, it is important to determinewhether optical equipment featuring a cer-tain spatial resolution can be used in prac-tice.

The first point above involves differentia-tion from fMRI and the second point involvesa question about the original goal of the brainresearch to be conducted by the researchersinvolved. Certainly, if optical measurement isused only to obtain data available using fMRI,optical measurement is not necessary. On theother hand, if the final purpose of the brainresearch is to search sites within the brain—

i.e., mapping—equipment offering high spa-tial resolution must be employed.

However, there are cases in which fMRImeasurement cannot be applied, as with meas-urement of infants. With optical measure-ment, an infant’s brain can be measured whilethe child is held by the mother. Moreover, wehave succeeded in measuring brain activitywhile the subject was exercising [21] [22]. Thisis data that cannot be provided by fMRI, EEG,or MEG devices. It is assumed that opticalmeasurement can reveal brain activitiesaccompanying actions, indicating potentialapplications in neuro-rehabilitation. Weexpect that future development will lead inthis direction. Furthermore, in an increasingnumber of studies the interpretation of fMRIdata is elucidated with the help of optical data[23]; we thus anticipate that optical measure-ment will develop into an essential componentof brain research.

EDA Hideo 87

References1 S. Ogawa, D. W. Tank, R. Menon, J. M. Ellermann, S. Kim, H. Merkle, and K. Ugurbil, “Intrinsic Signal

Changes Accompanying Sensory Stimulation: Functional Brain Mapping with Magnetic Resonance Imag-

ing”, PNAS 89: 5951-5955, 1992.

2 S. F. Johnston, “A History of Light and Colour Measurement: Science in the Shadows”, Institute of Physics

Publishing, 2001.

3 F. J. Janssen, “A study of the absorption and scattering factors of light in whole blood”, Med. & Biol. Engng.

Vol.10, pp231-240, 1972.

4 A. Ishimaru, “Wave Propagation and Scattering in Random Media”, Academic Press, 1978.

5 M. Born and E. Wolf, “Principes of Optics -Seventh edition”, Cambridge University Press, 1999.

6 K. M. Case and P. F. Zweifel, “Linear Transport Theory,Addison-Wesley”, 1967.

7 J. J. Duderstadt and L. J. Hamilton, “Nuclear Reactor Analysis”, JOHN WILEY & SONS, 1942.

8 S. Chandrasekhar, “Radiative transfer”, Dover, 1960.

9 J. M. Kaltenbach and M. Kashke, “Frequency- and Time-domain modeling of light transport in random

media”, Medical Optical Tomography, SPIE IS 11, 65-86, 1993.

10 K. Furutsu and Y. Yamada, “Diffusion approximation for a dissipative random medium and the applications”,

Phys. Rev. E, 50, 3634-3640,1990.

11 D. T. Delpy, M. Cope, P. van der Zee, S. Arridge, S. Wray, and J. Wyatt, “Estimation of optical pathlength

through tissue from direct time of flight measurement”, Phys. Med. Biol. 33, 1433-1442, 1988.

12 S. J. Matcher, C. E. Elwell, C. E. Cooper, M. Cope, and D. T. Delpy, “Performance comparison of several

published tissue near-infrared spectroscopy algorithms”, Analytical Biochemistry, 226, 54-68, 1995.

13 W. G. Zijlstra, A. Buursma, and W. P. Meeuwsen-van der Roest, “Absorption spectra of human fetal and

88 Journal of the National Institute of Information and Communications Technology Vol.51 Nos.3/4 2004

EDA Hideo, Ph.D.

Expert Researcher, Brain InformationGroup, Kansai Advanced ResearchCenter, Basic and Advanced ResearchDepartment

Biomedical Engineering, MedicalOptics

adult oxyhemoglobin, de-oxyhemoglobin, carboxyhemoglobin, and methemoglobin”, Clin. Chem. 37, 1633-

1638, 1991.

14 G. M. Hale and M. R. Querry, “Optical constants of water in the 200 nm to 200µm wavelength region”, Appl.

Opt., 12, 555--563, 1973.

15 F. F. Jobsis, “Noninvasive, Infrared Monitoring of Cerebral and Myocardial Oxygen Sufficiency and Circulato-

ry Parameters”, Science, 198, 1264-1267, 1977.

16 T. Tamura, H. Eda, M. Takada, and T. Kubodera, “New instrument for monitoring hemoglobin oxygenation”,

Adv Exp Med Biol, 248 : 103-107, 1989.

17 H. Eda, I. Oda, Y. Ito, Y. Wada, Y. Oikawa, Y. Tsunazawa, M. Takada, Y. Tsuchiya, Y. Yamashita, M. Oda,

A. Sassaroli, Y. Yamada, and M. Tamura, “Multichannel time-resolved optical tomographic imaging system”,

Review of Scientific Instruments, Sep/1999, Vol.70, Issue.9, pp.3595-3602,1999.

18 S. Homma, H. Eda, S. Ogasawara, and A. Kagaya, ”Near-infrared estimation of O2 supply and consumption

in forearm muscles working at varying intensity”, J. Appl. Physiol., 80(4),pp1279-1284,1996.

19 A. Maki, Y. Yamashita, Y. Ito, E. Watanabe, Y. Mayanagi, and H. Koizumi, “Spatial and temporal analysis of

human motor activity using noninvasive NIR topography”, Med Phys. 22(12), 1997-2005, 1995.

20 H. Eda, I. Sase, A. Seiyama, H. C. Tanabe, T. Imaruoka,Y. Tsunazawa, and T. Yanagida, “Optical Topogra-

phy System for Functional Brain Imaging:Mapping human occipital cortex during visual stimulation”, Pro-

ceedings of Inter-Institute Workshop on In Vivo Optical Imaging at the NIH, OSA, 93-99, 2000.

21 I. Miyai, H. C. Tanabe, I. Sase, H. Eda, I. Oda, I. Konishi, Y. Tsunazawa, T. Suzuki, T. Yanagida, and K. Kub-

ota, “Cortical Mapping of Gait in Humans: A Near-Infrared Spectroscopic Topography Study, NeuroImage”,

Volume 14, Issue 5, 1186-1192, 2001.

22 H Eda, I Miyai, K Kubota, and T Yanagida, ”Cortical mapping of gait by Near infrared spectroscopic imaging

Rinsyo Noha”, Vol.44, No.12, 751-757,2002.(in Japanese)

23 H Eda, I Sase, A Takatsuki, A Seiyama, T Yanagida, and S Miyauchi, ”BOLD Signal by fMRI agrees eith

totalHb signal by NIR imaging system”, Technical report of IEICE.MBE2001-97,95-99,2001. (in Japanese)