In vivo optical path lengths and path length resolved doppler shifts of multiply scattered light

Lasers in Surgery and Medicine 42:692–700 (2010)

In Vivo Optical Path Lengths and Path Length ResolvedDoppler Shifts of Multiply Scattered Light

Babu Varghese,1* Vinayakrishnan Rajan,1 Ton G. Van Leeuwen,1,2 and Wiendelt Steenbergen1

1MIRA Institute for Biomedical Technology and Technical Medicine, Biomedical Photonic Imaging Group,University of Twente, P.O. Box 217, 7500 AE Enschede, the Netherlands

2Academic Medical Center, Biomedical Engineering & Physics, Meibergdreef 9, 1105 AZ Amsterdam, the Netherlands

Background and Objectives: In laser Doppler measure-ments, perfusion values averaged over different and basi-cally unknown path lengths are recorded. To facilitatequantitativepathlengthresolvedperfusionmeasurements,we developed a phase modulated Mach–Zehnder interfer-ometer with spatially separated fibers for illumination anddetection. The goal of this study is to measure in vivo opticalpath lengths and path length resolved Doppler shifts and tocompare these with conventional laser Doppler perfusionmeasurements.Study Design/Materials and Methods: With a phasemodulated Mach–Zehnder interferometer, we performedpath length resolved perfusion measurements on humanskin and its variations to external stimuli and comparedthese with conventional laser Doppler perfusion measure-ments. The method was evaluated in three humansubjects on the dorsal side of the forearm to establishinter-individual within-site variations. Measurementswere performed at three different locations of oneindividual for observing intra-individual inter-sitevariations resulting from the heterogeneity of the tissue,both in the static matrix and in the microvasculararchitecture of the skin. In all measurements, perfusionwas simultaneously measured with a conventional laserDoppler perfusion monitor.Results: In this study, we show the first results of pathlength resolved perfusion measurements in skin and itsvariations to occlusion and Capsicum cream provocation.From our data, we deduced the Doppler shifted fraction ofphotons, which is related to the blood volume, and the pathlength dependent average Doppler shift, which is related tothe mean velocities of red blood cells. The Doppler shiftedfraction of photons is decreased from 28% to 18% duringocclusion and increases to 41% when capsicum cream wasapplied to the skin. Inter- and intra-individual inter-sitemeasurements demonstrated variations in optical pathlength distributions and path length resolved Dopplershifts. The Doppler shifted fraction of photons measuredon the fingertip is about 38% and that measured on thedorsal and palmar sides of the forearm are 32% and 17%,respectively. The path length distributions depend on theskin site that is being probed and the intra-individual inter-site variability is higher than the inter-individual within-site variability measured on comparable sites betweendifferent individuals.

Conclusions: In this study, we demonstrated, for the firsttime to our knowledge, that in vivo path length resolvedperfusion measurements are feasible. Optical path lengthdistributionsofmultiplyscatteredlight,spanningarangeof0–6 mm, and their response to external stimuli such asocclusion and capsicum cream provocation have beenmeasured. This method will enable better interpretationof inter- and intra-individual inter-site variations in theLDF readings that are introduced by the variance intissue optical properties. The inter- and intra-individualinter-site variations measured with our setup resultsindicate that that these variations should be taken intoaccount while comparing the perfusion readings fromcomparable sites between individuals and from differentsites of the same individual. Furthermore, the observedinter- and intra-individual inter-site variations in pathlength resolved Doppler measurements indicate theinherent limitation of conventional LDPM that restrict itsclinical usefulness, due to its dependence on the unknownphoton path length. Consequently, this method will enableto correctly interpret or counter-act the inter- and intra-individual inter-site variations in the LDF readingsintroduced by the variance in tissue optical properties.This approach also enables to discriminate between theDoppler-shifted photons resulting from interaction withthe moving red blood cells and the non-shifted lightscattered only by the surrounding static tissue matrices.Lasers Surg. Med. 42:692–700, 2010.� 2010 Wiley-Liss, Inc.

Key words: laser Doppler perfusion monitoring; opticaldiagnostics for medicine; medical optics instrumentation;low coherence interferometry

Contract grant sponsor: Netherlands Technology FoundationSTW; Contract grant number: TTF 5840; Contract grant sponsor:Institute for Biomedical Technology of the University of Twente,Perimed AB (Stockholm, Sweden).

*Correspondenceto:BabuVarghese,PhilipsResearchHighTechCampus34,RoomWB7.013,Postbox7.071,Eindhoven5656AE,theNetherlands. E-mail: [email protected]

Accepted 9 August 2010Published online 15 October 2010 in Wiley Online Library(wileyonlinelibrary.com).DOI 10.1002/lsm.20969

� 2010 Wiley-Liss, Inc.

INTRODUCTION

Laser Doppler blood flowmetry is a non-invasive tech-nique for monitoring blood microcirculation in biologicaltissues [1]. However, perfusion measurements using theseconventional laser Doppler techniques are affected by probeinduced variations [2] and variations in tissue optical prop-erties in terms of absorption and scattering [1,3]. Thephysical explanation of this dependence is the variance inthe optical path lengths of the photon migration in tissue. Inconventional laser Doppler techniques, perfusion valuesaveragedoverdifferentandbasicallyunknownpathlengthsare recorded, which creates an uncertainty in the relationbetween the measured perfusion signal and the real per-fusion. A longer path length will increase the probabilitythat a Doppler shift will occur, thus yielding a relative over-estimationofthebloodperfusion,comparedtotheshortpathlength situation. Techniques for monitoring perfusion withpath length selectivity may overcome this limitation andenable to correctly interpret or counter-act the inter- andintra-individual inter-site variations in the perfusion read-ings introducedbythetissueopticalproperties.Thedepthoflight penetration in the tissue depends on the wavelength ofthe light source, the distance between the transmitting andreceiving fibers and the optical properties of the tissue. Thisdependence leads to uncertainties in interpreting theDoppler shifted and non-shifted fraction of photons and alsoin discriminating the fraction of light scattered from super-ficial (nutrient) and deeper (thermoregulatory) layers ofskin. As such, depth resolved perfusion information couldbeobtainedbymulti-wavelengthsystems[4],byvaryingthedistance between transmitting and receiving fibers [3], byusingamultichannel laserDopplerprobe[5]orbycoherencelength modulation of a semiconductor laser [6]. However,these methods still give no control over the optical pathlength traveled by the detected light.

The path-length-specific technique used in this studyinvolves a phase modulated low coherence Mach–Zehnderinterferometer with two spatially separated light-deliver-ing and light-collecting fibers as used in conventional laserDoppler perfusion monitors [7–9]. Compared to both theMichelson [10–12] and the Mach–Zehnder [13,14] basedinterferometric measurements that could only record theDopplershiftedphotons, ourmethod isable tomeasurepathlength resolved information from mixed media [7], such asflow of moving red blood cells within static tissue matrices.In low coherence interferometry, the limited temporalcoherence of the light source acts as a band pass filter inselecting the photons that have traveled a specific pathlength in the medium.

The goal of this study is to measure in vivo optical pathlengths and path length resolved Doppler shifts. In thisstudy, we show the first results of path length resolvedperfusion measurements in skin and its variations to occlu-sion and capsicum cream provocation. The method wasevaluated in three human subjects on the dorsal side ofthe forearm to establish inter-individual within-site vari-ations. Measurements were performed at three differentskin locations of one individual for observing intra-

individual inter-site variations resulting from the hetero-geneity of the tissue, both in the static matrix and in themicrovascular architecture of the skin. In all measure-ments, perfusion was simultaneously measured with a con-ventional fiber-optic laser Doppler perfusion monitor.

MATERIALS AND METHODS

The fundamental output quantity of a laser Doppler per-fusion monitor is the first moment of the power spectrumP(v) of the detector signal; in general, the ith moment isbeing defined as

Mi ¼Zb

a

PðvÞvidv (1)

Here a and b are device dependent low and high cut-offfrequencies. With i ¼ 0, a quantity is obtained which isproportional to the concentration of moving red blood cells,while i ¼ 1 describes red blood cell flux, which is theproductof concentration and the root mean square of the red cellvelocity, at least for low blood concentrations [15].

Theconventional laserDoppler perfusionmonitorused inour measurements is a PF5000 monitor (Perimed AB,Jarfalla, Sweden) with a laser diode (780 nm) as the lightsource (Output optical power ¼ 1 mW). The PF5000 has abandwidth of 20 Hz–13 kHz (a and b in Eq. 1) and a timeconstant of 0.2 seconds was used for monitoring. The fiber-optic probe (Probe 408, standard probe) consists of twospatially separated fibers (core diameter ¼ 125 mm,NA ¼ 0.37) with a center-to-center separation of 250 mm.The calibration of the PF5000 was done with motility stand-ard (water suspension of polystyrene microspheres with adiameter of 320 nm). After the calibration, PF5000 dis-played a standard reading of 250 perfusion units whenmeasuring in the motility standard.

In our phase modulated low coherence interferometrictechnique, we use a fiber-optic Mach–Zehnder interferom-eter (Fig. 1) with a superluminescent diode (Inject LM2-850, l ¼ 832 nm, DlFWHM ¼ 17 nm, coherence lengthLC ¼ 18 mm) that yields 2 mW of power from the single-mode pigtail fiber as the light source. Single mode fibers(mode field diameter ¼ 5.3 mm, NA ¼ 0.14) are used forillumination, while multimode graded-index fibers (corediameter ¼ 100 mm,NA ¼ 0.29)areusedfordetection,pro-viding a large detection window and a small modal dis-persion. The fibers are spatially separated by a center-to-centerdistanceof300 mmandareembeddedinablackepoxyresinsurroundedbyametaltubeofinternaldiameter6 mm.The reference beam is polarized using a linear polarizer andthe phase is sinusoidally modulated at 6 kHz using an elec-tro optic broadband phase modulator (New Focus Model4002) with a peak optical phase shift of 2.04 radians appliedto the modulator. The AC photocurrent is measuredwith a 12 bit analogue to digital converter (NationalInstruments, Texas, USA), sampling at 40 kHz, averagedover 1,000 spectra. We have used 1,000 samples per spec-trum, with a total data acquisition time of 52 seconds to get

PATH LENGTH RESOLVED PERFUSION MEASUREMENTS 693

an average of 1,000 spectra. The path length resolution thatweusedinthesemeasurementsis200 mm.Forwidelydiffer-ent optical path lengths where no heterodyne signal isobserved, we used a path length resolution of 500 mm.

In our instrument, for large phase modulation angles(Df ¼ 2.04 radians) the power spectra contain interferencepeaks at both the phase modulation frequency and higherharmonics (Fig. 2). When path length resolved informationfrom these high-order harmonics is utilized, the signal tonoise ratio is increased byalmost one order ormagnitude, ascompared to a situation where the peak phase modulationangle is kept lower to avoid higher harmonic peaks(Df ¼ 0.51). Optical path length distributions are obtainedby adding the areas of all interference peaks (after subtrac-tion of the background noise, and within a bandwidth of�2 kHz around all center frequencies) in the power spec-trum [9]. The Full-width half maximum (FWHM) of theinterference signal is about 50 Hz in a statically scatteringmedium whereas in the case of dynamic media more andmore power is set to frequencies around the phase modu-lation frequency, resulting in Doppler broadening [7]. Thus,the area of the Doppler broadened peak, excluding the stati-cally scattered light contribution at the interference peaks,formsanestimationof theamountofDopplershifted lightatthat specific optical path length. The Doppler fraction ofphotons corresponds to the fraction of light scattered frommoving red blood cells and is proportional to the concen-tration of moving blood cells in the probed volume for shortoptical path lengths and low blood volumes. The non-shiftedfraction of light represents the light that has been only

statically scattered by the surrounding static tissuematrices. The average Doppler shift corresponding to theDoppler shifted light is calculated from the weighted firstmoments (M1/M0) of the heterodyne peak at the modulationfrequency, after noise correction. The analysis is performedin a bandwidth of 50 Hz–2 kHz close to the phase modu-lation frequency and its higher harmonics, using

Mi ¼X3

j¼1

Zjvmþb

jvmþa

PðvÞðv�jvmÞidv (2)

The Doppler shift measured from the weighted firstmoments averaged over all optical path lengths will leadto a relative overestimation for long path lengths. To avoidthis overestimation in the perfusion signal, and to measureflux independent of variations in optical path length, henceindependent of tissue optical properties, Doppler spectrashould preferably be measured for known optical pathlengths, a feature which is allowed by the low coherenceinterferometry technique presented here. However, thispath length selective property of LCI makes comparisonwith laser Doppler flowmetry difficult, since LDF takesall path lengths into account. Therefore, to allow for somecomparison of the two techniques, Doppler data and inten-sities are obtained for a large number of path lengths usingLCI, and the Doppler shifts for all optical path lengths areweighted with the corresponding intensities of Dopplershifted photons for that specific optical path length L asfollows:

Weighted average Doppler shift ¼1N

PNi¼1

MDoppl1 ðLiÞ

PNj¼1

MDoppl0 ðLjÞ

(3)

Fig. 2. Power spectra measured on the dorsal side of the

forearm and without and with external stimuli: occlusion and

capsicum cream provocation. The interference peak appear-

ing at the modulation frequency (inset).

Fig. 1. Schematic of the fiber optic Mach–Zehnder interfer-

ometer. Single-mode and gradient index multimode fibers

are shown by thin and thick lines, respectively. SLD denotes

superluminescent diode, PD is the photodetector, LP is a lin-

ear polarizer, PM is the electrooptic phase modulator, 90:10,

and 50:50 are single-mode and multimode fiber couplers,

respectively.

694 VARGHESE ET AL.

Optical Path Lengths and Path Length ResolvedDoppler Shift Measurements

Measurements were performed on the skin of the dorsalside of the right forearm of a healthy human volunteer (skintype—type II) in the sitting position [16]. All the measure-mentswereperformedatnormalroomtemperature.Aprobeholder (Perimed PH 08) was attached to the skin with adouble-sided adhesive tape. This probe holder is designedinsuchamannerthattheforceontheskin isminimized.Themeasurements (normal, occlusion, capsicum cream) withthe PF5000 and the LCI probe were done at the same tissuelocation by using the same probe holder. The subject restedapproximately 10 minutes prior to the measurements. Thefiber optic probe of the PF5000 was inserted into the probeholder andthe perfusion was measured for comparison withthe succeeding measurements with our developed probe.Measurements were performed on unprovoked skin andperfusion variations due to occlusion (for 50 seconds) weremeasured for a given optical path length and this wasrepeatedforallopticalpathlengths.Tomeasurethechangesinperfusionduringocclusion,arterialocclusionof theupperarm was done with a cuff inflated to a pressure of 170 mmHg. The perfusion readings indicated by PF5000 in parallelreduced from 28 � �1 to 5 � 0.18 during occlusion. Weensured that the readings recorded with PF5000 showthe same normal perfusion measured 10 minutes afterthe occlusion measurements. After this, we removed theprobe holder and 0.5 g of capsicum cream (Midalgan,15 mg methylnicotinate and 50 mg glycolmonosalicylateper gram, Remark Groep BV, Meppel, the Netherlands)was applied over an area of 9 cm2 to the same measurementlocation and we measured variations in perfusion after5 minutes. Capsicum cream was applied to the samemeasurement location 10 minutes after the occlusionmeasurements. The measurements with the capsicumcream were performed 5 minutes after application ofthe cream. Capsicum cream increases thermoregulatingactivity of the skin by increasing the mean concentrationand mean velocity of red blood cells in few minutes on theareawherethecreamisapplied[17].Skinsiteswereavoidedwith visible large superficial blood vessels, hair, and pig-ment variations.

Inter- and Intra-Individual Inter-Site Variations inOptical Path Lengths and Path Length ResolvedDoppler Shifts

For measuring inter-individual within-site variations,measurements were performed on the dorsal side of theforearm of three human subjects (A, B, C), one of whichfemale (B). From visible appearance, the skin type of personA (aged 25) belongs to type II whereas skin type of person B(aged 30) and person C (aged 28) belong to type IV. To assessthe intra-individual inter-site variations, measurementswere performed at three positions, viz. fingertip, dorsal,and palmar side of forearm of the same individual.Before the measurements with our setup, perfusion read-ings for 1 minute were taken at the same locations with thePF5000.

EXPERIMENTS AND RESULTS

Optical Path Lengths and Path Length ResolvedDoppler Shift Measurements

Figure 2 shows the power spectra measured on the fore-arm skin for an optical path length of 1 mm in the tissue.The power spectrum contains interference peaks at themodulation frequency (6 kHz) and at its multiples (12 and18 kHz). The inset of Figure 2 gives an enlarged view on thespectrum around the modulation frequency. The broaden-ing of the peak resulting from the Doppler shift imparted bythe moving red blood cells to the multiply scattered photonsis reduced during the arterial occlusion of the upper arm.The area under the peaks, which represents the intensity ofdetected photons within a certain optical path length, doesnot change during occlusion. The DC value of the photo-current generated by the light detected from the samplewas the same (160 mV) before and during occlusion. Thisindicates that the blood volume does not change due toocclusion, but the mean velocity of moving scatterers isdecreased. Thus for a given optical path length, theDoppler shifted fraction of photons is reduced during occlu-sion. The optical path length distribution of photonsmeasured in skin before and during occlusion is shown inFigure 3. With a fiber distance of 300 mm, optical pathlengths spanning a range of 0–6 mm are measured. Thefraction of Doppler shifted photons averaged over the entireoptical path length measured from the respective areas ofthe optical path lengths are decreased from 28% to 18%during occlusion. When capsicum cream was applied tothe skin, both the intensity and the average Doppler shiftof scattered photons, represented by the area and the widthof the peak increased (Fig. 2). In this case, the DC photo-currentvaluegeneratedbythelightdetectedfromthetissueincreased to 280 mV. The increase in the DC value duringCapsicum cream provocation indicates that the amount ofbackscattered light into the spatially separated detectionfiber increases. When the capsicum cream is applied to the

Fig. 3. Optical path length distributions of Doppler shifted

and non-shifted photons measured before and during

occlusion.


skin, it increases the perfusion locally. This can be clearlyobserved especially in lightly pigmented skin types. Theincreased local concentration of red blood cells increasesthe intensity of backscattered photons through an overallincreasing in scattering, which is higher than the concom-itant increase in absorption. Even though an increase inblood volume leads to an increased absorption, we observethat the effect of scattering dominates over the effect ofabsorption at this wavelength. Further investigation isneeded to confirm this hypothesis and to understand thesignificant increase in the DC value during capsicum creamprovocation. At a different wavelength, the situation couldbe different. We have also analyzed different sets of data onocclusionreperfusiontestandonMidalgantests.DuringtheMidalgan test the blood content of the skin changes muchmore than during occlusion–reperfusion. In general, thevariation in the DC value during occlusion–reperfusion testis within 1%, whereas in the case of Midalgan test thevariation is more than 10%. We also observed that the DCvalue measured with PF5000 decreases when the blood ispressed away. In the case of Midalgan, the effect is reversedleading to an overall increase in DC value. The optical pathlength distributions for Doppler and non-Doppler shiftedlight after application of capsicum cream are shown inFigure 4. As expected, the Doppler-shifted fraction ofphotons (41%) increased after application of capsicumcream.

As shown in Figure 5, the weighted first moment M1/M0 ofthe Doppler shifted light, which represents the averageDoppler shift, increased with the optical path length dueto the greater probability of interaction of photons withmoving scatterers for large optical path lengths. Theobserved decrease inaverageDoppler shiftduring occlusionis due to the decrease in the mean velocity of RBCs. For agivenopticalpathlength,anincreaseintheaverageDopplershift is observed with capsicum cream provocation. Thiscould beexplainedby the increase in thevelocity of red blood

cells and also by the increase in successive Doppler shifts(multiple scattering) imparted to the photons by the richlyperfused skin, resulting from an increased concentration ofred blood cells.

Inter- and Intra-Individual Inter-Site Variations

Optical path length distributions and the correspondingpath length resolved Doppler shifts measured on the dorsalside of the forearm of three individuals are shown inFigures 6 and 7, respectively. We have done a repeatabilitystudy for subject C, leading to standard deviations of theresults inFigures6and7.Theopticalpathlengthofthepeakofthedistributionisshiftedandthedistributionisrelativelybroader for subject A, indicating that the scattered photonsappear to emanate from a slightly deeper region in the skin.The fraction of Doppler shifted photons, which providesinformation about the blood volume is lower (28%) for sub-ject A than for subject B (45%) and C (44%). The path lengthdependent average Doppler shift measured for subject Cis higher for short optical path lengths, whereas thatmeasured in subject B is consistently lower. But for largeoptical path lengths, the same amount of Doppler shift ismeasured for subject A and C.

Figure 8 depicts the path-length variations in the threemeasurement sites (finger, dorsal, and palmar side of theforearm) of one individual C. The distributions normalizedwith their respective maximum values are shown in theinset of Figure 8. Optical path lengths and path lengthresolved Doppler shifts measured on the dorsal and palmarside of the forearm do not show significant variations. For agiven optical path length, the average Doppler shiftmeasured on the fingertip is higher than that measuredin the forearm skin (Fig. 9). Optical path lengths measured

Fig. 4. Optical path length distributions of Doppler shifted

and non-shifted photons measured after capsicum cream

provocation.

Fig. 5. Weighted first moments, which are equal to the aver-

age Doppler shift as a function of optical path length, without

and with external stimuli: occlusion and capsicum cream

provocation.

696 VARGHESE ET AL.

on the fingertip are 40% longer as compared to forearmmeasurements. The Doppler shifted fraction of photonsmeasured on the fingertip is about 38% and that measuredon the dorsal and palmar sides of the forearm are 32% and17%, respectively.

In Table 1, the first moment of Doppler shifted photons,the Doppler shifted fraction of photons and the weightedaverage Doppler shift measured with our LCI setup arecompared with the perfusion readings recorded with thePF5000. The weighted average Doppler shift is calculatedusing the optical path length distribution of Doppler shifted

photons, using Equation (3). The readings recorded on thesame person and its variation to external stimuli are nor-malized with the control resting values for comparison. Thenormalized perfusion readings measured with Perimed andLCIisaddedinTable1(value inbrackets).Forthe intra-andinter-individual within-site variations measured, normal-izationdonothaveanyphysical significance.While compar-ing our results with the perfusion readings of the PF5000, ithas to be noted that the differences in the wavelength andnumerical aperture will affect the overall sampling depthand measurement volume. As expected, there is a decrease

Fig. 9. Intra-individual inter-site variations in path-length

resolved Doppler shifts measured on different sites (finger,

dorsal, and palmar side of the forearm) of the same

individual.

Fig. 7. Inter-individual within-site variations in path-length

resolved average Doppler shifts measured on comparable

sites (dorsal side of the forearm) between individuals.

Fig. 6. Inter-individual within-site variations in optical path

length distributions of multiply scattered light measured

on comparable sites (dorsal side of the forearm) of three

individuals.

Fig. 8. Intra-individual inter-site variations in optical path

length distributions measured on different sites (finger,

dorsal, and palmar side of the forearm) of the same individ-

ual. Inset: Optical path length distributions normalized with

their respective maximum values.


in the measured concentration and velocity during occlu-sion. Similar decreasing trends are shown qualitatively inthe Doppler shifted fraction of photons and the weightedfirst moments which is measured with our setup. Theincrease in the concentration and velocity shown inPF5000 readings by the capsicum cream activity is similarto the increase in theDoppler shifted fractionofphotonsandthe average Doppler shift measured with our setup. Theaverage Doppler shift measured with our LCI setup andthe velocity measured with PF5000 shows similar trendsin inter- and intra-individual measurements and alsoduring occlusion and capsicum cream provocation. Thevelocity of subject B measured with PF5000 is lower thanthat of subjects A and C, which is similar to the observationthat the weighted first moments measured with our setupis lower.

DISCUSSION

In this article we have presented path length resolvedDopplermeasurementsofmultiplyscattered lightfromskinand the results are compared with a conventional laserDoppler technique. In general, the influence of photon pathlengthsonthemeasuredperfusionsignal isovercomebythistechnique. Also, our method allows us to discriminatebetween the Doppler-shifted and non-shifted fraction ofphotons in the detected photodetector signal. With this pre-sented path length-resolved laser-Doppler flowmetry, fluxcan be measured independent of variations in optical pathlength. This is a novel approach to solve a major problemin laser Doppler flowmetry, where the perfusion signaldepends on the optical path length. To prove the ability ofour technique in yielding perfusion data independent fromthepropertiesof thestatic tissuematrix,wehaveperformedpath length-resolved dynamic light scattering measure-ments in various media having a constant concentration

of dynamic particles inside a static matrix with differentscattering properties and we compared the results with aconventional laser Doppler device (PF5000). We showedthat LCI enables measurements independent of the effectof the scattering properties of static matrices in which themoving particles are embedded, whereas PF5000 showsclear dependence [18]. Earlier we have proven independ-ence of path length resolved Doppler measurements fromoptical absorption [7,14]. Hence our method enables fluxmeasurements independent of the optical properties ofthe media in which the particles are embedded. This is animportant step towards absolute measurements. Ina recentstudy, we demonstrated the feasibility of our method forreal-time monitoring of skin perfusion with path lengthsensitivity[19].Real-timemonitoringoftheDoppler-shiftedfraction of photons and the weighted first momentsmeasured with our setup for an optical path length of1.7 mm showed correlation with the perfusion signalmeasured using a conventional LDPM.

The direct comparison of our results with the perfusionsignalsofthePF5000isdifficult.Inourheterodynedetectiontechnique, the power spectrum of photocurrent fluctuationsresults from the interference between Doppler shiftedlight and unshifted reference light, whereas conventionallaser Doppler techniques are based on mutual interferencebetween all components of the light detected from the tissueonly, either Doppler shifted or unshifted, hence withoutreference beam. This leads to an extra mechanism of broad-ening of the power spectrum, depending on the fraction ofDoppler shifted photons. Furthermore, due to this mutualinterference in LDF the power in the fluctuations saturatesfor large fractions of Doppler shifted light. Secondly, bothtechniques sample different parts of the vascular bed. Thisdifference can mainly be attributed to the longer coherencelength (5 mm) of the light source used in the PF5000,measuring the perfusion averaged over all path lengths

TABLE 1. Comparison of Perfusion Readings Measured With LCI Setup and PF5000

Perimed (PF5000) LCI

Flux

(a.u.)

Concentration

(a.u.)

Velocity

(a.u.) M1

Doppler

fraction of

photons (%)

Average

weighted

Doppler shift

(Hz)

Norm. Norm. Norm. Norm. Norm. Norm.

Normal (A) 28 1 45 1 61 1 1792 1 28 1 151 1

Occlusion 5 0.18 8 0.18 43 0.7 1026 0.57 18 0.64 122 0.81

(A) 90 3.21 68 1.51 234 3.84 3034 1.69 41 1.46 210 1.39

Capsicum cream (A)

A 28 45 61 1792 28 151

B 4 12 32 1980 45 112

C 57 31 74 3168 44 179Finger (C) 230 109 224 4066 38 211

Forearm dorsal (C) 12 7.4 155 1472 32 122

Forearm palmar (C) 10 9.7 120 595 17 109

The readings recorded on the same person and its variation to external stimuli are normalized with the control resting values.

698 VARGHESE ET AL.

compared to the path length resolved perfusion measure-mentsthatareperformedwithourlowcoherentlightsource.We have provided Doppler shifts averaged over all opticalpath lengths to allow for comparison of our results with thatfromPF5000.InthepracticaluseofLCI,perfusionmeasure-ments at one or a few well-selected optical path lengths arepreferred. The preferred optical path lengths at which per-fusion can be measured are at 1 and 2 mm. At these opticalpath lengths, the intensity is nearly maximal and also ena-bles to discriminate between superficial and deeper layers.The optical path length at which the intensity is maximal(�1.7 mm) is not preferred since we observe that averagingover long time is required at this optical path length. Whenperfusion measured over all optical path lengths is aver-aged, both techniques are expected to show similar trends.However the measurements with two techniques will alsodiffer due to the dependence on the wavelength, numericalaperture of the detection fibers, fiber optic probe andsignal processing, but probably only to a small extent.Furthermore, even when the LCI probe and PF5000 probeare subsequently inserted in the same probe holder, differ-ences intheprobedvolumewillalsooccurduetothefact thatboth are not probing the same location exactly. All thesefactors have to be taken into account while comparing thesetwotechniques.Whenthemeasurementsarerepeatedafewtimes with the same probe by taking it out and repeating itagain, keeping all other factors same, the perfusion signalwill be different. Furthermore, since relatively longer timesare required for measuring the complete optical path lengthdistributions, the recordings are affected by the temporalvariations in the physiological circumstances of themeasurement site.

The effects of path length affecting the response to prov-ocation are clear from Figure 4, where the variation result-ing by the thermoregulating activity of Midalgan indicatesthat this effect is superficial. However, with PF5000, thesignal recorded does not show any discrimination betweenthe variations occurring at superficial and deeper layers oftissue.Moreover,duringarteryocclusion,thebiologicalzeromeasured (offset between the calibrated instrumental zeroof the instrument and the measured signal) was high forshort optical path lengths. This implies that the effect ofocclusion is more significant in deeper layers of tissue.

Optical Path Lengths and Path Length ResolvedDoppler Shifts of Multiply Scattered Photonsin Skin

Compared to the estimation of the Doppler-shifted frac-tion of photons based on the first-order [20] or second-orderstatistics [21] of the fluctuating speckle pattern, the pre-sented approach provided path length resolved informationabout the probed region. Using the speckle approach, thefraction of Doppler-shifted photons estimated in the lightbackscattered from the human skin of forearm was 15%(fiber distance ¼ 250 mm, fiber core diameter ¼ 125 mm),similar to our observation that most of the scattered lightin the perfusion signal is scattered from the surroundingstatic tissue matrices. However, their path length averagedapproach does not provide information about the probed

region and this will lead to uncertainties in interpretingthe physiological information, when measured on tissuewith different extent of perfusion at different depths.

Inter- and Intra-Individual Inter-Site Variations

The optical path lengths measured with fiber distancestypical for laser Doppler perfusion monitors (LDPM)showed inter- and intra-individual variations, which is con-tradictory to the observations made by Bonner and Nossalthat the path-length variation is not a large factor in typicalLDPMs [1]. This clearly shows that these variations shouldbe taken into account while comparing the perfusion read-ings from comparable sites between individuals and fromdifferentsitesofthesameindividual.ThemajordrawbackofLDF resulting from significant variations in optical pathlengths has been pointed out by Tenland et al. [22] andBraverman et al. [23]. Larsson et al. observed that theaverage path length varied up to �40% (max/min) betweenindividualsandtheaverageintra-sitevariabilityislessthanthe inter-site variability [20]. Methods for the local photonpath length estimation based on the diffuse reflectanceapproach can eliminate this influence, but their resultsshowed non-linearities for short optical path lengths dueto the fact that they considered the average photon pathlength and not the actual path length distribution [24]. Inour coherence-gated approach, we estimate the actual pathlength distribution rather than the average photon pathlength. The preliminary results presented in this manu-script indicate that the path length distributions dependson the skin site that is being probed and the intra-individualinter-site variability is higher than the inter-individualwithin-site variability measured on comparable sitesbetween different individuals. In general, fingertips havelower absorption and reduced scattering coefficients thanthe forearm skin resulting in an increase of optical pathlengths [24]. These increased optical path lengths thatthe photons travel in the finger are evident from the nor-malized distributions measured with our setup (inset ofFig. 8). Assuming a homogenous tissue perfusion, the con-ventional LDPM readings are thus expected to show arelative overestimation on fingertips compared to the fore-arm readings due to the linear dependence of perfusion onaverage photon path length [24]. This is shown in the fluxreadings obtained with the PF5000, which showed anincrease (factor of 23) in the flux recorded on finger ascompared to the forearm recordings. With LCI, this factoris reduced to 7. Since small source-detector separations of0.3 mm result in small sampling volume and superficialsampling depth, the perfusion readings will be sensitiveto the variations in the microvascular architecture.Previous theoretical investigation on the degree of signallocalization to the different layers of skin predicts that aprobe spacing of 250 mm samples primarily epidermallayers and papillary dermis, whereas spacings of 400–800 mm sample upper blood net dermis and dermis [25].We expect that the signals measured for short and largeoptical path lengths originate from superficial and deeperlayers of tissue, respectively. However, the results pre-sented in this manuscript are measured with path length


increments of 200 mm and thus there could be some overlapbetween measured signals from different layers. Theinterpretation of the velocity profile and the distinctionbetween capillaries and arterioles and venules from thesein vivo measurements are not possible with this method.However, our method suppresses the effect of inhomo-geneous tissue optical properties. The remaining perfusionvariations are only due to the inhomogeneities in theperfusion itself.

CONCLUSION

In this study, we demonstrated, for the first time to ourknowledge, in vivo path length resolved Doppler measure-ments of diffusely scattered light from skin under externalstimuli such as occlusion and capsicum cream provocation.The Doppler shifted fraction of photons, which is related tothe blood volume and the path length dependent averageDoppler shift, which is related to the mean velocities of redblood cells, are decreased during occlusion. Both quantitiesare significantly increased when capsicum cream wasapplied. Furthermore, the observed inter- and intra-indi-vidual inter-site variations in path length resolved Dopplermeasurements indicate the inherent limitation of conven-tional LDPM that restrict its clinical usefulness, due toits dependence on the unknown photon path length.Consequently, this method will enable to correctly interpretorcounter-act the inter-andintra-individual inter-sitevari-ations in the LDF readings introduced by the variance intissue optical properties. This approach also enables to dis-criminate between the Doppler-shifted photons resultingfrominteractionwiththemovingredbloodcellsandthenon-shifted light scattered only by the surrounding static tissuematrices.Finally, thisapproachallowsthedeterminationoftunable depth resolved perfusion information, which ena-bles the discrimination between the perfusion signal fromsuperficial and deeper layers of tissue. However, furtherdevelopments and fundamental research are required indeveloping this into a tool that is suitable for use in a clinicalenvironment, with acceptable measurement times andsuitable patient interfaces.

ACKNOWLEDGMENTS

This work was sponsored by the Netherlands TechnologyFoundation STW (grant TTF 5840) and the Institute forBiomedical Technology of the University of Twente.Perimed AB (Stockholm, Sweden) is acknowledged for pro-viding instrumentation.

REFERENCES

1. Shepherd AP, Oberg PA. Laser-Doppler Blood Flowmetry.Boston: Kluwer Academic; 1990.

2. Morales F, Graaff R, Smit AJ, Gush R, Rakhorst G. The influ-enceofprobefiberdistanceonlaserDopplerperfusionmonitor-ing measurements. Microcirculation 2003;10(5):433–441.

3. Larsson M, Steenbergen W, Stromberg T. Influence of opticalproperties and fibre separation on laser Doppler flowmetry.J Biomed Opt 2001;7:236–243.

4. Obeid AN, Boggett DM, Barnett NJ, Dougherty G, Rolfe P.Depth discrimination in laser Doppler skin blood flow

measurement using different lasers. Med Biol Eng Comput1988;26:415–419.

5. LiebertA,LeahyM,ManiewskiR.Multichannel laser-Dopplerprobe for blood perfusion measurements with depth discrimi-nation. Med Biol Eng Comput 1998;36:740–747.

6. DoughertyG.AlaserDopplerflowmeterusingavariablecoher-ence length to effect depth discrimination. Rev Sci Instrum1992;63:3220–3221.

7. Varghese B, Rajan V, Van Leeuwen TG, Steenbergen W. Pathlengthresolvedmeasurementsofmultiplescatteredphotonsinstatic and dynamic turbid media using phase modulated lowcoherence interferometry. J Biomed Opt 2007;12:024020.

8. Varghese B, Rajan V, Van Leeuwen TG, Steenbergen W.Quantification of optical Doppler broadening and optical pathlengths of multiply scattered light by phase modulated lowcoherence interferometry. Opt Expr 2007;15(15):9157–9165.

9. Varghese B, Rajan V, Van Leeuwen TG, Steenbergen W. Highangle phase modulated low coherence interferometry for pathlength resolved Doppler measurements of multiply scatteredlight. Opt Commun 2008;281(3):494–498.

10. Huang D, Swanson EA, Lin CP, Schuman JS, Stinson WG,Chang W, Hee MR, Flotte T, Gregory K, Puliafito CA,Fujimoto JG. Optical coherence tomography. Science 1991;254:1178.

11. Wax A, Yang C, Dasari RR, Feld MS. Path-length-resolveddynamic light scattering: Modeling the transition from singleto diffusive scattering. Appl Opt 2001;40:4222–4227.

12. Bizheva KK, Siegel AM, Boas DA. Path-length-resolveddynamic light scattering in highly scattering random media:The transition to diffusing wave spectroscopy. Phys Rev E1998;58:7664–7667.

13. Petoukhova AL, Steenbergen W, de Mul FFM. Path-lengthdistribution and path-length resolved Doppler measurementsof multiply scattered photons by use of low-coherence interfer-ometer. Opt Lett 2001;26:1492–1494.

14. Petoukhova AL, Steenbergen W, van Leeuwen TG, de MulFFM. Effects of absorption on coherence domain path lengthresolved dynamic light scattering in the diffuse regime. ApplPhys Lett 2002;81:595–597.

15. Nilsson GE. Signal processor for laser Doppler tissue flow-meters. Med Biol Eng Comput 1984;22:343–348.

16. Fitzpatrick TB. The validity and practicality of sun-reactiveskin types I through VI. Arch Dermatol 1988;124(6):869–871.

17. Serov AN, Nieland J, Oosterbaan S, de Mul FFM, vanKranenburg H, Bekman HH, Steenbergen W. Integrated opto-electronic probe including a vertical cavity surface emittinglaser for laser Doppler perfusion monitoring. IEEE TransBiomed Eng 2006;53(10):2067–2074.

18. Varghese B, Rajan V, Van Leeuwen TG, Steenbergen W.Measurementofparticlefluxinastaticmatrixwithsuppressedinfluence of optical properties, using low coherence interfer-ometry. Opt Expr 2010;18:2849–2857.

19. Varghese B, Rajan V, Van Leeuwen TG, Steenbergen W.Path-length-resolved optical Doppler perfusion monitoring.J Biomed Opt 2007;12:060508.

20. Briers JD. The statistics of fluctuating speckle patterns pro-duced by a mixture of moving and stationary scatterers. OptQuantum Electron 1978;10:364–366.

21. Serov A, Steenbergen W, de Mul F. Prediction of the photo-detector signal generated by Doppler-induced speckle fluctu-ations: Theory and some validations. J Opt Soc Am A 2001;18:622–630.

22. Tenland T, Salerud EG, Nilsson GE, Oberg PA. Spatial andtemporal variations in human skin blood flow. Int J MicrocircClin Exp 1983;2:81–90.

23. Braverman IM, Keh A, Goldminz D. Correlation of laserDoppler wave patterns with underlying microvascularanatomy. J Invest Dermatol 1990;95:283–286.

24. Larsson M, Nilsson H, Stromberg T. In vivo determination oflocal skin optical properties and photon path length by use ofspatially resolved diffuse reflectance with applications in laserDoppler flowmetry. Appl Opt 2003;42:124–134.

25. Meglinsky IV, Matcher SJ. Modelling the sampling volume forskin blood oxygenation measurements. Med Biol Eng Comput2001;39.

700 VARGHESE ET AL.

Documents

In vivo optical path lengths and path length resolved doppler shifts of multiply scattered light