Erc

NSa

b

c

a

ARRA

KMIDOB

1

ia[

1h

Medical Engineering & Physics 35 (2013) 1551– 1557

Contents lists available at ScienceDirect

Medical Engineering & Physics

jou rn al h om epage: www.elsev ier .com/ locate /medengphy

valuating and improving the quality of time-dependent, diffuseeflectance spectroscopic signals measured from in vivo brain duringraniotomy

itin Yadava,∗, Sanjiv Bhatiab, John Raghebb, Yinchen Songa, Adrian Romeroa,anghoon Ohc, Wei-Chiang Lina,b

Department of Biomedical Engineering, Florida International University, Miami, FL, USAThe Brain Institute, Miami Children’s Hospital, Miami, FL, USAMemorial Sloan-Kettering Cancer Center, New York, NY, USA

r t i c l e i n f o

rticle history:eceived 17 April 2012eceived in revised form 5 April 2013ccepted 17 April 2013

eywords:otion artifacts

n vivo brainiffuse reflectanceptical spectroscopyiomedical optics

a b s t r a c t

Background: Optical spectroscopy can be used to assess the pathophysiological characteristics of diseasedand injured biological tissue in vivo in a non-destructive way. It is often used in conjunction with a contactoptical probe for the purposes of operating and sensing in a sterile field. Since the probe is often held bythe hand of an investigator during data acquisition, any hand instability can affect the quality of acquireddata and, hence, degrade the accuracy of diagnosis. This study was designed to quantitatively characterizethese artifacts, and then propose an effective engineering solution to remove them.Methods: Time-dependent diffuse reflectance spectra (Rd(�,t)) were acquired from the normal cortexregion of pediatric patients undergoing epilepsy surgery. They were acquired at a rate of 33 Hz, andtheir range was 400 and 900 nm. Two distinct ways of collecting data were tested: one with the fiberoptical probe held by the surgeon’s hand during data acquisition, and the other with the probe heldby a specially designed probe holder. The probe holder was designed and constructed to minimize thevariations in probe contact pressure and contact point for the full duration of any given investigation.Spectral data acquired using versus not using the probe holder were characterized and compared in thetime, wavelength, and frequency domains, using both descriptive and inferential statistics.Results: Hand motion manifested as strong random variations in Rd(�,t) which impacted temporal andfrequency characteristics of Rd(�,t). The percentage standard deviation %STD of Rd(�,t) acquired withoutprobe holder could be as high as 60%, and they are significantly higher than those with probe holder at allwavelengths. This difference is especially prominent between 400 and 600 nm. Rd(�,t) acquired withoutthe probe holder also processed a higher spectral power energy in the frequency domain than thosewith the probe holder. The correlation analysis revealed that the hand motions induced synchronistic

variations in Rd(�,t) between 600 and 800 nm, but this synchronicity is not obvious between 400 and600 nm.Conclusion: The results of this investigation demonstrate the nature and the magnitude of hand motioninduced artifacts in in vivo diffuse reflectance spectra and propose one potential solution (i.e., a probeholder) to remove them. These findings allow us to improve the quality of time-dependent, diffusereflectance signals acquired to study the dynamic characteristics of biological tissues, like brain, in vivo.

. Introduction

Spectroscopy-based optical diagnostic technologies have been

nvestigated and evaluated widely over the past two decades; theyre primarily used to detect pre-cancer in various tissue organs1–6] and guide tumor resection intraoperatively [7–10]. These

∗ Corresponding author. Tel.: +1 7863031767.E-mail address: nyada001@fiu.edu (N. Yadav).

350-4533/$ – see front matter © 2013 IPEM. Published by Elsevier Ltd. All rights reservettp://dx.doi.org/10.1016/j.medengphy.2013.04.006

© 2013 IPEM. Published by Elsevier Ltd. All rights reserved.

technologies are built upon the principle of light-tissue interac-tions: light absorption and scattering inside a biological tissue aregoverned by the tissue’s structural and compositional characteris-tics and these intrinsic characteristics are altered by disease and/orinjury [4,11–15]. These technologies hold several advantages overexisting diagnostic technologies, in that they are relatively inex-

pensive, highly portable, and can be used real-time. To make thesetechnologies applicable within an in vivo or intraoperative envi-ronment, a sterilized contact probe often is employed to achieveremote sensing from in vivo tissue. The probe usually is held by

d.

1552 N. Yadav et al. / Medical Engineering & Physics 35 (2013) 1551– 1557

F 1 in. (a

ttmttoescsilori[ostbnp

btmwsmfishamawtii

ig. 1. (a) Schematic of the mechanical holder for the fiber optic probe. Here h = w =cquisition.

he hand of an operator. One common problem with such a prac-ice is motion artifacts in the acquired data; unintentional hand

ovements or tremors alter the pressure of the probe against thearget tissue (i.e., probe contact pressure) or shift the site of inves-igation. Consequently, these movements induce additional noiser artifacts in the in vivo spectra acquired. This phenomenon isspecially pronounced in studies involving time dependent mea-urements. A couple of earlier reports have suggested that probeontact pressure does not significantly affect fluorescence inten-ity of the cervix [17,18]. A similar finding was noted in a studyn which Raman spectroscopy was used to detect pre-cancerousesions within the gastrointestinal tract [19]. However, severalther studies have shown that excessive probe contact pressure,esulting from hand movements, can lead to strong alterationsn the hemodynamic and metabolic characteristics of local tissue20–26]. In addition, the effect of varying probe contact pressuren acquired optical spectral data is tissue type dependent: theofter the tissue, the greater the effect [16]. Despite this recogni-ion, the intrinsic characteristics of the spectral artifacts inducedy hand motions have not yet been thoroughly investigated, ando systematic approach to combat this obstacle has been pro-osed.

In this paper, we report the results of an in vivo study of therains of pediatric patients undergoing epilepsy surgery, so aso quantify the spectral and temporal artifacts induced by hand

otions. Specifically, time-dependent diffuse reflectance spectraere acquired from the normal cortex of patients during epilepsy

urgery using a fiber-optic spectroscopy system. Two distinctethods of spectral data acquisition were compared: one with the

ber-optic probe held by the surgeon’s hand during data acqui-ition, and the other with the probe held by a mechanical probeolder. The probe holder was devised to eliminate hand-motionrtifacts in the recording and to counteract any spontaneousovements of the in vivo brain. Recordings from the two data-

cquisition methods were compared in the time, frequency, and

avelength domains, using both descriptive and inferential statis-

ics. From these results, the characteristics of spectral artifactsnduced by hand movements were determined and character-zed.

b) The probe holder is attached to a Greenberg retractor system during in vivo data

2. Materials and methods

2.1. Instrumentation

The instrumentation for in vivo diffuse reflectance spectralacquisition consisted of a fiber-optical probe and a portable spec-troscopic system. It was calibrated using a calibrated tungstenhalogen light source (LS-1-CAL, Ocean Optics, Dunedin, FL) in orderto remove any spectral alterations induced by the instrument.Details of this system can be found in a previous publication byour group [7].

2.2. Design of the probe holder

A mechanical probe holder was designed and engineered toeliminate the need for a surgeon to hold the optical probe through-out a single data acquisition procedure. The introduction of thisholder enabled us to identify and, hence, quantify the spectralalterations induced by hand movements. The three critical designcriteria of the probe holder were: (1) it had to be able to with-stand the ethylene oxide gas sterilization procedure; (2) it shouldlimit the lateral movement of the probe and hence maintain aconstant site of investigation; and (3) it should maintain constantprobe-contact pressure, if feasible, against the in vivo brain tis-sue. The final design of the probe holder is depicted in Fig. 1. Tomeet the sterilization criterion, medical grade Stainless Steel 361was used to construct the entire holder. To meet the stability cri-terion, the probe holder was designed to be used in conjunctionwith the Greenberg self-retaining retractor system. This approachreduced the susceptibility of the probe holder to movements of thepatient during data acquisition, as the Greenberg system [26] wasattached directly to the Mayfield head clamp (Integra LifeSciencesCorp., Cincinnati, OH) which was attached to the surgical bed. Thecenter of the probe holder was a stainless-steel tube with an innerdiameter slightly larger than the diameter of the optical probe. The

optical probe was inserted through the tube and secured using aside thumbscrew. To enable movement of the optical probe alongits primary axis, the center tube was attached to the frame of theholder via a two-bar track system (Fig. 1). The range of the axial

N. Yadav et al. / Medical Engineering &

Table 1The effects of the damping springs on the probe contact force. The calculations wereconducted using the shear modulus of the brain = 5k Pa. Here k is the total springconstant of the probe holder, and k = 0 represents the probe holder without dampingmechanism.

Brain movement (mm) Probe contact force (N)

k = 0 k = 245 N/m k = 50 N/m

0.1 0.00635 0.00508 0.0019050.5 0.03175 0.0254 0.009525

mtfNbtshdlarsvcct

2

RfAaHcto((natttlbigluwrcs

abttps

explored. The power spectral energy of FTabs �i(f), P(�i), was calcu-lated by integrating [FTabs �i(f)]2 over the entire frequency range,

1 0.0635 0.0508 0.01905Reduction (%) 20% 80%

ovement of the center tube was limited to ±3 mm from its res-ing position. To damp the axial movement of the optical probe,our springs (CI-008B-08-S316, Lee Spring Company, Greensboro,C) were used in the track system to suspend the probe. The com-ined spring constant of this suspension system was 250 N/m. Sincehe free weight of the probe is about 26 g, the displacement of theprings was about 1 mm when the probe was loaded to the probeolder. Using the flat punch indentation model of viscoelastic solidserived by Harding and Sneddon [27] as well as the shear modu-

us of the brain reported by Green et al. [28], it was found that theddition of the damping springs in the probe holder would alsoeduce the variation in probe contact pressure, resulting from thepontaneous brain movements, by about 20% (Table 1). This type ofariation could be further minimized by reducing the overall springonstant. The trade-off of such an approach, however, is a signifi-ant increase in the loading displacement of the springs, and hencehe size of the probe holder.

.3. Data acquisition and preprocessing

In this study, in vivo time-dependent diffuse reflectance spectrad(�,t) were acquired from the in vivo brains in the operating roomrom pediatric patients who were undergoing epilepsy surgery.

total of 18 patients were enrolled in the study. The study waspproved by the Institutional Review Boards at Miami Children’sospital and Florida International University. Informed writtenonsent was obtained from the parents or guardians of each par-icipant. The 18 patients were divided evenly into two groups. Inne group, data acquisition was performed with the probe holderHolder Group). In the other group, the probe holder was not usedHand Group). In this trial, spectral data were acquired only fromormal cortex, as defined by intraoperative electrocorticographynd pre-operative imaging. Prior to data acquisition, each inves-igated site was rinsed with saline to remove surface blood, andhe surgeon gently placed the optical probe in direct contact withhe target tissue. Prior to spectral acquisition, the broadband whiteight source was turned on and pre-heated for at least 30 s to sta-ilize its output. Sequentially, 400 Rd(�,t) were acquired from the

nvestigated site at a rate of 30 spectra per second (i.e., 30 ms inte-ration time). Upon completion of spectral acquisition, the whiteight was turned off and a baseline spectrum was acquired to doc-ment the influence of ambient light. Only one spectral data setas acquired from each of the 16 patients, but two sets from the

emaining two (one in each group). Every stage of the study proto-ol was carried out under the careful supervision of the surgeons,uch that patient care was not compromised.

A standard data pre-processing routine was applied to allcquired Rd(�,t), so as to eliminate spectral variations inducedy the instrumentation and ambient light [7]. During the spec-

ral pre-processing procedure, the range of Rd(�,t) was reducedo 400–850 nm with a wavelength interval of 5 nm. These pre-rocessing procedures yielded a set of calibrated diffuse reflectancepectra Rdcal(�i,t).

Physics 35 (2013) 1551– 1557 1553

2.4. Data analysis

Rdcal(�i,t) was analyzed in time and frequency domains to iden-tify the characteristics of spectral alterations induced by handmovements. Furthermore, the degree of improvement in the qual-ity of Rdcal(�i,t) attributable to the probe holder was assessedquantitatively.

To characterize the temporal alterations induced by hand move-ments, Rdcal(�i,t) from the Holder and the Hand groups werecompared empirically by an experienced spectral analyzer (WCL).Time variations in Rdcal(�i,t) at a given wavelength were quantifiedfurther using its percentage standard deviation, which is calculatedas

%STD Rdcal(�i) = STD(Rdcal(�i, t))MEAN(Rdcal(�i, t))

, (1)

where STD is the standard deviation calculation along the time axis,MEAN is the overall average calculation along the time axis, and�i = 400–850 nm at 50 nm intervals. Finally, statistical comparisonswere performed of %STD Rdcal(�i) from the Holder and Hand groupsto estimate the degree of improvement which can be attributed tothe probe holder on the quality of acquired data.

To investigate wavelength-dependency of the temporal spec-tral alterations in Rdcal(�i,t) induced by hand movements, temporalcorrelation analysis of Rdcal(�i,t) was conducted. Here, two signalswithin a single recording set, Rdcal(�i,t) and Rdcal(�ref,t), were usedto calculate correlation coefficients r(�i,�ref,t) for a fixed time win-dow (�t), expressed as

r(�i, �ref , t) = COV [Rd(�i, t : t + �t), Rd(�ref , t : t + �t)]i

STD[Rd(�i, t : t + �t)] ∗ STD[Rd(�ref , t : t + �t)](2)

where COV is the covariance calculation. In this analysis,�ref = 800 nm was used as it is in the spectral region wherehemoglobin absorption is not prominent, and the window sizeused was 25 samples (i.e., �t = 750 ms). Again, r(�i,�ref,t) fromthe Holder and Hand groups were empirically evaluated to iden-tify wavelength-dependency of spectral alterations in Rdcal(�i,t)induced by hand movements. Furthermore, the mean of r(�i,�ref,t)over the entire recording duration, denoted as MEAN(r(�i,�ref,t)),was calculated. Here, the range of �i was 400–850 nm at 50 nmintervals. Finally, statistical comparisons were performed ofMEAN(r(�i,�ref,t)) from the Holder and Hand groups to estimate theextent of improvement which can be attributed to the probe holderon the quality of acquired data.

The frequency composition of Rdcal(�i,t) was analyzed to deter-mine the frequency characteristics of the spectral alterationsinduced by hand motions. To eliminate the contribution of theDC component to the analysis, the mean intensity over the entirerecording period was removed from Rdcal(�i,t). That is

Rdcal ac(�i) = Rdcal(�i, t) − MEAN(Rdcal(�i, t) (3)

Fourier transformation then was applied to Rdcal ac(�i), whichproduced a frequency spectrum FT �i(f). The absolute value ofFT �i(f) was computed to yield FTabs �i(f), which shows the mag-nitudes of the frequency components. From FTabs �i(f), locationsof the prominent frequency components were identified and theircorrelations with the physiological characteristics of the brain were

for �i = 400–850 nm in 50 nm intervals. Finally, statistical compar-isons were performed with P(�i) from the Holder and Hand groupsto estimate improvement in the quality of the acquired data causedby the probe holder.

1554 N. Yadav et al. / Medical Engineering & Physics 35 (2013) 1551– 1557

F especg h10 f

3

3

ictnfggaK%ahis

TR

igure 2. (a and b) The representative Rdcal(�,t) from the Holder and Hand groups, rroups, respectively. p1 to p10 represent data sets from the Holder group and h1 to

. Results

.1. Temporal variations in Rdcal(�i,t)

Temporal variations in Rdcal(�,t) from the Holder group werensignificant (Fig. 2a) at all wavelengths as shown in Table 2. Inontrast, strong variations often were observed in Rdcal(�,t) fromhe Hand group. These variations did not possess any predictableature (e.g., cycling) and often occurred at all wavelengths. Results

rom the quantitative analysis of Rdcal(�,t) in the time domain sug-est that time variations in Rdcal(�,t) from the Holder group, ineneral, were much less than those from the Hand group (Fig. 3cnd d). Two sample t-tests (for normally distributed data) and theolmogorov–Smirnov test (for non-normally distributed data) ofSTD Rdcal(�i) further confirmed that the decrease in temporal vari-

tions in Rdcal(�,t), resulting from the employment of the probeolder, was statistically significant (Table 2). In the Hand group,

t also is noted that %STD Rdcal(�i) between 400 and 600 nm wastronger than that beyond 600 nm, which suggests that Rdcal(�,t)

able 2esults of statistical comparisons (p values).

Domain Wavelength (nm)

400 450 500 550 6

%STD Rdcal(�i) 0.003* 0.001 0.001 0.001 0MEAN(r(�i ,�ref ,t)) 0.000 0.001* 0.000 0.001* 0P(�i) 0.000 0.000 0.000 0.000 0

* A non-parametric test (Kolmogorov–Smirnov) was used.

tively. (c and d) %STD Rdcal(�i) calculated using Rdcal(�,t) from the Holder and Handrom the Hand group.

between 400 and 600 nm was more susceptible to hand move-ments.

3.2. Temporal correlation in Rd(�,t)

The correlation coefficient maps, r(�i,�ref,t), from the Holdergroup were featureless, as shown in Fig. 3a; r(�i,�ref,t) fluctuatedrandomly between −0.5 and 0.5 along the time axis as well asthe wavelength axis. In contrast, r(�i,�ref,t) from the Hand groupoften possessed vertical bands, where high temporal correlations(|r| > 0.8) were found among the majority of wavelength bands.The temporal durations of these vertical bands varied. Moreover,this vertical band may be divided into two segments along thewavelength axis; r(�i,�ref,t) above 600 nm was greater than 0.8and r(�i,�ref,t) below 600 nm was less than −0.8. Because of these

vertical bands, MEAN(r(�i,�ref,t)) from the Hand group was gen-erally higher than that from the Holder group. This difference,according to a two sample t-test and Kolmogorov–Smirnov test(wherever applicable), was statistically significant (Table 2). Also,

00 650 700 750 800 850

.001 0.001 0.000* 0.001 0.000 0.001*

.000 0.000 0.000 0.003* – 0.003*

.000 0.000 0.000 0.000 0.000 0.000

N. Yadav et al. / Medical Engineering & Physics 35 (2013) 1551– 1557 1555

Fig. 3. (a and b) The representative correlation coefficient maps, r(�i ,�ref ,t), from the Holder and Hand groups, respectively. (c and d) MEAN(r(�i ,�ref ,t)) calculated usingRdcal(�,t) from the Holder and Hand groups, respectively. p1 to p10 represent data sets from the Holder group and h1 to h10 from the Hand group.

Fig. 4. (a and b) The representative FTabs �i(f) from the Holder and Hand groups, respectively. The green dots in (a) indicate the peaks associated with the breathing and heartrates. (c and d) P(�i) calculated using FTabs �i(f) from the Holder and Hand groups, respectively. p1 to p10 represent data sets from the holder group and h1 to h10 from thehand group. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)

1 ering &

ido

3

w(sg1jHiwhg(

4

tavtnsdmtiaactin

yrtTcatSfiFfcdcmaiwtcaii

m

556 N. Yadav et al. / Medical Engine

t is noted that the difference in MEAN(r(�i,�ref,t)) between the twoata groups was more pronounced between 400 and 600 nm thanver the remaining wavelengths.

.3. Frequency composition

Low temporal variations in Rdcal(�,t) from the Holder groupere clearly reflected in their corresponding frequency spectra

Fig. 4a); they were very similar to the frequency spectrum ofpectrometer noise (no light). However, FTabs �i(f) from the Holderroup often processed a prominent frequency component between

and 2 Hz, which was related to the heart rate of the studied sub-ect. As expected, strong temporal variations in Rdcal(�,t) from theand group led to more frequency components in their correspond-

ng FTabs �i(f). These random frequency components, however,ere not related to physiological characteristics of the brain (i.e.,eart rate and breathing rate). Statistically, P(�i) from the Holderroup was significantly smaller than P(�i) between 400 and 850 nmTable 2).

. Discussion and conclusions

In this study, the effects of hand motions on in vivo optical inves-igation were evaluated qualitatively and quantitatively in timend frequency domains. Hand motion induced significantly randomariations in acquired diffuse reflectance spectra, and its effects inhe short wavelength region (i.e., below 600 nm) are more promi-ent. These random variations in the time domain translate to aignificant increase in the power spectral energy in the frequencyomain. These variations can be strong enough to overwhelm theeaningful signals generated by the in vivo physiological charac-

eristics of the investigated tissue (e.g., cardiac cycle). Even morenteresting is that the temporal correlation of these spectral alter-tions exhibits peculiar wavelength dependencies: their phasest different wavelength bands may be completely opposite underertain circumstances. The introduction of the probe holder men-ioned in this paper significantly reduces these temporal variationsn the acquired diffuse reflectance spectra, while preserving theative signals from in vivo brain.

Through the frequency domain and temporal correlation anal-ses of Rdcal(�,t), it was found that oscillations in the diffuseeflectance spectra, acquired using the probe holder, coincide withhe rates of the beating heart (1–2 Hz) and breathing (0.2–0.5 Hz).his indicates that heart beats and breaths alter local hemodynamicharacteristics, namely blood volume and hemoglobin oxygen-tion. These changes, in turn, modify the absorption properties ofhe cortex and, hence, the measured diffused reflectance signals.pectral alterations induced by probe movement can be classi-ed into two categories, based upon the nature of the movement.or lateral probe movements, the spectral alterations originaterom spatial variations in the optical properties of the cerebralortex. This commonly appears as synchronized oscillations iniffuse reflectance signals at all wavelengths, and explains the high-orrelation-coefficient vertical bands in the correlation coefficientaps r(�i,�ref,t). For axial probe movement, the spectral alterations

re induced primarily by the changes in the regional hemodynam-cs [16]. The changes in hemoglobin oxygenation would lead to

avelength dependent increase/decrease in Rdcal(�,t), and hencehe band characteristics in its corresponding correlation coeffi-ient maps r(�i,�ref,t). The deformation of the cortex induced byxial probe movement, however, should not change its scatter-

ng properties because the biological tissue has been treated as anncompressible material in studies of tissue mechanical properties.

The potential to utilize signal processing techniques to removeotion artifacts from recorded signals like Rd(�,t) has been

Physics 35 (2013) 1551– 1557

explored by various groups, including ours. Because of the nonlin-ear and unpredictable nature of the motion artifacts, conventionalfrequency domain filtering would not provide satisfactory results.Consequently, nonlinear filters – such as adaptive filtering, waveletfiltering, and independent component analysis – must be used.Based on our experience thus far, none of the existing filteringtechniques effectively removes the motion artifacts in Rd(�,t). Thisis because of the nonlinear effects of motion artifacts on diffusereflectance signals at various wavelengths. Therefore, we believethat it is better to reduce the effects of motion artifacts during therecording stage (using a mechanical holder).

Attempts to stabilize the optical probe during in vivo spectralacquisition procedures have been made by several groups, where aplain mechanical structure, like a pincer, was used [16,29,30]. Suchan approach would eliminate the lateral and axial movements ofthe optical probe induced by the hand, but might not address theirregular probe contact pressure issue induced by the spontaneousmovements of the in vivo tissue. Based on our experience, in vivobrain movements in a surgical environment are significant, becauseof cardiac cycle and ventilation under anesthesia. A plane mechan-ical holder would yield inconsistent probe contact pressure. Dueto the softness of the brain, inconsistent probe contact pressureslead to detrimental tissue deformation and, thereby, regional alter-ations in hemodynamics [16]. The probe holder design proposedand tested in this paper took this concern into consideration; thesprings in the probe holder reduce contact pressure significantlywhen the probe is placed lightly in contact with the brain sur-face. In addition, it reduces the amount of force (and, hence, thedegree of brain deformation) required to move the probe alongits primary axis. In other words, the probe will move accordingto brain surface movement, without drastically changing probecontact pressure. This improvement consequently reduces motionartifacts in the optical data. Optical spectroscopy is being appliedincreasingly for differentiating various healthy body tissues fromunhealthy ones. Technological and theoretical advances make iteasier to apply it to the study of brain tissue. However, if attentionis not paid to sources of signal variations, researchers risk draw-ing the wrong conclusions from their data. This is especially truefor time-dependent recordings. In this study, the characteristics ofthe motion artifacts induced by an operator’s hand when holdingan optical probe, as well as the spontaneous movements of braintissue were quantitatively demonstrated. Moreover, a new probeholder design that can contain motion artifacts to a great extentwas introduced; hence, small time-dependent variations in diffusereflectance spectra associated with the intrinsic characteristics ofthe brain can be captured.

Funding

None.

Ethical approval

Not required.

Acknowledgements

This research was supported by the MCH Brain Institute,the Thrasher Research Fund, and the Ware Foundation ResearchEndowment Fund. We also acknowledge the Dissertation Year Fel-lowship from FIU.

Conflict of interest statement

None declared.

ering &

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[[

[

N. Yadav et al. / Medical Engine

eferences

[1] Bigio IJ, Bown SG. Spectroscopic sensing of cancer and cancer therapy: currentstatus of translational research. Cancer Biol Ther 2004;3(3):259–67.

[2] Sokolov K, Follen M, Richards-Kortum R. Optical spectroscopy for detection ofneoplasia. Curr Opin Chem Biol 2002;6(5):651–8.

[3] Brown JQ, Vishwanath K, Palmer GM, Ramanujam N. Advances in quantitativeUV–visible spectroscopy for clinical and pre-clinical application in cancer. CurrOpin Biotechnol 2009;20(1):119–31.

[4] Perelman LT. Optical diagnostic technology based on light scatter-ing spectroscopy for early cancer detection. Expert Rev Med Devices2006;3(6):787–803.

[5] Bakker Schut TC, Witjes MJ, Sterenborg HJ, Speelman OC, Roodenburg JL, MarpleET, et al. In vivo detection of dysplastic tissue by Raman spectroscopy. AnalChem 2000;72(24):6010–8.

[6] Wang TD, Van Dam J. Optical biopsy: a new frontier in endoscopic detectionand diagnosis. Clin Gastroenterol Hepatol 2004;2(9):744–53.

[7] Lin WC, Sandberg DI, Bhatia S, Johnson M, Oh S, Ragheb J. Diffuse reflectancespectroscopy for in vivo pediatric brain tumor detection. J Biomed Opt2010;15(6):061709.

[8] Lin WC, Toms SA, Johnson M, Jansen ED, Mahadevan-Jansen A. In vivobrain tumor demarcation using optical spectroscopy. Photochem Photobiol2001;73(4):396–402.

[9] Haglund MM, Berger MS, Hochman DW. Enhanced optical imaging of humangliomas and tumor margins. Neurosurgery 1996;38(2):308–17.

10] Sun Y, Hatami N, Yee M, Phipps J, Elson DS, Gorin F, et al. Fluorescence life-time imaging microscopy for brain tumor image-guided surgery. J Biomed Opt2010;15(5):056022.

11] Mayevsky A, Rogatsky GG. Mitochondrial function in vivo evaluated by NADHfluorescence: from animal models to human studies. Am J Physiol Cell Physiol2006;292:615–40.

12] Benaron DA, Parachikov IH, Friedland S, Soetikno R, Brock-Utne J, van der StarrePJ, et al. Continuous, noninvasive, and localized microvascular tissue oximetryusing visible light spectroscopy. Anesthesiology 2004;100(6):1469–75.

13] Knotzer H, Hasibeder WR. Microcirculatory function monitoring at the bedside– a view from the intensive care. Physiol Meas 2007;28(9):65–86.

14] Mallia R, Thomas SS, Mathews A, Kumar R, Sebastian P, Madhavan J, et al.Oxygenated hemoglobin diffuse reflectance ratio for in vivo detection of oral

pre-cancer. J Biomed Opt 2008;13(4):041306.

15] Wang HW, Jiang JK, Lin CH, Lin JK, Huang GJ, Yu JS. Diffuse reflectancespectroscopy detects increased hemoglobin concentration and decreased oxy-genation during colon carcinogenesis from normal to malignant tumors. OptExpress 2009;17(4):2805–17.

[

Physics 35 (2013) 1551– 1557 1557

16] Ti Y, Lin WC. Effects of probe contact pressure on in vivo optical spectroscopy.Opt Express 2008;16(6):4250–62.

17] Nath A, Rivoire K, Chang S, Cox D, Atkinson EN, Follen M, et al. Effect of probepressure on cervical fluorescence spectroscopy measurements. J Biomed Opt2004;9(3):523–33.

18] Rivoire K, Nath A, Cox D, Atkinson EN, Richards-Kortum R, Follen M. The effectsof repeated spectroscopic pressure measurements on fluorescence intensity inthe cervix. Am J Obstet Gynecol 2004;191(5):1606–17.

19] Shim MG, Song LM, Marcon NE, Wilson BC. In vivo near-infrared Ramanspectroscopy: demonstration of feasibility during clinical gastrointestinalendoscopy. Photochem Photobiol 2000;72(1):146–50.

20] Shangguan H, Prahl SA, Jacques SL, Casperson LW. Pressure effects on soft tis-sues monitored by changes in tissue optical properties. SPIE Proc Laser-TissueInteract IX 1998;3254:366–71.

21] Chan EK, Sorg B, Protsenko D, O’Neil M, Motamedi M, Welch AJ. Effectsof compression on soft tissue optical properties. IEEE J Quantum Electron1996;2(4):943–50.

22] Chen W, Liu R, Xu K, Wang RK. Influence of contact state on NIRdiffuse reflectance spectroscopy in vivo. J Phys D Appl Phys 2005;38:2691–5.

23] Reif R, Amorosino MS, Calabro KW, A‘Amar O, Singh SK, Bigio IJ. Analysis ofchanges in reflectance measurements on biological tissues subjected to differ-ent probe pressures. J Biomed Opt 2008;13(1):010502.

24] Lim L, Nichols B, Rajaram N, Tunnell JW. Time resolved study of probe pres-sure effects on skin fluorescence and reflectance spectroscopy measurements.Biomed Opt OSA Tech Dig (CD) 2010:BTuD102.

25] Ruderman S, Gomes AJ, Stoyneva V, Rogers JD, Fought AJ, Jovanovic BD, et al.Analysis of pressure, angle and temporal effects on tissue optical proper-ties from polarization-gated spectroscopic probe measurements. Biomed OptExpress 2010;1(2):489–99.

26] Greenberg IM. Self-retaining retractor and handrest system for neurosurgery.Neurosurgery 1981;8(2):205–8.

27] Harding JW, Sneddon IN. Proc Cambridge Philos Soc 1945;41:16.28] Green MA, Bilston LE, Sinkus R. In vivo brain viscoelastic properties

measured by magnetic resonance elastography. NMR Biomed 2008;21(7):755–64.

29] Coyle S, Markham C, Lanigan W, Ward T. A mechanical mounting systemfor functional near-infrared spectroscopy. Brain imaging studies Opto-Ireland

2005. Opt Sens Spectrosc 2005;5826:618–27.

30] Chang VT, Merisier D, Yu B, Walmer DK, Ramanujam N. Towards a field-compatible optical spectroscopic device for cervical cancer screening inresource-limited settings: effects of calibration and pressure. Opt Express2011;19:17908–24.