ISSN 0021�3640, JETP Letters, 2013, Vol. 97, No. 12, pp. 690–696. © Pleiades Publishing, Inc., 2013.

690

1. INTRODUCTION

Many qualitative and quantitative spectroscopictechniques have been used for biological�tissue diag�nostics [1–3] with their own set of advantages and dis�advantages [4]. For example, excisional biopsy canalter the original morphology and always can iatro�genic trauma [5]. Hence, using non�invasive opticalimaging modalities such as optical fluorescencemicroscopy, computing tomography, optical diffusetomography and optical coherence tomography(OCT) etc., which are capable to construct 2D or 3Dimages of tissue surfaces as well as the interior of thebody [6–9] can improve the diagnostics of diseases.Optical fluorescence cryomicrotome images havebeen used to reconstruct 3D coronary vasculatureusing point spread function (PSF) by extracting thevascular anatomy (phantom) for light scattering simu�lations in the tissue. The disadvantage includes whenmodeling without correcting for the optical blurring,resulted in 42.9% error on average for the vessel [8].During the past 15 years, OCT has been proved to be anew noninvasive imaging method on micron�scale bymeasuring of backscattered infrared light [10]. Thereason for OCT’s popularity included its high resolu�tion cross�sectional imaging capability (in situ, invitro, and in vivo) [11]. The parameters of opticalcoherence tomography like image depth, resolution,contrast, and acquisition rate imposed with a sourcepower that does not damage the tissue [12, 13] are cru�

¶The article is published in the original.

cial to be considered in different biomedical applica�tions. For example, OCT is currently well suited tohemodynamics of the microvasculature bed study withthe penetration depth ~1–2 mm as compared tomicroscopy [14] at resolution of 2–30 μm [15]. Opti�cal coherence tomography also has been used in situdetermination of the intrinsic optical attenuationcoefficient of atherosclerotic tissue componentswithin regions of interest (ROI) differentiatingbetween various plaque types within the vessel wall [9].In ophthalmology, spectral domain OCT (SD�OCT)or swept source OCT (SS�OCT) offer the segmenta�tion of retinal vessels by using complimentary infor�mation from fundus photographs [16]. The attractivefeature of the OCT is that it does not require any con�tact to object and is nondestructive because light wavesare used for investigations.

In this work, we present a pilot method using thestructural and speckle variance (SV�OCT) subse�quently applied primarily to image soft biological tis�sues of blood vessels in the normal and tumor inducedmice. Although it was challenging due to the sensitiv�ity of the technique to longitudinal motion but the invivo measurement for anesthetized mouse with sweptsource (SS�OCT) was carried out. The resultsobtained with this useful technique have potentialapplication in dermatology, glucose monitoring inblood analyzing Brownian motion of erythrocytesunder the phenomena of DLS or ballistic photon scat�tering fluids in M�mode imaging, and real time imag�ing during photodynamic therapy. Therefore, we have

Human Cervical Carcinoma Detection and Glucose Monitoringin Blood Micro Vasculatures with Swept Source OCT¶

H. Ullaha, b, E. Ahmedb, and M. Ikrama

a Department of Physics and Applied Mathematics, Pakistan Institute of Engineering and Applied Sciences, 45650 Islamabad, Pakistan

b Department of Physics, Bahauddin Zakariya University, Multan, Pakistane�mail: hafeezullah79@gmail.com

Received March 21, 2013; in final form, May 20, 2013

We report a pilot method, i.e., speckle variance (SV) and structured optical coherence tomography to visual�ize normal and malignant blood microvasculature in three and two dimensions and to monitor the glucoselevels in blood by analyzing the Brownian motion of the red blood cells. The technique was applied on nudelive mouse’s skin and the obtained images depict the enhanced intravasculature network forum up to thedepth of ~2 mm with axial resolution of ~8 µm. Microscopic images have also been obtained for both typesof blood vessels to observe the tumor spatially. Our SV�OCT methodologies and results give satisfactory tech�niques in real time imaging and can potentially be applied during therapeutic techniques such as photody�namic therapy as well as to quantify the higher glucose levels injected intravenously to animal by determiningthe translation diffusion coefficient.

DOI: 10.1134/S0021364013120126

JETP LETTERS Vol. 97 No. 12 2013

HUMAN CERVICAL CARCINOMA DETECTION 691

implemented on of the above mentioned applications,i.e., to quantify the glucose level in prior to glucoseinjection mouse as a base line. We have investigated thetranslational relaxation or decorrelation time inBrownian motion rather rotational relaxation. TheSV�OCT images have been used to visualize the bloodvessels in three dimensions so that light can be directlyfocused on the microvasculature for OCT signal. Theidea of quantification of the salt levels in simple stag�nant blood has been translated successfully for in vivocase scenario for non�injected glucose animals [2].This quantitative data will be proved very helpful intranslating the in vitro study into in vivo environmentfor measurements of higher glucose concentrationintravenously into the animal.

2. MATERIALS AND METHODS

2.1. Sample Preparation: Animal Model

The SV algorithm was applied to image the bloodvessels of normal and tumor induced mice (in vivo). Adorsal skin�fold windows chamber model (WCM) wasapplied on anesthetized mouse under institutionallaws of Princess Margret Hospital, Toronto approvedby the University Health Network Animal ResourceCentre. The mouse before surgery was anesthetizedwith a ketamine–xylazine (90–10 mg/kg) mixture. A10 mm diameter region of skin was placed between thetitanium plates of WCM. A 12 mm diameter and250 μm thick cover slip was used to protect theexposed fascia and vasculature. After recovery of theanimal from surgery, OCT imaging was performed byfixing the WCM into the removable lightweight alumi�num plate to keep the animal at 37°C. The mice wereinduced by tumor, ME180 human cervical carcinomacells transfected with the DsRed2 fluorescent protein.After injection the tumor into the fascia we waited forone week so that tumor would grow before imaging.

The purpose of fluorescent protein was to enhance thevisibility of tumor through microscope for subsequentcodetection of fluorescent and SV�OCT images.

2.2. Swept Source OCT Imaging System

OCT images data were obtained from 36 kHz SS�OCT system described earlier in our recent work [2].Briefly, it consists of frequency domain mode locking(FDML) fiber�ring laser source comprising of poly�gon�based tunable filter. The coherence length andspectral sweeping range were 6 and 112 nm, respec�tively, at a fundamental wavelength of 1310 nm. Theaxial resolution in tissue and the average output powerof the system were 8 μm and 50 mW. The coherencelength 6 mm is determined exclusively from sweepingresolution of laser source that was 2 nm.

Our OCT system uses a broadband light source andresolves the scattered signal by dispersing the signalinto many k�values (elementary emanating waves). InSS�OCT, the laser source does not directly sweep lin�early into k�space but rather time space. So, first therecalibration is required from the linear�in�time datato linear�in�k data. The common approach to achievethis goal is to fix the path length for interferometricsignal. Hence the SS�OCT’s signal in k�space can bewritten as

(1)

where S(k) is the incident intensity for each k�valueused (the swept source spectrum) and z0—offset dis�tance between reference plane and object surface. Fork, a positive integral multiple of π/2nz0 the maximaand minima of this equation are produced that specifythe location of equally spaced k values regardless of thetemporal dependence of the sweep. A photograph ofscanning head with animal under measurements isshown in Fig. 1a.

I k( ) S k( ) 2knz0( ),cos≈

Fig. 1. (a) A nude mouse with WCM implanted to image blood vessels with SV�OCT under the scanning head of OCT system.(b) A conceptual diagram of an acquired speckle variance data set of N frames and corresponding indices used to label the frame(i), transverse pixel location (j), and the axial pixel location (k).

692

JETP LETTERS Vol. 97 No. 12 2013

ULLAH et al.

2.3. 3D Image Processing Algorithm: Speckle Variance (SV�OCT)

The interframe OCT fluid contrast algorithmtermed as SV�OCT is based on the idea that the imagespeckle/texture of relatively solid regions will persistbetween consecutive images, whereas speckle willshow greater inter�image speckle washout in regions ofgreater fluidity [17] was designed by our researchgroup (mainly by Dr. Adrian Mariampillai [18]) atPrincess Margret hospital, Toronto, Canada. Thealgorithm for generation of speckle variance images ofOCT data needs to calculate the variance of pixelsfrom a set of N, B�mode images (N is gate length),acquired from the same spatial location

(2)SVijk1N��� Iijk

1N��� Iijk

i 1=

N

∑–⎝ ⎠⎜ ⎟⎛ ⎞

2

,

i 1=

N

∑=

where i, j, and k are indices for the frame (up to N),transverse, and axial pixels, and I is the correspondingpixel intensity value at the imaging speeds of 20 framesper second. A more clearly schematic representationof the data set and pixel indices for three dimensionalframes stacking according to Eq. (2) is shown inFig. 1b.

2.4. Brownian Motion and M�Mode Imagingof in Vivo Mouse

A particle executing Brownian (random) motion in

a liquid has squared displacement and for diffus�ing particle is given by [19]

(3)

r2⟨ ⟩

Δr2

t( )⟨ ⟩ 6DTτT.=

Fig. 2. (a) SV�OCT microvasculature image of a normal mouse in WCM, a low bulk tissue motion situation with gate lengthN = 8. This shows a 3D processed image of blood microvasculature to distinguish the individual blood vessels. Dimensions:(6400 × 1520 × 512, pixels) and/or (6 × 6 × 2.2 mm3). The ROI is represented with magnification and depth encoded color barused to specify the depth of the vessel. Scale bar: 250 µm (b) a microscopic z�stack image displaying the micro vessels with highlateral resolution but minimal axial information. Dimension: (6.2 × 6.2 mm) and scale bar: 1mm (color online).

JETP LETTERS Vol. 97 No. 12 2013

HUMAN CERVICAL CARCINOMA DETECTION 693

Fig. 3. (a) Blood vessels of nude tumor induced mouse obtained with SV�OCT. This shows a 3D processed image of bloodmicrovasculature to distinguish the individual blood vessels. Dimensions: (6400 × 1520 × 512, pixels) and/or (6 × 6 × 3.2 mm) andscale bar: 250 µm and (b) corresponding maximum�intensity projection image of a fluorescence confocal z�stack obtained using500 kD Fluorocein labeled dextran. Dimensions: (2.2 × 2.2 mm) and scale bar: 500 µm (color online).

Here, DT (translation diffusion coefficient) and τT

(translational relaxation or decorrelation time) arerelated by the formula [19]

(4)

This gives us the diffusion coefficient extracted fromthe translations decorrelation time. The decay is expo�nential in nature [2]. To get M�mode data, we selectedthe blood vessel of diameter ~160 μm at depth of~55 μm from top surface of the tissue and can be seenin SV�OCT microvasculature image (Fig. 2a). Thedata from three ROIs was averaged for each animalbecause of the little bit difference in decorrelationtime due to difference on flowing speed or shear effect.The measurement scheme for in vivo M�mode mea�surements was same as described recently by our group[2]. We examined a total of 5 animals in this experi�ment. The obtained M�mode OCT data was processedto calculate the autocorrelation function (ACF) frompower spectrum, P(ω) by means of Fourier transfor�mation using Weiner–Khinchin theorem [2, 20].

A double exponential fit function f1 =

B*exp(⎯t/ ) + C*exp(–t/ ) was applied on theACF because of the asymmetric shape of RBCs. Fromabove fitted function we got both and are indeedoccurring in dynamic light scattering for in vivo casebecause of transitional and rotational motion but weprocessed the data for translational diffusion coeffi�cient rather rotational. This was decided after examin�ing the B/C ratio (relative importance of translationalrelaxation versus rotational relaxation) which wasfound to be ~3.85 that strongly recommends thattranslational relaxation is dominant over rotationalmotion.

τT1

2k2DB

������������.=

τT' τR'

τT' τR'

3. RESULTS

Figure 2a shows a blood vessels image of normalnude mouse obtained with SV�OCT. This shows a 3Dprocessed image of blood microvasculature to distin�guish the individual blood vessels with dimensions of6 × 6 × 2.2 mm. Figure 2b shows a light microscopicimage displaying the micro vessels with high lateralresolution with dimension of 6.2 × 6.2 mm in WCM.Figure 3a shows blood vessels having tumor induced innude mouse obtained with SV�OCT with dimensionsof 6 × 6 × 2.2 mm. This reflects the capability of SV�OCT to map the abnormalities in blood vessels thatcould be implemented in diabetes having tumor dis�eases. Figure 3b shows corresponding maximum�intensity projection image of a fluorescence confocalz�stack image obtained using 500 kD Fluoroceinlabeled dextran with dimensions of 2.2 × 2.2 mm.

Figure 4a gives an original structured OCT imagein which different layers under laying the fascia can beobserved but the image is not clearer for diagnosticpoint of view. Anyhow, we get sufficient informationup to this limit for layers differentiations. The corre�sponding OCT signal decaying exponentially for nudefemale mouse skin in vivo is shown in Fig. 4b. The firstpeak is due to the backscattering of cover glass and sec�ond peck represents the surface of the tissue. Figure 4cshows the colorized image with efficient noiseremoved with code written by us for colorization invisual C++ resulting in an easy analysis of the image[21, 22].

The SV�OCT image used to select the ROI for invivo study of glucose monitoring in five animals withthe help of M�mode imaging the average value of DT

was yield ~5.85 × 10–14 m2/s (table). This value was~10% different from theoretical value of diffusioncoefficient (5.85 × 10–14 m2/s vs. 6.50 × 10–14 m2/s

694

JETP LETTERS Vol. 97 No. 12 2013

ULLAH et al.

[23]). This difference might be originated due toforced convection, shear effect and techniqueemployed for measurements. As the technique used in[23] to determine DT was diffuse wave spectroscopy(DWS) applied ex vivo on porcine kidney model atcontrolled arterial pressure and flow. Figure 2a pro�vides clear 3D image of glucose free mouse to visualizethe blood microvessells for easy selection of the ROI toobtain M�mode data. Thus, we get the threshold mea�surement to quantify glucose levels in case of injectedhigher glucose concentrations intravenously.

4. DISCUSSIONS

The SV algorithm represented in Fig. 1b providesan interfame algorithm to mitigate the bulk tissuemotions so that phase sensitive images can be envi�sioned with high precisions. Figure 2a shows a recon�structed image of the vascular structure in the skin ofmouse using SV�OCT analysis. This shows a 3D pro�cessed image with Eq. (2) of blood microvasculature todistinguish the individual blood vessels in depth withaxial resolution of ~8 μm and lateral resolution of~13 μm. In Fig. 2b, we see that light microscopicimage of blood vessels presents higher lateral resolu�tion but low axial information. Our reconstruction for3D view from structured images gives full map (axi�ally) of field of view and has capabilities to visualizethe individual vessels up to the depth of ~2.2 mm.Because in SV algorithm, we have segmented out thehigher inter�frame variance (~20 frames per image)possibly due to dynamic scatterers such as RBCs inblood vessel of various diameters. The size and depthof individual vessel is given by the color bar providedon the right hand side of the ROI image.

SV algorithm has potential application to demon�strate the life morphology by identifying the densebranching of microvasculature both in normal andmalignant skins. The microvasculature system under�goes a significant change in the density and shape afterinducing the tumor i.e. vascular plexus of blood circu�

lation is perturbed and thrombosis can be seen (Figs. 2and 3). This hierarchical network in hemodynamic hasinteresting potential application like monitoring ofanalytes in our recent study of glucose quantificationwith M�mode OCT and SV�OCT [2] before and afterinjection of glucose intravenously. This would changethe density of vessels after changing in the shape oferythrocytes when glucose injected in blood vessels.Hence, the visualization of vasculature with SV�OCTcan provide valuable information underneath the skinin diabetes either cancerous or non�cancerous.

The blood borne glucose provides the constantsupply to cerebral metabolism [24]. The glucose trans�ported to nerve tissue is separated by functional barrier(the blood�brain barrier, BBB), a characteristic of thecerebral endothelium. To know about its metabolicregulation and developmental expression in the BBB[25], the vasculature net is necessary to visualize in anyof the conditions like normal glucose supply, hyperg�lycemia, and hypoglycemia. Thus, we provide a basicroute to study modulations in micro blood vessels bothin 2D and 3D imaging with structures OCT and SV�OCT images. Another application of this study is theunderstanding of the role of mechanical factors in car�diovascular development [17] depending on theapparent morphological difference between the nor�mal and abnormal primitive vascular plexus. The lim�itations of SV�OCT include sensitivity for high inten�sity fluctuations due to shadowing effect [26], because,the interframe variance is produced due to its sensitiv�ity to bulk tissue motion (BTM). This BTM was min�imized by using WCM in this study or a high speedusing imaging system can be used, i.e., high A�scanacquisition speed to ensure sufficient number ofacquired frames for calculation of speckle decorrela�tion. Another cause of this variance in the intensity ofeach voxel might be the shape of the scattering parti�cles such as red blood cells, keratinocyte like StarumGranulosum, Starum Spinosum, and Stratum Basalein epidermis lie in different shapes, sizes and refractiveindexes [6].

It is worth to mention that OCT has increased dra�matically the efficacy of preclinical studies to monitorthe disease on real time. During imaging the live ani�mals, anesthesia is much necessary to minimize themotion artifacts (i.e., to restrain the animal) and isimportant challenge in contrast to human studies. Theinjection of anesthesia drug to animal reasonablychanges the animal’s physiology that can export moreor less different data to imaging device which is muchdifficult in human beings. The repeated anesthesia,exposure of ionizing radiations and the use of contrastagents and/or imaging biomarkers can also affect thephysiology of animal [27] that should be consideredduring the results analysis.

Therefore, speckle contrast levels in SV�OCT pro�vide a better 3D understanding than original struc�tured unprocessed images for in vivo skin. Our B�scan

Summary of Brownian motion analysis results in dynamiclight scattering regime including translational decorrelationtime, translational diffusion coefficient and r2�values for allfive animals

AnimalsTranslational decorrelation

time, ms

Translational diffusion

coefficient,10–14 m2/s

r2�value

Animal 1 38.20 1.02 0.93

Animal 2 44.00 6.68 0.95

Animal 3 48.80 6.03 0.98

Animal 4 46.70 6.30 0.99

Animal 5 52.20 9.20 0.96

Average 45.98 5.85 0.96

JETP LETTERS Vol. 97 No. 12 2013

HUMAN CERVICAL CARCINOMA DETECTION 695

swept source method is sufficient and straightforwardfor transverse sampling of the full speckle pattern inpractical implementations. The technique can be

applied during photodynamic therapy (PDT) proce�dure by monitoring the blood vasculature up to a depthof few millimeters in case of superficial tumors [28].But during PDT, there are multiple tissue reactionspost PDT. Our fundamental in vivo work of glucosemonitoring has a significant application in the patientsof hyperglycemia, where, fibrinogen depletion canreduce the blood viscosity after intravenous adminis�tration of ancrod. We have reported therefore the ref�erence value of glucose diffusion by measuring diffu�sion coefficient of healthy animals that can be imple�mented initially in lab experiments as a non�invasivemethod. Therefore OCT’s capability for monitoringglucose in blood vessels of diameters in micron rangesprovides the route to enhance the application for fur�ther glucose concentration’s quantification.

5. CONCLUSIONS

The potential application of speckle variance OCTand structured OCT was validated to visualize the skinof live nude mouse. The quantitative metrics devel�oped and implemented in this study uncover and mea�sure vascular characteristics from 3D SV�OCTimages. Our method demonstrates that SV�OCT isapplicable for microvasculature detection in normaland tumor induced vessels within WCM and carriesmore information than structured OCT. That is moreaccurate and unique representation of the vascularstructure underlying biology of vasculature develop�ment and response. For in vivo study of glucose mon�itoring the RBCs can have the same dynamics as areused in flowing blood (in vitro study) through the cap�illary tube to reinforce the concept that RBCs are onlyasymmetrical particles that cause the change in theOCT signal due to Brownian motion underlying theDLS. Thus, the translational diffusion coefficientshave been determined to provide a base line in this per�spective to image the deformation in microvasculaturefor further higher glucose levels.

This work was supported by Higher EducationCommission Pakistan, Islamabad, Pakistan. One ofthe authors, Dr. Hafeez Ullah highly acknowledgesDr. Prof. Alex Vitkin, Leigh Conroy, Adrian Mari�ampillai, Michael Wood, and Azusa Maeda at Depart�ment of Medical Biophysics, University of Toronto,Canada, who allowed and helped him to use the Bio�photonics laboratory.

REFERENCES

1. A. Lemelle, B. Veksler, I. S. Kozhevnikov, et al., LaserPhys. Lett. 6, 71 (2009).

2. H. Ullah, G. Gilanie, M. Attique, et al., Laser Phys. 22,1002(2012).

3. S. Oeltze and B. Preim, IEEE Trans. Med. Imaging 24,540 (2005).

4. N. Sharma and L. M Aggarwal, J. Med. Phys. 35, 3(2010).

Fig. 4. (a) Original gray scale structured OCT image ofmouse skin. Dimensions (640 × 512 pixels) pixels, where1 pixel = 3.9 µm (color online). (b) The OCT signal decay�ing exponentially for nude female mouse skin in vivo. Thefirst peak is due to the backscattering of cover glass and sec�ond peck represents the surface of the tissue. The depthcontains a total of 512 pixels in ~3 mm depth, again 1 pixel ~3.9 µm. (c) Noise removed hard segmented OCT imagewith dimensions (256 × 256) pixels, where 1 pixel = 3.9 µm.

696

JETP LETTERS Vol. 97 No. 12 2013

ULLAH et al.

5. T. Gambichler, G. Moussa, M. Sand, et al., J. Derma�tol. Sci. 40, 85 (2005).

6. H. Ullah, M. Atif, S. Firdous, et al., Laser Phys. Lett.7, 889 (2010).

7. B. Veksler, E. Kobzev, M. Bonesi, and I. Meglinski,Laser Phys. Lett. 5, 236 (2008).

8. A. Goyal, J. Lee, P. Lamata, et al., IEEE Trans. Med.Imaging 32, 56 (2013).

9. F. J. van der Meer, D. J. Faber, D. M. B. Sassoon, et al.,IEEE Trans. Med. Imaging 24, 1369 (2005).

10. A. F. Fercher, W. Drexler, C. K. Hitzenberger, andT. Lasser, Rep. Progress Phys. 66, 239 (2003).

11. M. Atif, H. Ullah, M. Y. Hamza, and M. Ikram, LaserPhys. Lett. 8, 629 (2011).

12. J. G. Fujimoto, Handbook of Optical Coherence Tomog�raphy (Marcel Dekker, New York, 2002).

13. Z. Hamdoon, W. Jerjes, T. Upile, and C. Hopper, Pho�todiagn. Photodyn. Ther. 8, 49 (2011).

14. S. Gu, M. W. Jenkins, L. M. Peterson, et al., Dev. Dyn.241, 534 (2012).

15. J. G. Fujimoto, Nature Biotechnol. 21, 1361 (2003).16. H. Zhihong, M. Niemeijer, M. D. Abramoff, and

M. K. Garvin, IEEE Trans. Med. Imaging 31, 1900(2012).

17. N. Sudheendran, S. H. Syed, M. E. Dickinson, et al.,Laser Phys. Lett. 8, 247 (2011).

18. A. Mariampillai, PhD Thesis (Dept. Med. Biophys.,Univ. of Toronto, Toronto, 2010).

19. B. J. Berne and R. Pecora, Dynamic Light Scatteringwith Applications to Chemistry, Biology, and Physics(Dover, New York, 2000).

20. M. Brezinski, Optical Coherence Tomography Principlesand Applications (Elsevier, San Diego, 2006).

21. M. Attique, G. Gilanie, Hafeez�Ullah, et al., PLoSONE 7, e33616 (2012). doi:10.1371/jour�nal.pone.0033616.

22. G. Gilanie, M. Attique, Hafeez�Ullah, et al., PatternRecogn. Lett. (2013).

23. M. Ninck, M. Untenberger, and T. Gisler, Biomed.Opt. Exp. 1, 1502 (2010).

24. R. Dermietzel, D. Krause, M. Kremer, et al., Dev. Dyn.193, 152 (1992).

25. A. J. Barber and E. Lieth, Dev. Dyn. 208, 62 (1997).26. J. Fingler, D. Schwartz, C. Yang, and S. E. Fraser, Opt.

Express 15, 12636 (2007).27. L. Tremoleda, A. Kerton, and W. Gsell, Eur. J. Nucl.

Med. Mol. Imaging 2, 1 (2012).28. G. Hüttmann, Photodiagn. Photodyn. Ther. 8, 152

(2011).