+ All Categories
Home > Documents > Discrimination of prostate carcinoma from benign prostate tissue fragments in vitro by estimating...

Discrimination of prostate carcinoma from benign prostate tissue fragments in vitro by estimating...

Date post: 20-Jan-2017
Category:
Upload: carlos-augusto
View: 213 times
Download: 1 times
Share this document with a friend
9
ORIGINAL ARTICLE Discrimination of prostate carcinoma from benign prostate tissue fragments in vitro by estimating the gross biochemical alterations through Raman spectroscopy Landulfo Silveira Jr & Kátia Ramos M. Leite & Fabricio Luiz Silveira & Miguel Srougi & Marcos Tadeu T. Pacheco & Renato Amaro Zângaro & Carlos Augusto Pasqualucci Received: 26 August 2013 /Accepted: 10 February 2014 # Springer-Verlag London 2014 Abstract Raman spectroscopy has been proposed for detect- ing biochemical alterations in prostate cancer (PrCa) com- pared to benign prostate tissue in in vitro fragments from surgery for diagnostic purposes. Freezer-stored fragments of human prostate tissues were unfrozen and submitted to Raman spectroscopy with a dispersive spectrometer (830-nm and 200-mW laser parameters, 30-s exposure time). Fragments were fixed and submitted to histopathology to grade PrCa according to Gleason score. A total of 160 spectra were taken from 32 samples (16 benign tissues and 16 PrCa tissues). The relative concentrations of selected biochemicals were estimat- ed using a least-squares fitting model applied to the spectra of pure compounds and the tissue spectrum. A discrimination model was developed employing the most statistically relevant compounds with capability of separating PrCa from benign tissues. The fitting model revealed that actin, hemo- globin, elastin, phosphatidylcholine, and water are the most important biochemicals to discriminate prostate depending on the Gleason score. A discrimination based on Euclidean dis- tance using the relative concentrations of phosphatidylcholine and water showed the higher accuracy of 74 % to discriminate PrCa from benign tissue. Raman spectroscopy is an analytical technique with possibility for identifying biochemical consti- tution of prostate and could be used for diagnostic purposes. Keywords Raman spectroscopy . Prostate cancer . Discrimination . Biochemical model . Euclidean distance Introduction Prostate cancer (PrCa) is the sixth most common type of cancer worldwide and the most prevalent in men, representing around 10 % of all cases of cancer [1]. The incidence rate of this cancer is around sixfold higher in developed countries compared to that in developing nations, varying greatly from country to country and between continents, with a mortality ratio that varies from 0.13 in North America to 0.80 in Africa [2]. The PrCa is considered a cancer of the elderly, since around three quarters of cases occur in men over 65 years of age [3], whereas cases diagnosed in the age of 5564 are increasing [4]. In the United States, the estimated number of new cases of PrCa for the year 2009 was 155 cases per 100,000 men [4]. In Brazil, excluding the non-melanoma skin cancer, PrCa is the most frequent neoplasia affecting men in all regions, with 62 cases per 100,000 men [5]. Researchers have developed several tools to diagnose and assess the potential for PrCa disease progression, such as the L. Silveira Jr : F. L. Silveira : M. T. T. Pacheco : R. A. Zângaro Biomedical Engineering Institute, Universidade Camilo Castelo BrancoUNICASTELO, Parque Tecnológico de São José dos Campos, Estr. Dr. Altino Bondesan, 500, Eugênio de Melo, São José dos Campos, SP 12247-016, Brazil K. R. M. Leite : M. Srougi Laboratory of UrologyLIM 55, University of São Paulo Medical SchoolFMUSP, Av. Dr. Arnaldo, 455, São Paulo, SP 01246-903, Brazil C. A. Pasqualucci Laboratory of Cardiovascular PathologyLIM 22, University of São Paulo Medical SchoolFMUSP, Av. Dr. Arnaldo, 455, São Paulo, SP 01246-903, Brazil L. Silveira Jr (*) Universidade Camilo Castelo BrancoUNICASTELO, Parque Tecnológico de São José dos Campos, Estr. Dr. Altino Bondesan, 500, São José dos Campos, SP 12247-016, Brazil e-mail: [email protected] L. Silveira e-mail: [email protected] Lasers Med Sci DOI 10.1007/s10103-014-1550-3
Transcript

ORIGINAL ARTICLE

Discrimination of prostate carcinoma from benign prostate tissuefragments in vitro by estimating the gross biochemical alterationsthrough Raman spectroscopy

Landulfo Silveira Jr & Kátia Ramos M. Leite & Fabricio Luiz Silveira &

Miguel Srougi & Marcos Tadeu T. Pacheco & Renato Amaro Zângaro &

Carlos Augusto Pasqualucci

Received: 26 August 2013 /Accepted: 10 February 2014# Springer-Verlag London 2014

Abstract Raman spectroscopy has been proposed for detect-ing biochemical alterations in prostate cancer (PrCa) com-pared to benign prostate tissue in in vitro fragments fromsurgery for diagnostic purposes. Freezer-stored fragments ofhuman prostate tissues were unfrozen and submitted to Ramanspectroscopy with a dispersive spectrometer (830-nm and200-mW laser parameters, 30-s exposure time). Fragmentswere fixed and submitted to histopathology to grade PrCaaccording to Gleason score. A total of 160 spectra were takenfrom 32 samples (16 benign tissues and 16 PrCa tissues). Therelative concentrations of selected biochemicals were estimat-ed using a least-squares fitting model applied to the spectra ofpure compounds and the tissue spectrum. A discriminationmodel was developed employing the most statistically

relevant compounds with capability of separating PrCa frombenign tissues. The fitting model revealed that actin, hemo-globin, elastin, phosphatidylcholine, and water are the mostimportant biochemicals to discriminate prostate depending onthe Gleason score. A discrimination based on Euclidean dis-tance using the relative concentrations of phosphatidylcholineand water showed the higher accuracy of 74 % to discriminatePrCa from benign tissue. Raman spectroscopy is an analyticaltechnique with possibility for identifying biochemical consti-tution of prostate and could be used for diagnostic purposes.

Keywords Raman spectroscopy . Prostate cancer .

Discrimination . Biochemical model . Euclidean distance

Introduction

Prostate cancer (PrCa) is the sixth most common type ofcancer worldwide and the most prevalent in men, representingaround 10 % of all cases of cancer [1]. The incidence rate ofthis cancer is around sixfold higher in developed countriescompared to that in developing nations, varying greatly fromcountry to country and between continents, with a mortalityratio that varies from 0.13 in North America to 0.80 in Africa[2]. The PrCa is considered a cancer of the elderly, sincearound three quarters of cases occur in men over 65 years ofage [3], whereas cases diagnosed in the age of 55–64 areincreasing [4]. In the United States, the estimated number ofnew cases of PrCa for the year 2009 was 155 cases per100,000 men [4]. In Brazil, excluding the non-melanoma skincancer, PrCa is the most frequent neoplasia affecting men inall regions, with 62 cases per 100,000 men [5].

Researchers have developed several tools to diagnose andassess the potential for PrCa disease progression, such as the

L. Silveira Jr : F. L. Silveira :M. T. T. Pacheco : R. A. ZângaroBiomedical Engineering Institute, Universidade Camilo CasteloBranco—UNICASTELO, Parque Tecnológico de São José dosCampos, Estr. Dr. Altino Bondesan, 500, Eugênio de Melo, São Josédos Campos, SP 12247-016, Brazil

K. R. M. Leite :M. SrougiLaboratory of Urology—LIM 55, University of São Paulo MedicalSchool—FMUSP, Av. Dr. Arnaldo, 455, São Paulo, SP 01246-903,Brazil

C. A. PasqualucciLaboratory of Cardiovascular Pathology—LIM 22, University ofSão Paulo Medical School—FMUSP, Av. Dr. Arnaldo, 455, SãoPaulo, SP 01246-903, Brazil

L. Silveira Jr (*)Universidade Camilo Castelo Branco—UNICASTELO, ParqueTecnológico de São José dos Campos, Estr. Dr. Altino Bondesan,500, São José dos Campos, SP 12247-016, Brazile-mail: [email protected]

L. Silveirae-mail: [email protected]

Lasers Med SciDOI 10.1007/s10103-014-1550-3

magnetic resonance imaging [6], but only two are widelyrecognized to be good predictors of cancer progression: pros-tate specific antigen (PSA) testing and Gleason score [7, 8].The most important risk factors are PSA serum levels, clinicalstage, and Gleason score. The treatment of PrCa relies uponthe result of the Gleason scoring and the risk of progression,ranging from active surveillance through palliative anti-androgen therapy [9]. On the other hand, although traditionalprostate examinations (digital rectal examination and PSAtesting) are being widely used as a screening test and are usedroutinely for the detection of PrCa, they are not diagnostic andlack of sensitivity and specificity [10]. Gleason gradingthrough histopathological examination is considered the bestprognostic indicator in prostate cancer. The Gleason gradingevaluates prostate cancer cells (either from needle biopsies orfrom radical prostatectomy) on a scale of 1 (well differentiat-ed) to 5 (poorly differentiated), based on their glandular dif-ferentiation at a low-power microscope [11]. The Gleasonscore combines the Gleason grade for the first and second mostpredominant patterns within the biopsies from about 12 or morecore samples, removed from different parts of the prostate,resulting in scores of 2 (1+1) to 10 (5+5). Gleason scores of 2to 4 (well differentiated) almost never develop aggressive dis-ease, whereas 8 to 10 usually die of the disease [11]. However,inter-observer variation can occur, and the grading of the biopsymay not correlate with the prostatectomy specimen because ofsampling problems, and in some cases morphologically identi-cal prostate cancer can behave differently [12–14].

Optical techniques such as fluorescence and Raman spec-troscopy have been proposed for an early diagnosis of humandiseases [15–23], by detecting and quantifying biomarkersresponsible for or generated as a response of the disease.Raman spectroscopy is an optical technique with the potentialfor detecting diseases in vivo in real time, by obtaining infor-mation of the presence of molecular groups of proteins, lipids,and nucleic acids from bio-tissues [21]. Raman spectroscopyhas been proposed as a diagnostic tool of a variety of humanillnesses in different organs and tissues [24–28]. Cancer diag-nosis through Raman spectroscopy has received special con-sideration due to the molecular specificity obtained from theRaman spectrum. By using dispersive and FT-Raman spec-troscopy, different neoplastic tissues have been characterizedsuch as breast [22], gastric [29], laryngeal [30], colonic [31],urological [23, 32], and epithelial from esophagus, colon,prostate, and breast [33].

The process of Raman scattering can be viewed as aninelastic phenomenon in which the incident photon is scatteredwith a frequency shift as it either gains energy from or losesenergy to a particular vibrational mode of the molecule[21–23, 34]. Thus, Raman spectroscopy can be used to accessthe molecular constitution of a specific sample, revealing thepresence of specific molecular groups, and then classify itaccording to differences observed in the spectra [21].

The PrCa tissues have been studied in order to develop amethodology for optical biopsy using Raman spectroscopy[23, 32, 33, 35–37]. Raman studies on prostate tissues re-vealed variations in the molecular content (bands of glycogenand nucleic acid) between adenocarcinoma and benign pros-tatic hyperplasia (BPH) [35], despite the small spectral differ-ences between neoplasia and benign/normal tissues. Crowet al. [38] aimed to determine whether Raman spectroscopycould differentiate between PrCa cell lines of various degreesof biological aggressiveness, determining the differences inthe molecular composition of different cell lines classifiedthrough an algorithm using principal component analysis(PCA). Crow et al. [32] showed that Raman spectroscopycould be used to accurately identify BPH and three differentgrades of prostatic PrCa in vitro through a diagnostic algo-rithm based on PCA. Stone et al. [23] showed that Ramanspectroscopy may provide a methodology for noninvasivedetection of diseases by quantifying the biochemicals whichare present in normal and diseased urological tissues, such asproteins, lipids, and nucleic acids, with specific informationfor classifying and grading the malignant neoplasm. Lopeset al. [37] developed a simplified spectral model to estimatethe concentration of lipid-like and protein-like spectral fea-tures to distinguish normal, hyperplastic prostate, and prostatecarcinoma biopsy fragments in vitro and compared to theresults obtained by principal component analysis. This maylead to the development of an optical diagnostic tool for thestudy of the evolution of pre-cancerous and cancerous lesionsin vivo [22, 32].

The objective of the present work was to develop a spectralmodel based on dispersive Raman spectroscopy (at 830-nmexcitation) to probe the prostate biochemical constitution andreveal the chemical differences between benign and PrCatissues obtained from surgical specimens, correlating suchspectral differences to the Gleason score (histopathology).The proposed diagnostic model was based on the relativeamount of the biochemical compounds with significant dif-ferences between benign PrCa in two different Gleason scoreranges (≤7 and >7), calculated by least-squares fitting thebiochemicals spectra and the tissue spectra, discriminated byEuclidean distance, thus, obtaining a classification of benignand PrCa samples with different Gleason grades according tothe differences in the relative concentration of these selectedcompounds.

Material and methods

Prostate tissue samples

Fragments of prostate tissue from 32 patients of approximate-ly 10 mm×10 mm×5 mm (width×length×thickness) wereobtained in the bench, 10 min after being removed from the

Lasers Med Sci

patient. The pathologist examined the specimens, and thehardened areas of the peripheral zones, considered suspiciousfor tumor, were chosen to be studied, snap-frozen, and storedin liquid nitrogen (−196 °C) prior to Raman spectroscopyanalysis. After the Raman spectra collection, the sample frag-ment was fixed with 10 % buffered formalin, paraffin embed-ded, and examined by an experienced uropathologist forGleason grading. The Gleason score was obtained by micro-scopically evaluating about five spots of the fragment.

Raman spectroscopy

At the time for Raman data collection, frozen samples weremoisturized with 0.9 % saline to reach room temperature.Tissue fragments were placed in an aluminum sample holder.Raman spectra were obtained using a table-top dispersiveRaman spectrometer (model Dimension P-1, LambdaSolutions, MA, USA) as described elsewhere [15, 16].Briefly, the Raman excitation is done by a diode laser operat-ing at 830-nm wavelength and 200-mW output power. Thelaser is coupled to a fiber-optic-based Raman probe (modelVector Probe, Lambda Solutions, MA, USA), which directsand focuses the excitation light to the sample, filters unwantedRaman scattering generated by the fiber, and collects thescattered light from the tissue, blocking the Rayleigh scatterfrom the sample. Then, the collection fiber is coupled to thespectrometer that disperses and directs the Raman signal to a1,320×100 pixels CCD camera (back illuminated, deep de-pleted CCD chip, −75 °C working temperature). All spectrawere obtained with an integration time of 30 s under a spec-trometer resolution of about 4 cm−1, in the spectral range from800 to 1,800 cm−1. Five spectra have been collected in differ-ent spots of each fragment, at the tissue surface, totaling 80spectra for benign and 80 spectra for PrCa. Using absorptionand scattering coefficient data of prostate tissue from recentliterature measured at 785 nm (μa~0.5 cm−1, μs′~7.0 cm−1,and μeff~3.0 cm−1) [38], the laser penetration depth of thelaser for this wavelength was estimated to be about 0.3 to0.5 cm [39, 40]. At this integration time, tissue damage(burning) was not observed as laser wavelength is not stronglyabsorbed by prostate chromophores, as the absorption coeffi-cient ranges from 0.3 to 0.5 cm−1 and pulse duration is longenough to have thermal diffusion [40]. Fluorescence back-ground was observed in most of the samples, which could beresponsible for the reduced signal-to-noise ratio.

After intensity and wavenumber calibration, all spectrawere preprocessed, through a fluorescence background re-moval procedure employing a fifth-order polynomial filter,which has been fitted over the useful spectral range andsubtracted. The unwanted cosmic ray spikes were handlyremoved. Finally, each spectrum was normalized using thearea under the curve, where all Raman bands in the usefulrange were rectified, the intensities were summed, and the

result divided by the number of datapoints. Due to the strongfluorescence in some samples, three spectra of benign and fivespectra of PrCa were withdrawn from the data set.

Estimation of prostate biochemical constitutionthrough the Raman of basal compounds

In order to estimate the contribution of the basal compounds tothe observed Raman spectrum of prostate tissue, a spectralmodel, as described elsewhere [15, 22, 23, 37], was performedby calculating the relative amount of most relevant biochem-icals found in prostate tissues and responsible for the Ramanbands appearing in the bulk spectrum. Actin, elastin, leucine,glycogen, blood (hemoglobin), phosphatidylcholine(phospholipids), beta-carotene, and water, some of them de-scribed by Stone et al. [23], were obtained and the Ramanspectra measured. These spectra were then modeled in orderto estimate the “Raman concentration,” i.e., the relative con-centration based on the Raman scattering of each molecule inthe tissue spectrum. The differences in the amount of tissuebiochemicals estimated by the Raman spectral model havebeen correlated to the Gleason score of the prostate fragment[23].

The spectral model has been developed by fitting thespectra of most relevant tissue constituents to the spectra ofprostate tissues using ordinary least-squares fitting, assumingthat the components selected are the major components of thespectra, and then solving the expression X=C ·SC, where X isthe matrix of the original spectra, SC is the matrix of tissueconstituents spectra, and C is the matrix of Raman coefficientsto be predicted. The predicted coefficients are obtained in aleast-squares sense: C=X/SC. Then, these coefficients areplotted, and a suitable discriminant analysis technique suchas the Euclidean distance is applied to separate tissue type ingroups according to the histopathology. Observation of theresidual of the fitting enabled the quality of the fit to beevaluated. The disadvantage of this fitting is that any colin-earity in the components selected or exclusion of importantcompounds would skew the fit [23]. The fitting was per-formed using the software Matlab 7.0.

Results and discussion

The mean normalized Raman spectra of in vitro benign pros-tate and PrCa tissues with Gleason score ≤7 and Gleason >7are plotted in Fig. 1. The spectra of prostate exhibit bandsassignable to structural proteins, amino acids, lipids/phospholipids, and nucleic acids. Observed prominent bandsat 860, 879, 941, 1,006, 1,120 to 1,220, 1,260 to 1,300, 1,318,1,334, 1,455, 1,498, 1,519, 1,568, 1,610, and 1,659 cm−1 aredue to molecular compounds which are described in Table 1.A great similarity in the spectra of both benign and PrCa with

Lasers Med Sci

Gleason ≤7 has been observed, indicating a similarity inconstitution. Higher spectral differences were observed forthe PrCa with Gleason >7. Despite the small differences inthe spectra of benign and PrCa samples, the most significantdifferences (t test, p<0.05) with potential use for the proposeddiscriminant model due to the statistical significance betweenbenign and Gleason >7 are the ones at 860, 879, 941, 1,006,1,120 to 1,200, 1,455, 1,498, 1,519, 1,568, and 1,610 cm−1,attributed basically to proteins/amino acids, lipids, carbohy-drates, carotenoids, and nucleic acids.

The Raman data set (77 spectra of benign and 75 spectra ofPrCa) were used to develop the model based on the relativeconcentration of spectrally relevant biochemicals. The fittingof the tissue spectra with the basal compounds resulted inconcentrations C relative to the Raman cross section, referred

as “Raman concentrations,” found in each tissue spectrum.Figure 2 shows the normalized spectra of the basal com-pounds actin, elastin, beta-carotene, blood (hemoglobin), gly-cogen, leucine, phosphatidylcholine (phospholipid), and wa-ter, showing bands in the positions of N and PrCa tissues andused in the fitting model. Phosphatidylcholine was used tomodel both phospholipids and unsaturated lipids from cell fat,since the lipidic portion of this compound is unsaturated(oleic) acid.

Figure 3 presents the mean and standard deviation of theRaman concentration C of basal compounds for each group ofprostate tissue benign and PrCa, being the PrCa divided in twosubgroups: Gleason score ≤7 and Gleason score >7.Significant differences in the Raman concentration of actin,blood, elastin, phosphatidylcholine, and water for both benign

Fig. 1 Normalized mean Ramanspectra of prostate tissuesobtained in vitro: benign,carcinoma with Gleason ≤7, andcarcinoma with Gleason >7.Labeled peaks have attributionaccording to Table 1. Asterisk andfilled circle indicate peaks withstatistically significantdifferences: asterisk betweenbenign and Gleason >7 and filledcircle between Gleason ≤7 andGleason >7 (t test, p<0.05)

Table 1 Raman peaks related to molecular compounds of prostate tissue referred to the recent literature [33, 50, 51]

Raman bands(cm−1)

Biomolecular compounds

860 PO4 stretch (lipids—phosphate group); CCH bending (aromatic); and C–C stretching (proteins—leucine)

879 CH2 rocking and C–C stretching (proteins); CH3 and C–C–N+ symmetric stretching (phospholipids—choline); and C–O–C ring(glycogen)

941 P(CH3) and C–C stretching (proteins); C–O–C ring (glycogen)

1,006 C–C stretching of aromatic ring (proteins—phenylalanine)

1,120–1,220 C–C and C–N stretching (proteins); C–C skeletal stretching—trans conformation (phospholipids); C=C stretch mode (carotenoids);C–H in-plane bending and CH2 wagging (proteins); C–O and C–C stretching of aromatic ring (glycogen)

1,260–1,300 C–N stretching and N–H in-plane bending (proteins); CH modes (CH2 twisting and wagging) (lipids and proteins); =C–Hbending—cis conformation (phospholipids)

1,318 CH2 and CH3 twisting and wagging (proteins and lipids); C–H deformation (proteins)

1,344 CH deformation (proteins and carbohydrates); CH2 and CH3 wagging and C–C stretching of aromatic ring (proteins—actin)

1,455 C–H bending mode (proteins); CH2 and CH3 deformations—bending and scissoring (lipids/phospholipids and proteins)

1,498 C=C stretching in benzenoid ring (proteins)

1,519 C–C stretch mode and C=C stretching (carotenoids)

1,568* Red blood cells (heme group); C–N and N–H modes (proteins)

1,610* Red blood cells (heme group); C=O stretching and C=C stretching and bending modes (proteins); NH2 (proteins)

1,659 C=O stretching (proteins and lipids); C=C stretching (lipids); nucleic acids; H2O intermolecular bending

Lasers Med Sci

and Gleason ≤7 groups to the Gleason >7 group (t test,p<0.05) have been found, indicating that these compoundshave a discrimination capability only for lesions of higherGleason scores. Figure 4 presents the scatter plot of theRaman concentration C for the two biochemicals that present-ed the highest discrimination capability (lower error rate)using the Euclidean distance as a discriminator, being thephosphatidylcholine and water, and also the discriminationline based on the mean Euclidean distance for separating thegroups benign and PrCa (black line) and benign, Gleasonscore ≤7 and Gleason score >7 (red lines). The results of theclassification in each histopathological group are summarizedin Table 2. This spectral model showed sensitivity and

specificity of 75 and 74 %, respectively, for discriminatingbenign from PrCa, with an overall accuracy of 74 %. Fordiscriminating benign, Gleason ≤7, and Gleason >7, the mod-el resulted in sensitivity of 64, 35, and 69%, and specificity of79, 77, and 89 %, respectively, with an overall accuracy of57 %. Higher sensitivity and specificity were obtained fordiscriminating Gleason >7.

The high autofluorescence found in most samples could beresponsible in some extent to the lower signal-to-noise ratio ofthe Raman spectra compared to other works [23, 32, 33] andthe higher error bars of the spectral model. This high fluores-cence background was not reported before and was expectedin epithelial tissues with the use of 830-nm excitation, so the

Fig. 2 Raman spectra of the biochemicals used in the spectral model, which appeared with peaks in the same positions as the tissue spectra. The verticaldotted lines indicate the peak positions of Raman bands of interest in the prostate tissues

Lasers Med Sci

use of longer wavelengths (1,064-nm excitation with the newInGaAs detectors) could bring better classification.

Although considered to have a relatively good prognosis ifdiagnosed opportunely, the PrCa remains the most prevalenttype of cancer in men and continues being a serious healthproblem worldwide [1, 5]. Currently available screeningmethods such as PSA testing have not been found to besuccessful in reducing mortality and, moreover, lead to manyunnecessary surgeries, resulting in financial losses and re-duced quality of life [7, 8]. It is clear that there is a continuedneed for investment in the development of technology appliedin early diagnosis and treatment of PrCa.

Recently, several studies have demonstrated that Ramanspectroscopy could be used to differentiate prostate diseasesfrom normal tissues in vitro [23, 32, 33, 35–38], using dis-criminant analysis [23, 32, 36, 37]. Those works aimed theidentification, through Raman spectroscopy, of main bio-chemicals presented in both normal and cancerous tissuesmacroscopically, aiming tissue diagnosis [23, 32, 33, 35,

36], and the identification of the biochemical alterations inmalignant cells microscopically, aiming disease grading [37,38, 41]. Raman spectroscopy has been capable of identifyingendogenous biological markers related to diseases such asproteins/amino acids, carbohydrates, lipids/phospholipids,and nucleic acids [23, 33, 42].

The discrimination of tissues using the estimated Ramanconcentration and the Euclidean distance showed to be a veryuseful way to find the spectral variables that are common to aparticular group and differs from the other groups, because thedifferences in the tissue biochemistry associated to the dis-eases can be easily identified and quantified using a suitablespectral model [22–24, 29]. Changes in the molecular constit-uents that are associated with the disease led to a few butsignificant differences in the spectra [33, 38]. This workshowed the major differences between N and PrCa spectrain vitro being related to vibrational bands of proteins, lipids/phospholipids, blood, and water. The spectral features ofcollagen were obtained and included in the model. Theyresulted in negative Raman estimators, so they were notincluded in the final model.

Fig. 3 Plot of the relativecontribution of selected basalcompounds estimated by theRaman spectral model (mean±standard deviation) applied to thein vitro prostate spectra of benignand PrCa in two ranges ofGleason scores. Asterisk indicatesstatistically significant differencesbetween labeled groups, and nsindicates not significant(ANOVA, p<0.05)

Fig. 4 Scatter plot of the relative contribution of phosphatidylcholineand water found by the spectral model for the prostate tissues. Discrim-ination lineswere based on the Euclidean (mean) distance among groups:black line separates benign and PrCa, and red lines separate benign,Gleason ≤7, and Gleason >7

Table 2 Results of the discriminating model in terms of correct classifi-cation of prostate spectra for each histopathological group

Spectral model

Histopathology Benign PrCa

Benign (n=77) 57 20

PrCa (n=75) 19 56

Overall accuracy 74 %

Spectral model

Histopathology Benign Gleason ≤7 Gleason >7

Benign (n=77) 49 16 12

Gleason ≤7 (n=40) 15 14 11

Gleason >7 (n=35) 1 10 24

Overall accuracy 57 %

Lasers Med Sci

In this study, highest sensitivity and specificity valuescould be achieved using phospholipid (phosphatidylcholine)and water features to distinguish PrCa with Gleason score >7.The differences in the hydration of prostate tissues can berelated to the tubular structure of the peripheral zone, which inhealthy prostate tissue allows extensive diffusion of watermolecules within the gland tubules. Cancer tissue destroysthe normal glandular structure of the prostate and replacesducts [6]. It also has a higher cellular density than does healthyprostate peripheral zone tissue [43]. It has been found thatactin was increased in benign and Gleason ≤7 and decreasedin Gleason >7. As Gleason scores increase, the amount ofepithelial cells increases (higher cellular density), and theamount of stroma decreases. Thus, the decreasing of actincould be explained by the occupation of the fibromuscularspace by the cancer cells, also diminishing the vascularization.The increased amount of elastin in Gleason >7 tumors couldbe a signal of higher vascular formation.

Phosphatidylcholine increased following the Gleasonscore. Choline is a biomarker in proton resonance imaging,and choline compounds are often increased in prostate cancertissue [44]. Choline compounds are involved in the biosyn-thesis and degradation of phospholipids, which are requiredfor the buildup and maintenance of cell membranes. Anincreased cell turnover in prostate cancer results in an in-creased concentration of choline-containing molecules withinthe cytosol and the prostate interstitial tissue [6]. Stone et al.[23] found that actin was not different from tumor and non-tumor, but choline and triolein increased from benign to PrCa,and triolein was even more intense for the Gleason >7. Interms of tissue histochemistry, studies suggest that the growthof certain tumors is dependent on maintaining sufficient lipidlevels and that the lipid-mobilizing effect of the tumor may benecessary to sustain tumor growth [45].

The biochemical model was implemented using the Ramanspectra of commercially available biochemical compoundsfitted to the spectra of in vitro fragments of human tissues.This model has simple implementation, as new tissue bio-chemicals and new tissue samples could be added anytimewithout the need for reprocessing of all tissue spectra; newspectra could be included for statistical purposes without theneed for model rebuilt as needed for models based on princi-pal component analysis [37], since the Raman spectra ofbiochemicals are already the basis of the tissue spectral infor-mation [23]. One should be aware of including spectra oftissue compounds or structures which have spectral informa-tion with orthogonal (non-similar) information, in order toavoid cross correlation [23].

Despite small spectral differences between healthy andneoplastic tissues, it has been verified that one can accessbiochemical information regarding the composition of pros-tate tissue on site nondestructively, and such information isemployed in tissue discrimination of prostate tissues with

higher Gleason scores. Since biochemical changes precedingand accompanyingmorphological tissue changes are extreme-ly complex [46], spectral analysis could help extracting infor-mation tissues and would provide as much information aspossible to facilitate the most accurate prediction of histopa-thology using a few spectral components [21]. Since waterwas important in tissue diagnosis, the use of higher wavenum-ber Raman spectroscopy would bring some insight in thewater features at 3,400 cm−1 (O–H stretching) [47]. In urolo-gy, the current status of the Raman instrumentation does notpermit noninvasive measurements due to tissue localization;nevertheless, fiber optics incorporated to injection needles[48] and flexible side-viewing fibers [49] have been demon-strated for collecting Raman spectra in human tissues aimingin vivo diagnostic applications in solid and hollow tissues. Atthe present time, Raman diagnosis in ex vivo biopsy frag-ments would ultimately bring rapidness to the tissue evalua-tion at the bench, aiming histopathological diagnosis (Gleasonscoring) and tumor margin detection, since the measurementcould be done in real time in several spots.

Conclusions

The results demonstrated that dispersive Raman spectroscopycould be employed to help reveal prostate biochemistry and beused to discriminate between benign and prostate cancerin vitro. Despite small spectral differences between healthyand neoplastic prostate, tissue biochemistry is altered in thehigh-grade lesions, and a spectral model based on the least-squares fitting of selected biochemicals revealed that relativeconcentration of actin, elastin, phosphatidylcholine, blood,and water was most relevant to discriminate prostate withGleason score >7. A discrimination model based onEuclidean distance using the relative concentrations of phos-phatidylcholine and water showed sensitivity of 75 % andspecificity of 74 % to discriminate prostate cancer and sensi-tivity of 69 % and specificity of 89 % to discriminate Gleasonscore >7. Raman spectroscopy is an analytical technique withpossibility for identifying biochemical constitution of pros-tate, providing an insight into the molecular changes of neo-plasias while being used to discriminate malignant from be-nign tissues.

Acknowledgments L. Silveira Jr. thanks FAPESP (São Paulo ResearchFoundation) for the financial support (process nos. 2009/01788-5 and2012/20666-0).

References

1. Jemal A, Siegel R, Ward E, Murr T, Xu J, Thun MJ (2007) Cancerstatistics, 2007. CA Cancer J Clin 57(1):43–66

Lasers Med Sci

2. Zeigler-Johnson CM, Rennert H, Mittal RD, Jalloh M, Sachdeva R,Malkowicz SB, Mandhani A, Mittal B, Gueye SM, Rebbeck TR(2008) Evaluation of prostate cancer characteristics in four popula-tions worldwide. Can J Urol 15(3):4056–4064

3. Zerbib M, Zelefsky MJ, Higan CS, Carroll PR (2008) Conventionaltreatments of localized prostate cancer. Urology 72(Suppl 6):S25–S35

4. Howlader N, Noone AM, Krapcho M, Neyman N, Aminou R,Altekruse SF, Kosary CL, Ruhl J, Tatalovich Z, Cho H, MariottoA, Eisner MP, Lewis DR, Chen HS, Feuer EJ, Cronin KA (2012)SEER cancer statistics review, 1975-2009 (vintage 2009 popula-tions). National Cancer Institute. http://seer.cancer.gov/csr/1975_2009_pops09. Accessed 09 January 2012

5. Brasil. Ministério da Saúde. General Office of Attention to theHealth. Coordination of Prevention and Surveillance (2011)Estimate/2012—Cancer incidence in Brazil. National Institute ofCancer—INCA. http://www.inca.gov.br/estimativa/2012/estimativa20122111.pdf. Accessed 09 January 2012

6. Hoeks CMA, Barentsz JO, Hambrock T, Yakar D, Somford DM,Heijmink SWTPJ, Vos TWJ, Scheenen PC, Huisman H, van OortIM, Witjes JA, Heerschap A, Fütterer JJ (2011) Prostate cancer:multiparametric MR imaging for detection, localization, and staging.Radiology 261(1):46–66

7. Philip J, Manikandan R, Javlé P, Foster CS (2009) Prostate cancerdiagnosis: should patients with prostate specific antigen >10 ng/mLhave stratified prostate biopsy protocols? Cancer Detect Prev 32(4):314–318

8. Albertsen PC (2009) The treatment paradigm shifts again on prostatecancer. Eur Urol 55(1):9–11

9. Bostwick DG, Foster CS (1997) Evaluating radical prostatectomyspecimens: therapeutic and prognostic importance. Virchows Arch430(1):1–16

10. Beneduce L, Prayer-Galetti T, Giustinian AM, Gallotta A, Betto G,Pagano F, Fassina G (2007) Detection of prostate-specific antigencoupled to immunoglobulin M in prostate cancer patients. CancerDetect Prev 31(5):402–407

11. Epstein JI, Allsbrook WC, Amin MB, Egevad LL, GradingCommittee ISUP (2005) The 2005 International Society ofUrological Pathology (ISUP) Consensus Conference on GleasonGrading of Prostatic Carcinoma. Am J Surg Pathol 29(9):1228–1242

12. Leite KRM, Camara-Lopes LH, Dall'OglioMF, Cury J, Antunes AA,Sañudo A, Srougi M (2009) Upgrading the Gleason score in extend-ed prostate biopsy: implications for treatment choice. Int J RadiatOncol Biol Phys 73(2):353–356

13. Allsbrook WC Jr, Mangold KA, Allsbrook WC, Johnson MH, LaneRB, Lane CG, Amin MB, Bostwick DG, Humphrey PA, Jones EC,Reuter VE, SakrW, Sesterhenn IA, Troncoso P,Wheeler TM, EpsteinJI (2001) Interobserver reproducibility of Gleason grading of pros-tatic carcinoma: urologic pathologists. Hum Pathol 32(1):74–80

14. Murphy AM, McKiernan JM, Olsson CA (2004) Controversies inprostate cancer screening. J Urol 172(5 pt 1):1822–1824

15. Silveira L, Silveira FL, Bodanese B, Zângaro RA, Pacheco MTT(2012) Discriminating model for diagnosis of basal cell carcinomaand melanoma in vitro based on the Raman spectra of selectedbiochemicals. J Biomed Opt 17(7):077003, Erratum in: J BiomedOpt 18(3):039801 (2013)

16. Oliveira FSS, Giana HE, Silveira L (2012) Discrimination of selectedspecies of pathogenic bacteria using near-infrared Raman spectros-copy and principal components analysis. J Biomed Opt 17(10):107004

17. Das K, Stone N, Kendall C, Fowler C, Christie-Brown J (2006)Raman spectroscopy of parathyroid tissue pathology. Lasers MedSci 21(4):192–197

18. Lattermann A, Matthäus C, Bergner N, Beleites C, Romeike BF,Krafft C, Brehm BR, Popp J (2013) Characterization of atheroscle-rotic plaque depositions by Raman and FTIR imaging. JBiophotonics 6(1):110–121

19. Krafft C, Belay B, Bergner N, Romeike BF, Reichart R, KalffR, Popp J (2012) Advances in optical biopsy-correlation ofmalignancy and cell density of primary brain tumors usingRaman microspectroscopic imaging. Analyst 137(23):5533–5537

20. Souza RA, Xavier M, Silva FF, Souza MT, Tosato MG, Martin AA,Castilho JCM, Ribeiro W, Silveira L (2012) Influence of creatinesupplementation on bone quality in the ovariectomized rat model: anFT-Raman spectroscopy study. Lasers Med Sci 27(2):487–495

21. Hanlon EB, Manoharan R, Koo TW, Shafer KE, Motz JT,Fitzmaurice M, Kramer JR, Itzkan I, Dasari RR, Feld MS (2000)Prospects for in vivo Raman spectroscopy. PhysMed Biol 45(2):R1–R59

22. Haka AS, Shafer-Peltier KE, Fitzmaurice M, Crowe J, Dasari RR,Feld MS (2005) Diagnosing breast cancer by using Raman spectros-copy. Proc Natl Acad Sci U S A 102(35):12371–12376

23. Stone N, Prieto MCH, Crow P, Uff J, Ritchie AW (2007) The use ofRaman spectroscopy to provide an estimation of the gross biochem-istry associated with urological pathologies. Anal Bioanal Chem387(5):1657–1668

24. Buschman HP, Motz JT, Deinum G, Romer TJ, Fitzmaurice M,Kramer JR, van der Laarse A, Bruschke AV, Feld MS (2001)Diagnosis of human coronary atherosclerosis by morphology-basedRaman spectroscopy. Cardiovasc Pathol 10(2):59–68

25. Nogueira GV, Silveira L, Martin AA, Zângaro RA, Pacheco MT,Chavantes MC, Pasqualucci CA (2005) Raman spectroscopy studyof atherosclerosis in human carotid artery. J Biomed Opt 10(3):031117

26. Pichardo-Molina JL, Frausto-Reyes C, Barbosa-García O, Huerta-Franco R, González-Trujillo JL, Ramírez-Alvarado CA, Gutiérrez-Juárez G, Medina-Gutiérrez C (2007) Raman spectroscopy and mul-tivariate analysis of serum samples from breast cancer patients.Lasers Med Sci 22(4):229–236

27. Rossi EE, Silveira L, Pinheiro ALB, Zamuner SR, Aimbire F, MaiaM, Pacheco MTT (2010) Raman spectroscopy for differential diag-nosis of endophthalmitis and uveitis in rabbit iris in vitro. Exp EyeRes 91(3):362–368

28. Lieber CA, Majumder SK, Ellis DL, Billheimer DD, Mahadevan-Jansen A (2008) In vivo nonmelanoma skin cancer diagnosis usingRaman microspectroscopy. Lasers Surg Med 40(7):461–467

29. Huang Z, Teh SK, ZhengW, Lin K, Ho KY, TehM, Yeoh KG (2010)In vivo detection of epithelial neoplasia in the stomach using image-guided Raman endoscopy. Biosens Bioelectron 26(2):383–389

30. Stone N, Stavroulaki P, Kendall C, Birchall M, Barr H (2000) Ramanspectroscopy for early detection of laryngeal malignancy: prelimi-nary results. Laryngoscope 110(10 pt 1):1756–1763

31. Widjaja E, Zheng W, Huang Z (2008) Classification of colonictissues using near-infrared Raman spectroscopy and support vectormachines. Int J Oncol 32(3):653–662

32. Crow P,MolckovskyA, StoneN, Uff J,Wilson B,Wongkeesong LM(2005) Assessment of fiberoptic near-infrared Raman spectroscopyfor diagnosis of bladder and prostate cancer. Urology 65(6):1126–1130

33. Stone N, Kendall C, Smith J, Crow P, Barr H (2004) Raman spec-troscopy for identification of epithelial cancers. Faraday Discuss 126:141–157

34. Moreira LM, Silveira L, Santos FV, Lyon JP, Rocha R, Zângaro RA,Villaverde AB, Pacheco MTT (2008) Raman spectroscopy: a pow-erful technique for biochemical analysis and diagnosis. SpectroscopyInt J 22(1):1–19

35. Stone N, Kendall C, Shepherd N, Crow P, Barr H (2002) Near-infrared Raman spectroscopy for the classification of epithelial pre-cancers and cancers. J Raman Spectrosc 33(7):564–573

36. Crow P, Stone N, Kendall CA, Uff JS, Farmer JA, Barr H (2003) Theuse of Raman spectroscopy to identify and grade prostatic adenocar-cinoma in vitro. Br J Cancer 89(1):106–108

Lasers Med Sci

37. Lopes RM, Silveira L, Silva MASR, Leite KRM, Pasqualucci CA,Pacheco MTT (2011) Diagnostic model based on Raman spectra ofnormal, hyperplasia and prostate adenocarcinoma tissues in vitro.Spectrosc Int J 25(2):89–102

38. Crow P, Barrass B, Kendall C, Hart-Prieto M, Wright M, Persad R,Stone N (2005) The use of Raman spectroscopy to differentiatebetween different prostatic adenocarcinoma cell lines. Br J Cancer92(12):2166–2170

39. Svensson T, Andersson-Engels S, Einarsdóttír M, Svanberg K (2007)In vivo optical characterization of human prostate tissue using near-infrared time-resolved spectroscopy. J Biomed Opt 12(1):014022

40. Moore CM, Mosse CA, Allen C, Payne H, Emberton M, Bown SG(2011) Light penetration in the human prostate: a whole prostateclinical study at 763 nm. J Biomed Opt 16(1):015003

41. Crow P, Uff JS, Farmer JA, Wright MP, Stone N (2004) The use ofRaman spectroscopy to identify and characterize transitional cellcarcinoma in vitro. BJU Int 93(9):1232–1236

42. Kast RE, Serhatkulu GK, Cao A, Pandya AK, Dai H, Thakur JS,Naik VM, Naik R, Klein MD, Auner GW, Rabah R (2008) Ramanspectroscopy can differentiate malignant tumors from normal breasttissue and detect early neoplastic changes in a mouse model.Biopolymers 89(3):235–241

43. Somford DM, Fütterer JJ, Hambrock T, Barentsz JO (2008) Diffusionand perfusionMR imaging of the prostate.MagnReson Imaging ClinN Am 16(4):685–695

44. Cornel EB, Smits GA, Oosterhof GO, Karthaus HF, Deburyne FM,Schalken JA, Heerschap A (1993) Characterization of human pros-tate cancer, benign prostatic hyperplasia and normal prostate byin vitro 1H and 31P magnetic resonance spectroscopy. J Urol150(6):2019–2024

45. Mulligan HD, Tisdale MJ (1991) Effect of the lipid-lowering agentbezafibrate on tumor growth rate in vivo. Br J Cancer 64(6):1035–1038

46. Bird B, Miljkovic M, Romeo MJ, Smith J, Stone N, George MW,DiemM (2008) Infrared micro-spectral imaging: distinction of tissuetypes in axillary lymph node histology. BMC Clin Pathol 8:8. doi:10.1186/1472-6890-8-8

47. Duraipandian S, Zheng W, Ng J, Low JJ, Ilancheran A, Huang Z(2012) Simultaneous fingerprint and high-wavenumber confocalRaman spectroscopy enhances early detection of cervical precancerin vivo. Anal Chem 84(14):5913–5919

48. Day JC, Stone N (2013) A subcutaneous Raman needle probe. ApplSpectrosc 67(3):349–354

49. Lima CJ, Sathaiah S, Pacheco MTT, Zângaro RA, Manoharan R(2004) Side-viewing fiberoptic catheter for biospectroscopy applica-tions. Lasers Med Sci 19(1):15–20

50. Movasaghi Z, Rehman S, Rehman IU (2007) Raman spectroscopy ofbiological tissues. Appl Spectrosc Rev 42(5):493–541

51. Bankapur A, Zachariah E, Chidangil S, Valiathan M, Mathur D(2010) Raman tweezers spectroscopy of live, single red and whiteblood cells. PLoS ONE 5(4):e10427

Lasers Med Sci


Recommended