Cervical Smear Photodiagnosis by Fluorescence

Abstract

Objective: Fluorescence is widely investigated and can characterize general changes occurring in the state of the cell and the tissue during physiological and/or pathological processes, but the method has not been adopted for clinical use. We present a photodiagnostic device and a relevant evaluation method fitted to classical screening procedures. This method is based on the discovery of smear content intrinsic fluorescent markers. Background data: Meaningful spectral components of cervical smear samples differ from those measured in the entire live cervix. This article deals with the identification of changes in smear spectra in cervicitis and CIN2+ (cervical intraepithelial neoplasia of the second degree or higher) at 355 nm excitation. Methods: Methods used in the study: microlaser-induced fluorescence spectroscopy of liquid cytology samples and histological evaluation of the biopsies of the same cervix (and/or only cytological evaluation) was performed for 78 cases. The fluorescence spectra of cervical cytology supernatant sediment were approximated with several Gaussian components. The ratios of the area under each Gaussian component to the whole area under the experimental curve were calculated and compared among histological groups by using the Mann–Whitney test and receiver operating characteristic (ROC) analysis. Results: The results of this study are a concise summary of the essential features verified by the data: the spectral regions 402–416 nm and 424–438 nm are important for the discrimination of both normal and CIN2+ groups (in terms of sensitivity, specificity, and positive predictive value). The spectral regions 480–515 nm and 595–625 nm are important for the identification of cervicitis. Conclusions: Cervical smear autofluorescence diagnostics could be useful for cancer screening at the point of care, in a simple cytology laboratory, and for the monitoring of treatment. We suggest possible fluorophores in the smear content.

Introduction

C ervical cancer is one of the most common forms of cancer in women. Although prevention of this disease by cytological smear microscopy has been very successful, the test does give a significant proportion of false conclusions. Sensitivity of conventional cytology ranges from 50% to 80%, and the specificity ranges from 70% to 90%.¹ In recent years, many different methods have been used to improve the diagnostics of pre-cancer. However, extensive application of immunohistochemical markers in the cytological laboratory or at the point of care is still rather problematic.^2,3 Global trends in the development of relevant devices demand that equipment be able to catch and interpret natural molecular signals at the point of care.^4,5 Great success was achieved by applying optical methods for the investigation of live tissue conditions, the gathering and treatment of molecular spectra.^6

–10 Imaging techniques for the selection of a proper biopsy site are complicated; therefore, it would be reasonable to apply them in case of accurately predicted pathology. Cell absorbance evaluation based on vibrational spectroscopic techniques can elucidate cellular differences originating from proteins, nucleic acids, and lipids as a support to current methods for cervical cancer screening.¹¹ Smear Fourier transform infrared (FTIR) absorption spectroscopy is proposed as an alternative method for cervical cancer screening.¹²

Fluorescence has been widely investigated. A review¹³ describes general changes occurring in the state of cell and tissue during physiological and/or pathological processes. They produce modifications in the amount and distribution of endogenous fluorophores as well as in the chemical–physical properties of their microenvironment. Elastin,¹⁴ collagen VI,¹⁵ collagen VII,^16,17 flavins,¹³ lysophospholipids,¹⁸ and lipofuscin¹⁹ are the molecules containing fluorescent groups which have been confirmed as key players in pathological processes including cancerogenesis and inflammation. Most of the fluorescent spectral properties of these products of cell activity are well known.²⁰

Our experience and that of other researchers has confirmed that ultraviolet excitation-based diagnostics, which exploit live cervical tissue fluorescence spectra, are complicated because of the overlap of fluorescence from the epithelium and the stroma.^10,21

–24 Content of the cervical smear exhibits strong visible fluorescence derived from cervical epithelium.^25,26 Compared with other optical techniques, the fluorescence detection-based equipment is less complicated. The aim of this work was to study the value of fluorescence-based photodiagnosis for cervical pre-cancer screening at the point of care. The main task of the study was to discover the spectral regions where the differences between pathology groups are most expressed and suitable for the creation of a diagnostic algorithm.

Methods

Samples and setup

With the permission (N61) of the Lithuanian Bioethics Committee and informed patient signed consent, 78 samples of normal and high-grade squamous intraepithelial lesion (HSIL) by cytology were investigated: 38 samples constituted a “normal” group, and 40 samples constituted an HSIL group. HSILs were confirmed by histology as a CIN2+ (precancerous state, second degree or higher cervical intraepithelial neoplasia) group of 28 samples, and a cervicitis group of 12 samples; 1 sample of CIN1 (first degree cervical intraepithelial neoplasia) did not allow us to form a separate group. Diagnosis of the “normal” group was confirmed only by a cytological test, as a biopsy is not performed unless a cytological test has indicated pathology. Sediments of the liquid cervical smear supernatant dried on a quartz plate were used as the specimens for this investigation.

The fluorescence of a sample was excited by the third harmonics (wavelength 355 nm) of sub-nanosecond (<1 ns), high pulse repetition rate (10 kHz), single longitudinal mode passive Q-switched Nd:YAG microlaser STA-01-TH (Standa, Ltd.). A UFS-1 filter was used for the elimination of the first and the second harmonics. A fiberoptic reflection probe was used to excite and to collect the fluorescence of samples. The fluorescence signal was registered in three locations of each sample by Ava-Spec-2048 spectrometer.

Evaluation of spectra

Measurement of the fluorescence of samples deposited on a substrate requires accurate removal of the substrate contribution. It was found that the sample fluorescence was absent in a region marked “a” in Fig. 1 (>800 nm). The removal of the quartz fluorescence contribution was based on the evaluation of the magnitude of quartz signal according to the specific quartz fluorescence peak shown in Fig. 1 as a “b” range. In order to get the specimen spectrum, the fluorescence of quartz was normalized multiplying it by the factor obtained as a ratio of peak intensity in the sample to peak intensity in quartz and subtracted from the sample spectrum. The residual signal in the region >800 nm was accepted as a noise level. The total noise of the fluorescence spectrum was reduced by averaging up to 25 adjacent points.

FIG. 1.

Spectra of (1) a substrate response and (2) a specimen. Spectrum of sample (3) was obtained by subtracting the spectrum of substrate. “a” and “b” mark the regions used for the substrate spectrum normalization. A peak at 810 nm is the scattered light of the LEDs that pump the laser.

Data approximation

The fluorescence spectra of biomedical samples are complex as a result of the existence of different fluorophors in the tissue; therefore, different methods have been used for their analysis. We use an approximation to divide the spectrum into a series of Gaussian components.²⁷ This allowed us to follow the changes of the spectrum in its definite regions and to use a standard procedure of analysis. The spectral components were characterized using the parameters of Gaussian curves: A_j – total area under the Gaussian curve from the baseline, w_j – width of the peak at a half height, x_j – position of the peak maximum where j indicates the j ^th Gaussian curve. This approximation is quite formal. However, the positive results speak in favor of the use of this method, although more detailed investigation into the nature of the components is still necessary. By means of Gaussian curves, two steps of the analysis of experimental spectra were performed, and each of them had certain preferences.

Method I

The division of the spectrum into components was performed by the method of multi-peaks analysis (Origin software version 7, OriginLab Corporation). To ensure equal conditions for all cases, a maximum number of eight Gaussian curves was chosen. After the approximation of all spectra, Gaussian curves were grouped into components according to the peak center positions (xj). An example of a sample spectrum with Gaussian components is presented in Fig. 2.

FIG. 2.

The example of experimental spectra approximated with a set of Gaussian components using method I. Black line, experimental data; solid grey line, the approximation; dashed lines, the Gaussian components.

Method II

Based on the results of method I, this method also involves curve fitting. It was evident that some distributions of the parameters of fitting curves exhibited a doublet structure. Limits of the variation of parameters (A_j, w_j, x_j) were introduced and optimal parameters (initial values and the boundaries) were found. Boundaries within which the parameters of Gaussian curves were allowed to vary were introduced into method II. Optimization was conducted on the basis of the assumption of 12 components, and initial parameter values and boundary positions were manipulated during roughly 25 approximations. The following optimization criterion was to be met: a bell-shaped distribution of parameters w_j and x_j found through the curve fitting, which involves the manipulation of boundaries and initial parameters. Figure 3 demonstrates the distribution of peak values of the components. Exceptions were made for the centers of three components with non-bell-shaped distributions: at 595–625 nm (probably this region could have been separated into two components, but we did not have enough data for the decision), 702–745 nm, and 745–770 nm (Fig. 3). These components mostly had small amplitude, their areas under the components curve in respect to the area under the whole experimental curve were small, and the statistical peak center was not clearly expressed. However, these components were important to ensure optimal parameters of other components.

FIG. 3.

The distribution of Gaussian components' peak values obtained after optimization procedure in method II. Relative frequency is given to each component. The values of peak centers are grouped at intervals of 2 nm. The positions of components' peak centers are in a range: (1) 380–400 nm, (2) 402–416 nm, (3) 424–438 nm, (4) 446–457 nm, (5) 466–477 nm, (6) 478–494 nm, (7) 503–550 nm, (8) 595–625 nm, (9) 655–672 nm, (10) 680–696 nm, (11) 705–742 nm, (12) 745–770 nm. The center position distributions of components 1–7, 9, 10 are bell shaped. The distributions of components 8, 11, 12 are not so clearly expressed, and the lines are a guide for the eyes.

The ratio of spectral components' area to the area under the whole experimental curve (RA) was established for every component. As mentioned previously, the spectra from three points of one sample were registered and RAj values (here j indicates the j ^th component) from the same specimen were averaged before performing the Mann–Whitney test. If the fluorescence intensity in one or two points (out of three) was low and the noise was strong, these points were excluded from further analysis, and the average of RAj was calculated from two points or one point, respectively. Low fluorescence intensity was observed in nine cases.

A standard linear correlation analysis (Pearson's correlation) was performed for the averaged RA_j values of components in method II. In cases in which the Pearson's correlation coefficient was higher than zero, the RA_j values of those components were summed. These sums were called “complex components.” If the correlation coefficient was positive, higher values were summed with higher ones, and lower values with lower ones, so that the differences would become more obvious.

Every component of histological/cytological groups (CIN2+ vs. cervicitis; CIN2+ vs. normal and cervicitis vs. normal) as well as combined groups (CIN2+ and cervicitis as positive vs. normal; cervicitis and normal as negative vs. CIN2+; CIN2+ and normal as negative vs. cervicitis) were compared using the Mann–Whitney test, with confidence level α=0.05. The parameter compared is the averaged RAj values. For the cases in which the Mann–Whitney test showed statistically reliable differences between compared groups, receiver operating characteristic (ROC) curves were plotted and area under the curve (AUC) values were calculated using the XLSTAT 2009.5.01 program (Addinsoft SARL). The Youden index J²⁸ was used to set a statistical cutoff point of the RA_j value for the determination of optimal value of sensitivity (Se) and specificity (Sp) for making a decision. J equals to a maximal value of (Se+Sp-1).

Results

Results of the statistical analysis are presented in Tables 1 and 2. Some general comments can be made for each of the methods used.

Table 1.

Results of ROC Analysis, Comparing the Best AUC Values for Each Group; the Group Considered as a Positive Case is Underlined

	Component
	No	nm	Compared groups	AUC	Se	Sp	ACC	PPV	NPV
Method I	2	402–420	Normal vs. CIN2+	0.74	0.90	0.61	0.78	0.77	0.81
	8	680–750	Normal vs. others	0.70	0.65	0.75	0.70	0.72	0.68
	2	402–420	CIN2+ vs. others	0.69	0.61	0.81	0.74	0.63	0.79
	5	480–515	Cervicitis vs. others	0.68	0.67	0.72	0.71	0.30	0.93
	2	402–420	Cervicitis vs. CIN2+	0.53	0.42	0.82	0.70	0.50	0.77
Method II	2	402–416	Normal vs. CIN2+	0.82	0.86	0.66	0.74	0.65	0.86
	2	402–416	Normal vs. others	0.78	0.80	0.66	0.73	0.71	0.76
	2	402–416	CIN2+ vs. others	0.77	0.86	0.60	0.69	0.55	0.88
	8	595–625	Normal vs. Cervicitis	0.72	0.42	1.00	0.86	1.00	0.84

ROC, receiver operating characteristic; AUC, area under the curve; Se, sensitivity; Sp, specificity; ACC, accuracy; PPV, positive predictive value; NPV, negative predictive value; RA_i, relative area of the j ^th component, where “j” is indicated in column “No”.

Table 2.

Results of ROC Analysis, Comparing the Best PPV Values for Each Group; Case Considered as Positive is Underlined

	Component
	No	nm	Compared groups	AUC	Se	Sp	ACC	PPV	NPV
Method I	2	402–420	Normal vs. CIN2+	0.74	0.90	0.61	0.78	0.77	0.81
	7	650–680	CIN2+ vs. others	0.64	0.81	0.46	0.69	0.74	0.57
	8	680–750	Normal vs. others	0.70	0.65	0.75	0.70	0.72	0.68
	2	402–420	Cervicitis vs. CIN2+	0.53	0.42	0.82	0.70	0.50	0.77
	5	480–515	Cervicitis vs. others	0.68	0.67	0.72	0.71	0.30	0.93
Method II	8	595–625	Normal vs. Cervicitis	0.72	0.42	1.00	0.86	1.00	0.84
	(2,3)	402–416; 424–438	Normal vs. others	0.74	0.43	0.97	0.69	0.94	0.62
	(2,3)	402–416; 424–438	Normal vs. CIN2+	0.77	0.50	0.97	0.77	0.93	0.73
	(2,3)	402–416; 424–438	CIN2+ vs. others	0.74	0.50	0.94	0.78	0.82	0.77

ROC, receiver operating characteristic; AUC, area under the curve; Se, sensitivity; Sp, specificity; ACC, accuracy; PPV, positive predictive value; NPV, negative predictive value. Component number (2,3) means that the RA of component j=2 and j=3 were summed.

Method I

Eight Gaussian components were used in this approximation. The Mann–Whitney test showed 13 cases in which RAs of cytological/histological groups differed significantly. ROC analysis revealed that the maximum AUC and positive predictive value (PPV) were achieved comparing normal versus CIN2+ using the component at 402–420 nm (AUC=0.74, se=90.0%, sp=60.7%, index J=77.9%; PPV=76.6%, negative predictive value [NPV]=81.0%). The lowest AUC=0.527 was found comparing CIN2+ versus cervicitis. The lowest PPV=29.6% and the highest NPV=92.5% were reached following the comparison cervicitis versus others using the 480–515 nm component.

Method II

This approximation was performed with 12 Gaussian components. RAs of cytological/histological groups differed significantly in 25 cases according to the Mann–Whitney test. The maximum AUC value was reached in comparing normal versus CIN2+ using the component at 402–416 nm (AUC=0.82, se=85.7%, sp=65.8%, index J=74.2%). If normal groups versus cervicitis were compared and the component at 595–625 nm was used, the maximal values of specificity and accuracy were reached (AUC=0.72, se=41.7%, sp=100%, index J=86.0%). High accuracy was reached in comparing CIN2+ to others using a complex component, which is the RA sum of components at 402–416 nm and 424–438 nm (AUC=0.74, se=50.0%, sp=94.0%, index J=78.2%). The highest PPV was reached by comparing normal versus cervicitis using the 595–625 nm component (PPV=100%, NPV=84.4%). Comparing normal to others and CIN2+ to normal using the complex component 402–416 nm and 424–438 nm, the PPV >90 was reached, and the best AUC (0.82) and the best PPV (100%) values were achieved by applying method II. Method I can give higher sensitivity and method II can give higher specificity. Method I requires experienced personnel to perform the multipeak analysis because of the necessity of manual setting of the initial data. Method II is more automatic.

Discussion

The success of this investigation into the liquid cervical smear supernatant sediment in comparison with earlier works in which a live cervix surface was investigated, may be accounted for by the principle of investigation. The smear consists of all products produced by the processes taking place in the cervix epithelium; therefore, the investigation eliminates the influence of stromal and blood vessel collagen and elastin, which contribute to the cervix's subepithelial autofluorescence. In addition, cell viability-related fluctuations are eliminated in a fixed smear. It summarizes interfering fluorophore information from the squamous, glandular, transition zone (TZ) epithelium; therefore, it eliminates diagnostic problems related to the transformation zone complex anatomy.^21,22,24,29 The results presented earlier demonstrate the differences between normal, cervicitis, and CIN2+ groups that have been established by detecting the main differences in spectral regions at 402–420 nm, 480–515 nm, and 595–625 nm, and also by cooperative response of contributions in the combination of regions 402–416 nm and 424–438 nm. These regions narrow down the spectral regions used for diagnostics in previous works^30,31 that point out the overlap of different fluorophores' contribution,^18,20 and the resulting narrower regions reveal the differences of the previously mentioned contributions better.

Similarly, Millot et al.³² used confocal fluorescent microscopy involving a 363 nm laser beam for malpighian epithelium cells cytoplasm prepared by Thin-Prep technique. Instead of the whole emission spectrum, the dual wavelength emission images at 425 and 525 nm were compared, and it caused an increase in the analyzed cell number and improve the classification.

In the present study, the region of 355 nm/395 nm (where collagen I fluorescence is maximal according to the excitation/emission matrice) was not found to be important for diagnostics because spectra started to be analyzed at 376 nm, and a contribution of the substrate signal was big in this region (Fig. 1). We can suggested that collagen VII and elastin in 402–416 nm and 424–438 nm spectral region under 355 nm excitation^{3,20,33

–38} can be most important fluorophores for the differentiation of cervical smear supernatant sediment.

Additionally, based on our earlier spectroscopy study that demonstrated the difference of emission near 420 nm under 355 nm excitation in cervical smear related to hormone fluctuation during cervical maturation in pregnancy,³⁸ we suggest the effect of sexual hormones on the group differences in diagnostics. Sexual hormone relations to stromal component changes are established in other works.^39
–41

The fluorescence of smear supernatant sediment at 595–625 nm is important for normal versus cervicitis differentiation and probably corresponds to collagen VI⁴² and mucin I,²⁰ whereas the fluorescence at 480–515 nm, important for the diagnostics of cervicitis, could be related to lipid fluorescence;^17,43 it can cause flavins, the main cell and secretary granules' florescent components.⁴⁴ The diagnostic significance of spectral region 402–416 nm and 424–438 nm fluorescence in a smear is in harmony with the dynamics of main fluorescent constituents of the neoplastic epithelium proven elsewhere, and could serve as a cervical smear prognostic marker at the point of care and in the cytology laboratory.^2,33,45

Fluorescence decay time measured by us under 280 nm excitation in the same samples revealed differences in spectral region 320–360 nm (collagen IV, keratin) between normal and CIN2+ groups.^25,45 For this reason, combination of a few excitation waves is possible in Nd:YAG based microlaser for the purpose of improving the accuracy of epithelium diagnostic. We can use this dangerous excitation beam because we perform our measurements in vitro.

The absence of a CIN1 group is a weak point in this investigation, but inflammation and low- grade intraepithelial lesion (LGSIL) were found similar in earlier investigation. Cervical biopsies excited by 488 nm fluorescence intensity Student t test showed the chronic inflammation (CC) and LGSIL ratios having no significant spectroscopic difference. The sensitivity and specificity of the high-grade intraepithelial lesion (HGSIL) intensity compared with the mean ratio of CC and LGSIL was 89% and 100%.⁴⁶

Excitation by two waves and decay time evaluation would be the advancement elements of the device, but equipment cost enhancement is not suitable in terms of cancer screening economics. The application of smear/biopsy autofluorescence diagnostics in other high-risk human papilloma virus (HPV) related cancers' prevention at the point of care is reasonable to test, because global figures of head and neck cancer are higher than those for cervical cancer.^47,48 Moreover, unprocessed fresh samples are suitable for photodiagnostic purposes. The application of laser-induced fluorescence spectroscopy (LIFS) upon excitation of 405 nm, the comparison of emission spectra of bone cells, revealed that fluorescence intensity and the area under the spectra of malignant bone cells was lower than normal. This indicates the usefulness of simple fluorescence evaluation technology for the detection of malignancy at the point of care in the suspension of live cells.⁴⁹

Conclusions

In conclusion, a few diagnostic algorithms of high accuracy can be created by using the fluorescence spectra analysis by comparing its Gaussian components and revealing the flourescent markers. The autofluorescence spectral analysis of cervical liquid smear content permits the diagnostics of cervical epithelium: normal, cervicitis, and CIN2+.

Footnotes

Acknowledgments

The authors express their gratitude to the National Pathology Center for the cytological and histological evaluation of biopsies.

Author Disclosure Statement

No competing financial interests exist.

References

Hamashima

, Aoki

, Miyagi

et al. 2010. The Japanese Guideline for Cervical Cancer Screening. Jpn. J. Clin. Oncol., 40:485–502.

Branca

, Syrjänen

2009. Novel markers as predictors of CIN and prognosis of cervical cancer. Selected Papers on Gynaecologic Oncology. In Honour of 70th birthday of Professor Peter Bosze. Syrjänen

Budapest: PressCon., 51–78.

al–Saleh

, Delvenne

, van den Brule

F.A.

, Menard

, Boniver

, Castronovo

. 1997. Expression of the 67 KD laminin receptor in human cervical preneoplastic and neoplastic squamous epithelial lesions: an immunohistochemical study. J. Pathol., 181:287–293.

Greenwald

2007. A favorable view: progress in cancer prevention and screening. Recent Results Cancer Res., 174:3–17.

Brennan

, Justice

, Corbett

, McCarthy

, Galvin

. 2009. Emerging optofluidic technologies for point-of-care genetic analysis systems: a review. Anal. Bioanal. Chem., 395:621–636.

Kendrick

J.E.

, Huh

W.K.

, Alvarez

R.D.

2007. LUMA cervical imaging system. Expert Rev. Med. Devices, 4:121–129.

Robichaux–Viehoever

, Kanter

, Shappell

, Billheimer

, Jones

3rd , Mahadevan-Jansen

2007. Characterization of Raman spectra measured in vivo for the detection of cervical dysplasia. Appl. Spectrosc., 61:986–993.

Crystal Redden

, Richard

A.S.

, Atkinson

E.N.

et al. 2008. Model-based analysis of reflectance and fluorescence spectra for in vivo detection of cervical dysplasia and cancer. J. Biomed. Opt., 13:64016.

Werner

C.L.

, Griffith

W.F.

3rd , Ashfaq

et al. 2008. Comparison of human papilloma virus testing and spectroscopy combined with cervical cytology for the detection of high-grade cervical neoplasia. J. Low. Genit. Tract. Dis., 11:73–79.

10.

Soutter

W.P.

, Diakomanolis

, Lyons

et al. 2009. Dynamic spectral imaging: improving colposcopy. Clin. Cancer Res., 15:1814–1820.

11.

Ostrowska

K.M.

, Malkin

, Meade

et al. 2010. Investigation of the influence of high-risk human papillomavirus on the biochemical composition of cervical cancer cells using vibrational spectroscopy. Analyst, 135:3087–3093.

12.

El–Tawil

S.G.

, Adnan

, Muhamed

Z.N.

, Othman

N.H.

2008. Comparative study between Pap smear cytology and FTIR spectroscopy: a new tool for screening for cervical cancer. Pathology, 40:600–603.

13.

Monici

2005. Cell and tissue autofluorescence research and diagnostic applications. Biotechnol. Annu. Rev., 11:227–256.

14.

Mongiat

, Marastoni

, Ligresti

et al. 2010. The extracellular matrix glycoprotein elastin microfibril interface located protein 2: a dual role in the tumor microenviro nment. Neoplasia, 12:294–304.

15.

Sherman–Baust

C.A.

, Weeraratna

A.T.

, Rangel

L.B.A.

et al. 2003. Remodeling of the extracellular matrix through overexpression of collagen VI contributes to cisplatin resistance in ovarian cancer cells. Cancer Cell, 3:377–386.

16.

Waterman

E.A.

, Sakai

, Nguyen

N.T.

et al. 2007. A laminin-collagen complex drives human epidermal carcinogenesis through phosphoinositol-3-kinase activation. Cancer Res., 67:4264–4270.

17.

Nerlich

A.G.

, Lebeau

, Hagedorn

H.G.

, Sauer

, Schleicher

E.D.

1998. Morphological aspects of altered basement membrane metabolism in invasive carcinomas of the breast and the larynx. Anticancer Res., 18:3515–3520.

18.

Peyruchaud

2009. Novel implications for lysophospholipids, lysophosphatidic acid and sphingosine 1-phosphate, as drug targets in cancer. Anticancer Agents Med. Chem., 9:381–391.

19.

DaCosta

R.S.

, Andersson

, Cirocco

, Marcon

N.E.

, Wilson

B.C.

2005. Autofluorescence characterisation of isolated whole crypts and primary cultured human epithelial cells from normal, hyperplastic, and adenomatous colonic mucosa. Clin. Pathol., 58:766–774.

20.

DaCosta

R.S.

, Andersson

, Wilson

B.C.

2003. Molecular fluorescence excitation-emission matrices relevant to tissue spectroscopy. Photochem. Photobiol., 78:384–392.

21.

Pavlova

, Sokolov

, Drezek

, Malpica

, Follen

, Richards–Kortum

2003. Microanatomical and biochemical origins of normal and precancerous cervical autofluorescence using laser-scanning fluorescence confocal microscopy. Photochem. Photobiol., 77:550–555.

22.

Palsson

, Stenram

, Thompson

M.S.

et al. 2006. Methods for detailed histopathological investigation and localization of biopsies from cervix uteri to improve the interpretation of autofluorescence data. J. Environ. Pathol. Toxicol. Oncol., 25:321–340.

23.

Klinteberg

, Wang

, Lindquist

, Svanberg

, Vaitkuviene

1997. Laser induced fluorescence studies of premalignant and benign lesions in the female genital tract. Proc. SPIE, 3197:34–40.

24.

Vaitkuviene

, Andersen–Engels

, Auksorius

et al. 2005. Photo diagnosis of early pre cancer (LSIL) in genital tissue. Proc. SPIE, 5968:8–19.

25.

Vaitkuviene

, Gegzna

, Juodkazis

et al. 2009. Fluorescence spectrum and decay measurement for HSIL vs NORMAL Cytology differentiation in liquid PAP smear supernatant. AIP Conf. Proc., 1142:21–25.

26.

Vaitkuviene

, Auksorius

, Ramasauskaite

et al. 2004. LIF analysis of cervical mucus and amniotic fluid for maturity monitoring in pregnancy. Proc. SPIE, 5610:6–13.

27.

Vaitkuviene

, Gegzna

, Varanius

, Vaitkus

2011. Optimal Spectral regions for laser excited fluorescence diagnostics for point of care application. AIP Conf. Proc., 1380:31–35.

28.

Akobeng

A.K.

2007. Understanding diagnostic tests 3: receiver operating characteristic curves. Acta Paediatr., 96:644–647.

29.

Mirkovic

, Lau

, McGee

et al. 2009. Effect of anatomy on spectroscopic detection of cervical dysplasia. J. Biomed. Opt., 14:044021.

30.

Alfano

R.R.

, Katz

2000. Noninvasive fluorescence-based instrumentation for cancer and precancer detection and screening. Proc. SPIE, 3913:223–226.

31.

Gavriušinas

, Vaitkus

, Vaitkuviene

2000. Methodology of medical diagnostics of human tissues by fluorescence. Lithuanian J. Physics, 40:232–236.

32.

Millot

, Bondza–Kibangou

, Millot

J.-M.

, Lallemand

, Manfait

2003. Autofluorescence spectroscopy of malpigian epithelial cells, as a new tool for analysis of cervical cancer precursors. Histol. Histopathol., 18:479–485.

33.

Kamemoto

L.E.

, Misra

A.K.

, Sharma

S.K.

et al. 2010. Near-infrared micro-Raman spectroscopy for in vitro detection of cervical cancer. Appl. Spectrosc., 64:255–261.

34.

Lee

D.Y.

, Cho

K.H.

2005. The effects of epidermal keratinocytes and dermal fibroblasts on the formation of cutaneous basement membrane in three-dimensional culture systems. Arch. Dermatol. Res., 296:296–302.

35.

Martins

V.L.

, Vyas

J.J.

, Chen

2009. Increased invasive behaviour in cutaneous squamous cell carcinoma with loss of basement-membrane type VII collagen. J. Cell Sci., 122:1788–1799.

36.

Fülöp

, Larbi

2002. Putative role of 67 kDa elastin-laminin receptor in tumor invasion. Semin. Cancer Biol., 12:219–229.

37.

Branca

, Giorgi

, Ciotti

et al. 2006. Relationship of up-regulation of 67-kd laminin receptor to grade of cervical intraepithelial neoplasia and to high-risk HPV types and prognosis in cervical cancer. Acta Cytol., 50:6–15.

38.

Hornebeck

, Emonard

, Monboisse

J.C.

, Bellon

2002. Matrix-directed regulation of pericellular proteolysis and tumor progression. Semin. Cancer Biol., 12:231–241.

39.

Battlehner

C.N.

, Caldini

E.G.

, Pereira

J.C.

, Luque

E.H.

, Montes

G.S.

2003. How to measure the increase in elastic system fibres in the lamina propria of the uterine cervix of pregnant rats. J. Anat., 203:405–418.

40.

Breyer

B.N.

, Wang

, Lin

et al. 2010. The effect of long-term hormonal treatment on voiding patterns during filling cystometry and on urethral histology in a postpartum, ovariectomized female rat. BJU Int., 106:1775–1781.

41.

Kociecka

, Surazynski

, Miltyk

, Palka

2010. The effect of Telmisartan on collagen biosynthesis depends on the status of estrogen activation in breast cancer cells. Eur. J. Pharmacol., 628:51–56.

42.

Lorencini

, Silva

J.A.

, Almeida

C.A.

, Bruni–Cardoso

, Carvalho

H.F.

, Stach–Machado

D.R.

2009. A new paradigm in the periodontal disease progression: gingival connective tissue remodeling with simultaneous collagen degradation and fibers thickening. Tissue Cell, 41:43–50.

43.

Gierse

, Thorarensen

, Beltey

et al. 2010. A novel autotaxin inhibitor reduces lysophosphatidic acid levels in plasma and the site of inflammation. J. Pharmacol. Exp. Ther., 334:310–317.

44.

Zipfel

W.R.

, Williams

R.M.

, Christie

, Nikitin

A.Y.

, Hyman

B.T.

, Webb

W.W.

2003. Live tissue intrinsic emission microscopy using multiphoton-excited native fluorescence and second harmonic generation. Proc. Natl. Acad. Sci. U.S.A., 100:7075–7080.

45.

Gegzna

, Vaitkuviene

, Miasojedovas

, Kurtinaitiene

, Vaitkus

2010. Autofluorescence decay measurement of cervical smear supernatant for normal versus CIN2+ pathology discrimination. Biomedical Engineering – 2010: Proceedings of International Conference, Kaunas: Technologija, 167–169.

46.

Rodero

A.B.

, Silveira

Jr , Rodero

D.A.

, Racanicchi

, Pacheco

M.T.

2008. Fluorescence spectroscopy for diagnostic differentiation in uteri's cervix biopsies with cervical/vaginal atypical cytology. J. Fluoresc., 18:979–85.

47.

Syrjänen

2010. The role of human papillomavirus infection in head and neck cancers. Ann. Oncol., 21,Suppl 7:vii243–vii245.

48.

Shin

, Vigneswaran

, Gillenwater

, Richards–Kortum

. 2010. Advances in fluorescence imaging techniques to detect oral cancer and its precursors. Future Oncol., 6:1143–1154.

49.

Khosroshahi

M.E.

, Rahmani

2012. Detection and evaluation of normal and malignant cells using laser-induced fluorescence spectroscopy. J. Fluoresc., 22:281–288.