Dysplastic Cervical Intraepithelial Neoplasia (CIN ... - OSA Publishing

JW3A.24.pdf

Frontiers in Optics 2013/Laser Science XXIX © OSA 2013

Dysplastic Cervical Intraepithelial Neoplasia (CIN) Tissues Detection using Spatial Frequency Analysis Yang Pu1, Jaidip Jagtap2, Asima Pradhan2 and R. R. Alfano2* 1

Institute for Ultrafast Spectroscopy and Lasers, Departments of Electrical Engineering and Physics of The Graduate Center at The City College of the City University of New York, 160 Convent Avenue, New York, NY 10031 2 The Department of Physics, Indian Institute of Technology (IIT), Kanpur, India  The corresponding author: R. R. Alfano may be reached by email at [email protected]

Abstract: Spatial frequency spectra from precancer cervical tissue images are presented and used to detect differences among different grades of human cervical tissues. OCIS Codes: 170.0170, 170.6480, 330.6100, 330.5000

1. INTRODUCTION Cervical dysplasia, Cervical Intraepithelial Neoplasia (CIN), is premalignant and abnormal squamous cells on surface of cervix [1]. The current techniques for CIN grading include the Papanicolaou or “Pap” smear and colposcopy, by which the CIN grading can be only confirmed by biopsy [1]. There is a need to detect early cervical cancer without taking biopsy. In this presentation, the spatial frequencies of normal and CIN tissues were obtained by a Fourier Transform analysis of the light intensity distribution of the cervix ex vivo samples. Since spatial frequency spectra provide information of the periodic and/or random structures of two dimensional (2D) light intensity distributions, and the periodic structure in the stromal region of tissue is related to the cancer stages progressing [1, 2], the spatial frequency spectra may offer alternative way to analyze CIN grades. 2. METHODS A set of human cervix of Normal, CIN 1, CIN 2, and CIN 3 tissues stained by H&E is used in this study. Figures 1 (a), (b), (c), and (d) show confocal microscope images of the normal, CIN 1, 2, and 3 tissues, respectively. The intensities distribution of these images can be expressed as 2D functions f(x,y) in spatial coordinates (x,y)[3]:       2ux 2vy  2ux 2vy  , f ( x, y )   b sin    au , v cos  L L  u ,v  L L  u0 v0 y  y   x  x

(1)

where u and v are the numbers of cycles of f(x,y) with a periodic Lx and Lx in the x and y directions, respectively. Fast Fourier Transform (FFT) is usually-used mathematical tools to convert the 2D spatial function f(x,y) into the 2D spatial spectrum F(u,v) of a finite extent N × N sampling of 2D intensity distribution [3]: F( u ,v ) 

1 N

N 1 N 1



x 0 y 0



 2 ( ux  vy )   2 ( ux  vy )    j sin  . N N   

  f ( x, y )cos

(2)

The magnitude spectra, |F(u,v)|, are then calculated using [3]:

F ( u , v )  R 2 ( u ,v )  I 2 ( u , v ) .

(3)

where R(u,v) and I(u,v) are the real and imaginary parts, respectively.

140 120 100 80

3. RESULTS AND DISCUSSION

60

40 To obtain the information of spatial frequency for cervical tissue at different CIN grade, the 2D DFT 20 of Figs. 1 was achieved by sampling N = 256. (a) Normal (b) CIN 1 (c) CIN 2 (d) CIN 3 After the 2D amplitudes spectra of normal, CIN 1, 2, and 3 cervical tissues are obtained, the Fig. 1. Portion of typical confocal microscope image of (a) normal, digital spatial cross section frequency (c) CIN 1, (c) CIN 2, and (d) CIN 3 tissues. distributions of the normal, CIN 1, 2, and 3 tissues crossing the areas of the most dominant frequency along horizontal direction can be extracted, which are shown as Figs. 2 (a), (b), (c), and (d), respectively. The salient differences among Figs. 2 (a) to (d) are that higher frequency components exist in CIN tissues than those in normal tissue, as well as those in higher grade CIN tissue

JW3A.24.pdf

Frontiers in Optics 2013/Laser Science XXIX © OSA 2013

than those in lower grade CIN tissue. The spatial frequency of different types of tissue images shows that the higher grade of CIN tissue, the wider spatial frequency range [3].

150

normal

100 50 0 -80

-60

-40

-20

0

20

40

60

Spatial frequency (Cy/pixel)

80

250

(b)

200

CIN 1 150 100 50 0 -80

-60

-40

-20

0

20

40

60

80

250

(c) Magnitude (Arbitrary Unit)

200

Magnitude (Arbitrary Unit)

250

(a)



250

200

CIN 2 150 100 50 0 -80


-60

-40

-20

0

20

40

60

80

(d)

200 150

CIN 3

100 50 0 -80

-60

-40

-20

0

20

40

60

80



Fig. 2. Spatial frequency of (a) normal, (b) CIN 1, (c) CIN 2, and (d) CIN 3 obtained by the digital spatial cross section frequency distributions at the most dominant frequency along horizontal direction of 2D map of spatial frequency of their corresponding type of tissues.

It is well known that the patterns of normal and low grade CIN tissues consist of evenly placed uniform epithelia cells supported by a well-structured extracellular matrix (ECM) [1]. With grade advancing, ECM degrades and causes loss and randomness of collagens [1]. Since the collagen in the normal tissue is more ordered in layers and uniform in shape and size while those in CIN precancer tissues are aperiodic random, anti-symmetrical, different sizes, and disordered in structure with more structure parameters [1], the higher grade CIN tissues should have wider spatial frequency range in comparison with lower grade CIN and normal tissues. The wider spatial frequency of higher CIN tissue may be provide a diagnostically criterion for grading CIN tissues.

4. CONCLUSION In conclusion, this pilot investigation on spatial Fourier analysis of human normal, CIN 1, CIN 2, and CIN 3 tissue in vitro could discriminate the normal and three grades CIN tissues. Further based on “width of the spatial frequency” as a function of CIN grade, a spatial spectral grading in parallels with CIN grading could be established. One can envision the potential of this method for in vivo applications. Moreover, the features of disorder, aperiodic random, anti-symmetrical collagens in different size are also hallmarker of prostate [2, 4], breast [5] and other types of dysplasia and/or tumor, this study may impact on the detection of other types of cancer. 5. ACKNOWLEDGMENTS

Spatial Frequency at 10% of Maximum

To evaluate this potential, Fig. 3 shows the width of range for the spatial frequency from the full maximum decreasing to 10% of the maximum as a function of normal and CIN grade. It is important to note that the “width of the spatial frequency signal” exhibits feature of growth with the CIN grades. This property can be schematically shown as the dash dote line in Fig. 3, which can be characterized by correlation coefficient: R2 = 0.97. An attempt was achieved to establish to grade the CIN tissues using spatial frequency analysis. 60

2

50

Linear fitting with R = 0.97

40 30 20

Normal

CIN 1 CIN 2 CIN 3 Tissue Type Fig. 3. The levels of “width of the spatial frequency” as a function of normal and CIN grade.

This research is supported in part by U. S. Army Medical Research and Material Command (USAMRMC) under grant of # W81XWH-11-1-0335 (CUNY RF # 47204-00-01). REFERENCES [1] Barbara Hoffman, John Schorge, Joseph Schaffer, Lisa Halvorson, Karen Bradshaw, F. Cunningham, “Williams Gynecology, Second Edition,” McGraw-Hill Professional, ISBN 978-0-07-171672-7, Copyright @ 2012. [2] Yang Pu, W. B. Wang, Yuanlong Yang, and R. R. Alfano, “Stokes Shift Spectroscopic analysis of multi-fluorophores for human cancer detection in breast and prostate tissues,” J. Biomed. Opt., 18(1), 017005-1-8 (2013). [3] Joseph Goodman, “Introduction to Fourier Optics, Third Edition,” Roberts & Company, ISBN 0-9747077-2-4, Copyright @ 2005. [4] 6. D. F. Gleason and G. T. Mellinger, “Prediction of prognosis for prostatic adenocarcinoma by combined histological grading and clinical staging”, J. Urol, 111, 58-64 (1974). [5] Yang Pu, W. B. Wang, Yuanlong Yang, and R. R. Alfano, “Native fluorescence spectra of human cancerous and normal breast tissues analyzed with nonnegative constraint methods,” Appl. Opt., 52(6), 1293-1301 (2013).