, pp.181-186 http://dx.doi.org/10.14257/astl.2017.147.26 Earlier Detection of Cervical Cancer from PAP Smear Images Asmita Ray 1, Indra Kanta Maitra 2 and Debnath Bhattacharyya 1 1 Assistant Professor Dept of Computer Science & Engineering, Vignan Institute of Information Technology, A.P. 1,2 Professor Dept of Computer Science & Engineering, Vignan Institute of Information Technology, A.P. 2 Assistant Professor Department of Information Technology, B.P. Poddar Institute of Management and Technology, Kolkata Abstract. Cervical cancer is the abnormal cell growth on cervix. Cervical cancer cannot be detected at an early stage as it does not show any symptom unlike other cancers so it is one of the most common causes of mortality among woman. PAP smear test is an efficient and common screening procedure for detection of abnormal cells in cervix. In this work an algorithm has been proposed to detect the cervical cancer. The proposed method comprises of three stages: preprocessing, segmentation and feature extraction. In case of cervical cancer cell nuclei are the useful key indicators of disease progression so segmentation of cell nucleus is an essential preliminary phase to detect the cervical cancer. In this proposed work segmentation of cell nucleus from the PAP smear images has been performed successfully. Texture features have been extracted using Gray Level Co-occurrence Matrix (GLCM). On the basis of features extraction method of the segmented image, cervical cancer effected images have been distinguish from the normal image. Keywords: Cervix, Cervical cancer, PAP Smear Images, Feature Extraction, GLCM. 1 Introduction Cervical cancer is the fourth most common form of cancer that occurs in woman belongs to all age groups. Symptom cannot be identified at an initial stage thus presently morbidity is increasing consistently. Real formation of cervical cancer is very slow. Early detection and confirmation of this kind of cancer is treatable and preventable. Human Papilloma Virus (HPV) is the main cause of cervical cancer [1]. PAP smear test is the preliminary manual screening method to identify the precancerous lesions of the cervix region [2] Cervical cancer is one of the highest leading causes of morbidity and mortality among the female worldwide. Cervical cancer has the significant impact, according to World Health Organization (WHO) it is estimated that approximately 530,000 new ISSN: 2287-1233 ASTL Copyright 2017 SERSC
cases in the world in 2012, more than 27,000 women died each year due to cervical cancer and around 80% of all cervical deaths occur in the developing country [3]. Several studies for segmentation of nuclei in cytological images have taken place [4-9]. Accurate detection and segmentation of cell nuclei from the PAP smear images is one of the challenging task due to of its high degree of overlapping with other cells, more than one nucleus in a cell, lack of homogeneity in image intensity. Current manual screening methods are very expensive and may have chances of producing inaccurate diagnosis result caused by human error. In this regard the computer assisted diagnosis system has a significant benefit which can reduce the financial cost as well as increases the screening accuracy. 2 Methodology The block diagram of proposed method is demonstrated in Figure 1. The figure represents the different steps of preprocessing, segmentation and feature extraction techniques to detect the cervical cancer from Pap smear images. Step 1: Consider Pap smear image. Step2 : (a) Convert image into gray image. (b) Remove noise. (c) Filtration is applied for enhancement. Step 3: (a) Watershed has been performed for extracting different pixels from background. (b) Threshold technique has been applied on segmented image. (c) Finally boundaries have been detected. Step 4: Feature analysis has been performed of the segmented image. 182 Copyright 2017 SERSC
2.1 Detail Description of the Proposed Method Preprocessing: Pap smear images play a crucial role for the past half century for screening of cervical cancer. Qualities of the Pap smear images are degraded due to various limitation and interference like variation of stain, uneven light across the field of view and noise during the transfer. So preprocessing of the image is the important step to assure the accurate analysis in segmentation step [10-11]. Pap smear images are the colored images, so the first and different phases of the preprocessing stage are as follows: (a) Covert it into gray scale image. (b) To enhance the contrast and remove the noise from the image, image adjustment and median filter of windows [3 X 3] have been applied, which are necessary to obtain stable boundaries and edges of nucleus. Original Image Adjusted Image Filter Image Preprocess Images of Normal Cell Original Image Adjusted Image Filter Image Preprocess Images for Abnormal Cell Copyright 2017 SERSC 183
: Image segmentation divides the image into multiple regions to serve the purpose of extracting of the important features from the image. has been done using following methods: Otsu Threshold : Otsu Thresholding method is a global thresholding method, is very simple and effective, useful for fast segmentation. In this method, optimal gray-level threshold value has been selected for separating the object of interest from the background. Watershed : In this segmentation, pixels are grouped depending on their intensities. Pixels having same intensities are grouped together. Edge Detection: Edge detection operation has been applied on the resultant image. Watershed Threshold Edged-Detected Image Segmented Images for Normal Cell Watershed Threshold Segmented Images for Abnormal Cell Edge-Detected Image Feature Extraction: Feature extraction method is a decision making process. Features are extracted to quantify the change in tissue or cell level. The GLCM features which are used in this paper are: Contrast: Contrast is measurement of sudden change of intensity value in image, which is calculated given below, i j 2 p(i. j) 184 Copyright 2017 SERSC
Correlation: It measures how the reference pixel correlates with its neighbor over an image. (i µi)(j µj)p(i, j) ( ) σ i σ j Energy: Energy is also called as angular second moment which defines the summation of squared elements in the GLCM. Energy is 1 for a constant. pi, j 2 Homogeneity: Homogeneity describes the distribution closeness of the element in the GLCM to GLCM diagonal. p(i, j) i + i j Table 1. GLCM features for normal and cancerous cell Type Contrast Correlation Energy Homogeneity Normal 0.6457 0.9412 0.1051 0.7995 Cancer 0.5345 0.9432 0.0924 0.8344 This table represents the different GLCM features like contrast, correlation, energy and homogeneity of normal and cancerous cell. 3 Conclusion This paper proposes computer-aided cervical cancer diagnosis system based on efficient detection and feature extraction approach where nucleus has been segmented successfully. This paper has been enabled to develop an algorithm which suppress the noise and enhance the contrast effectively for clear distinguishable boundary of the nucleus. In our study we have developed texture feature extraction method. This study not only will help the pathologist but also exempted them from tedious time consuming sometimes difficult and error prone manual interpretation. Early detection of the cervical cancer can be treated successfully and save the life, in this respect our new proposed method is very useful for fast and proper treatment by accurate detection of cancerous cell. Copyright 2017 SERSC 185
References 1. NCCC, Cervical cancer, http://www.nccc-online.org/index.php/cervicalcancer. 2. ACS, What is cervical cancer? 2011, American Cancer society, http://www.cancer.org/cancer/cervical 3. Cancer/DetailedGuide/cervical-cancer-what-is -cervical-cancer World Health Organization, http://www.who.int/en/. 4. Bamford P., Lovell B., Unsupervised cell nucleus segmentation with active contours, Signal Processing 71(2), pp 203-213, 1998. 5. Bamford P., Lovell B., A water immersion algorithm for cytological image segmentation, Proceedings of the APRS Image segmentation workshop, pp. 75-79, University of Technology Sydney, Sydney 1996. 6. Mouroutis T., Roberts S. J., Robust cell nuclei segmentation using statistical modelling, IOP Bioimaging, 6, pp. 79-91,1998. 7. Garrido A., Perez de la Blanca N., Applying deformable templates for cell image segmentation, Pattern Recognition 33, pp. 821-832, 2000. 8. Lee K.M., Street W.N., Learning shapes for automatic image segmentation, Proc. INFORMS- KORMS Conference, pp. 1461-1468, Seoul, Korea, June 2000. 9. Begelman G., Gur E., Rivlin E., Rudzsky M., Zalevsky Z., Cell nuclei segmentation using fuzzy logic engine, International Conference on Image Processing, Vol. 5, pp 2937-2940, October 2004. 10. R. Chouhan, R. K. Jha, and P. K. Biswas, Enhancement of dark and low-contrast images using dynamic stochastic resonance, IET Image Processing, vol. 7, no. 2, pp. 174 184, 2013. 11. J. Mukherjee and S. K. Mitra, Enhancement of color images by scaling the DCT coefficients, IEEE Transactions on Image Processing, vol. 17, no. 10, pp. 1783 1794, 2008. 186 Copyright 2017 SERSC