Diagnosis of cervical cancer using CLAHE and SGLDM on RGB pap smear images through ANN

Abstract

Cervical cancer is one of the most preventable Cancers in the present day. PAP and HPV tests are followed for early diagnosis for screening procedure. In medical field, it is difficult to detect Pap smear images through microscope. Image processing is a vital role to take right decision by using various technique and algorithms. In this paper, the proposed technique is used to enhance the Pap smear images by comparing Histogram Equalization in Contrast Stretching algorithm, Power Law Transformation for Gamma Correction, Shading Correction, Contrast Limited Adaptive Histogram Equalization (CLAHE). The Quality measurement MSE, PSNR, SC and NAC value has been calculated to find performance analysis of enhanced Pap smear images. Then proposed four different feature extraction algorithms SGLDM-Spatial Gray Level Difference Method, RDM-Run Difference Method, LBP-Local Binary Pattern and HOG-Histogram of Oriented Gradients are used to extract features of Pap images. Experimental results are obtained for a data set of 215 Pap images taken from the Management and Decision Engineering Laboratory (MDE-LAB) database. MATLAB R2016a used as a programming tool. ANN Classification is used for each feature extraction algorithm to evaluate the accuracy level. Thus, CLAHE achieved the good result for enhancement and SGLDM feature extraction algorithm reached 93% accuracy using ANN.