Cervical Cancer Classification Using Improved Ridge Polynomial Neural Network

Abstract

Pap smear test is a standard examination for cervical cancer diagnosis. However, this method is very time-consuming and is very subjective in interpreting an image. This paper developed a system to diagnose the cervical cancer phase based on pap smear images. Five classes were investigated, namely: normal, precancerous (CIN1, CIN2, and CIN3), and malignant. The flow of the model is as follows: (i) pre-processes image using amoeba median filter and Gaussian filter (ii) nuclei detection, and segmentation (iii) extracts characteristics image using texture and shape analysis (iv) classify the pap smear image using Ridge Polynomial Neural Network pre-trained by Chaos Optimization. Based on experiments conducted, the proposed method could detect and classify the pap smear images with a sensitivity of 96.8%, specificity of 97.8%, and accuracy of 97%.