Computer Vision Based Diagnostic System For Cervical Cancer Using Colposcopy Images

Abstract

Cervical cancer is common and preventable disease that poses a significant threat to women’s health and well-being. It is the fourth most prevalent cancer among women worldwide, with approximately 604,000 new cases and 342,000 deaths in 2020, according to the World Health Organization. This project aims to develop a computer vision–based diagnostic system for cervical cancer using a colposcopy images . Early detection plays a pivotal role in improving patient treatment by timely intervention and effective treatment. This paper explores two distinct strategies for enhancing cancer detection. The dataset contains 350 to 500 images 80% images are kept in train and 20 % images are kept in test set. This data is use to classify the core of cancer. Firstly ,Convolution Neutral Network Consists of Customized CNN model ,not a pre-trained architecture. CNN layers contains Dense layer, flatten layer,conv2D layer, pooling layer, batch normalization, Dropout layer, fully connected layer. Secondly ,Region base Convolution Neutral Network it uses the CNN layers as a backbone for feature extraction. Through this the accuracy of the model is increased to the 0.95 to 0.98 in all three classifications they are severe, moderate ,mild. Finally, the confusion matrix is shown between CNN and partial RCNN