3D CNN-BN: A Breakthrough in Colorectal Cancer Detection with Deep Learning Technique

Abstract

The Convolutional neural network (CNN) has made significant strides in the medical domain. CNN excels at the extraction of highly representative features in acute medical pathology. Amidst the network layers, CNN allows classification Through the process of filtering, selecting, and implementing these characteristics at the final layer level that is fully connected. Colon cancer rises from cells that cover the inner lining of the colon. Frequently originating from a noncancerous growth known as a polyp, it progresses gradually and eventually becomes cancerous. Our study aims to develop a computer-aided detection (CAD) system using a CT colonography dataset for colorectal cancer (CRC) prevention by classifying scans as polyp or polyp-free. After the preprocessing phase, we developed a deep-learning model with two variations: a 3D CNN-BN and a 3D CNN-BN & Dropout. 3D CT images of the abdomen classification according to the presence of polyps or its absence using CNN is primordial to enhance the chance of early detection. Thus, move toward the appropriate treatment. Our primary emphasis is on enhancing the training process of deep learning models and improving their performance during the testing phase. The findings suggest that the 3DCNN-BN demonstrated superior performance, achieving an accuracy of 92%.