A Comprehensive Deep Learning Approach for Colon Cancer Prediction using Histopathological Image Analysis

Colon cancer, a significant contributor to both mortality and disability, results from a complex interplay of genetic and metabolic factors. The timely detection of this condition relies on histological examination, particularly the analysis of glandular structures within tissue regions. This leads to the development of different screening tests for investigating polyps images and colorectal cancer. In this study, we present a comprehensive approach driven by deep learning to identify colon cancer from histopathological images. In this work, the deep learning techniques, including MobileNet and Convolutional Neural Networks (CNNs) were employed, to train a dataset for the prediction of colon cancer. Furthermore, the deep learning models are assessed using standard performance metrics such as Accuracy, F1 score, MCC, Error rate, and Kappa. The results of the experiment show that the MobileNet model exhibited an outstanding accuracy of 98% whereas, the CNN model achieved an accuracy of 76% when differentiating between adenocarcinoma and benign tissues. The developed model shows promise in providing pathologists with a powerful tool for accurate and timely diagnoses.