Colorectal Cancer Disease Classification and Segmentation Using A Novel Deep Learning Approach

Abstract

In the world, with a death rate of about 35% of colorectal cancer aids as the most widespread type of tumor. The third most commonly diagnosed cancer compared to breast and lung cancer is colorectal cancer. Specifically, in the minimization of health inequalities, it can be supported by the clinical care of AI guidance. Therefore in this paper an effective classification is performed by the deep learning technique called MobileNetV2. Before the classification task, cycle GAN based data augmentation technique is applied to solve the data imbalance problem. Moreover, Mask R-CNN based segmentation is applied to improve the classifier’s performance. For the classification of colorectal cancer, the proposed method is examined on a colorectal histopathological dataset (MNIST). The performance of the work will be estimated using precision, accuracy, F-score and recall metrics and the simulation results shows that the proposed classifier achieves 99.91% accuracy, 100% precision, 99.87% recall and 0.999% f1-score over the other state-of-art techniques.