Classification Of Colon Cancer Using Deep Learning Techniques On Histopathological Images

Abstract

Colon cancer is widely spread and deadly type of cancer in humans. Smoking contributes to the growth of lung cancer, which leads to an unbalanced diet, which can lead to colon cancer. Identification of this type of cancer depends largely on the histological diagnosis. One of the most important research goals is colon cancer identification and classification, especially in the field of health systems. The proposed model suggests a deep learning model based on histopathological images for the identification of colon cancer. The proposed model introduces a novel approach utilizing Vision Transformer (ViT) and a new ViT version known as Swin Transformer. The model achieves high accuracy as benign or malignant by using a modified version of the Swin Transformer model. A comparative analysis of several models Swin Transformer, ViT, ResNet-101 and a modified Swin transformer is also presented in this paper. LC25000 dataset is used for training phase and testing of the proposed model. The final results demonstrate that the test model's accuracy is 99.80% using the modified version fo Swin Transformer model, 99.64% using the Swin Transformer model, 99.36% using the ViT model, and 98.27% using ResNet-101.