Early Detection and Classification of Colon Cancer using Region Aware Cross Attention U-Net

In recent days colon cancer has become a major health issue that is affecting millions of individuals worldwide. It is a type of cancer that affects the areas of the large intestine, which causes the patients to lose their lives. It is also found that colon cancer is the second leading cause of all cancer, particularly in the United States. In the year 2024, approximately about 53,010 people have lost their lives because of this deadliest cancer. Early prediction and proper diagnosis are very important for this type of cancer. By doing so, effective treatments can be given to prevent the development of the disease. Various techniques have emerged tremendously by utilizing Artificial Intelligence (AI). Especially, deep learning applications have gained popularity in recent days, as they analyze images and further promote disease detection and diagnostic processes earlier. In this study, colon cancer is detected by using the hybrid approach, which is named Region aware cross attention U-Net. It helps in predicting and classifying the medical images that are obtained for disease detection. Additionally, a Tissue morphology map as a pre-processing technique along with multi-mask RCNN as a segmentation technique is also used to enhance the classification process. Moreover, this study also serves as a comparative analysis as this compares the works of some existing algorithms namely, Res-Net, VGGNet, U-Net with the proposed algorithm. Furthermore, the works are evaluated by using the evaluation metrics, namely accuracy, precision, recall, and F1 score.