Hybrid Deep Learning for Lung and Colon Cancer Classification with XAI

Globally, lung and colon cancer are two of the most prevalent and deadly diseases which require efficient and accurate detection for proper treatment in early stages. Histopatho-logical images analysis is important for the early detection and classification of numerous diseases inlcuding lung and colon cancer. But it is challenging to differentiate between tissue types based on staining, architecture and overlapping morphological features variations on histopathological slides. The proposed method provides a hybrid deep learning model with combination of MobileNetV3 and Swin Transformer-T for the classification of cancer subtypes using lung and colon cancer histopathological images dataset. The images are preprocessed with transformation methods to enhance the quality of image and fit into the model. Features are extracted from both architectures which are used for the binary and multiclassification. This methodology have shown accuracy of 99.58% for classification. To increase the interpretability and transparency of model performance, SHapley Additive exPlanations(SHAP)is integrated which highlights the regions determining model predictions. This approach can act as highly effective tool in cancer diagnosis by enhancing clinical decision-making.