Automated Histopathological Colon Cancer Image Classification Diagnosis using Advanced Federated Deep Learning Techniques

Colon cancer is also considered as one of the main reasons for cancer-related deaths, therefore, correct and early diagnosis significantly influences patients’ prognosis. Original diagnostic techniques employ human interpretation of histopathological images in which pathologists are involved; due to controversy and may be inconclusive, and slow. The colon cancer diagnostic automated system based on CNNs and modern image analysis techniques is proposed in this research. A fully automated system that offers histopathological diagnosis on images of tissue samples as being cancerous or non-cancerous exists with the primary objective of improving on diagnostic outcomes and duration. CNN was trained on an acquired data set of histopathological images which includes thousands of images annotated and divided to train, validate, and the test data set to use during training data augmentation and regularization. The authors successfully implemented the proposed approach in experiments that also demonstrated its high accuracy, sensitivity, specificity, and readiness for further use in clinical pathology settings. Further, the network and controller are designed for real-time computation with the least processing delay. The extension of this system to other types of cancer and the incorporation of explainable Artificial Intelligence methods to the models to improve clinicians’ trust.