EAI4CC: deciphering lung and colon cancer categorization within a federated learning framework harnessing the power of explainable artificial intelligence

Abstract

Advances between medical imaging and artificial intelligence (AI) have led to improvements in cancer diagnosis and classification. This paper provides a new framework called Explainable AI for cancer categorization (EAI4CC), which has been developed to define lung and colorectal cancer classification in an integrated manner, addressing privacy concerns by enabling collaborative model training using Federated Learning. In this study, EAI4CC used convolutional neural networks (CNNs) such as VGG 16, VGG19, ResNet50, DenseNet121 and Vision Transformer to analyze histopathological images from lung and colon tissue. In Federated Learning architecture it ensures data privacy while enabling model training on dispersed dataset. Furthermore, state-of-the-art artificial intelligence (XAI) presentation techniques are used. In particular gradient-weighted class activation mapping (GradCAM) combined with EAI4CC to elucidate the decision-making process of the model. The evaluation system shows good performance in important evaluation measures such as accuracy, precision, specificity, sensitivity, and F1 score. More importantly, it enhances model interpretation capabilities, explaining each prediction. This gives doctors clarity and confidence in AI-assisted diagnosis. Interpretable and reliable methods allow AI technologies to be responsibly integrated into the critical cancer research workflow to demonstrate the performance of model measures. In summary, this breakthrough sets a standard to establish a framework for AI to achieve more accurate, transparent, and equitable clinical decision-making.