Detection of IC-NST

The paper addresses automated detection of invasive carcinoma of no special type (IC-NST) in breast histopathology using federated learning to overcome data scarcity and privacy constraints. It proposes a model that combines residual neural network features with Gabor-based features through late fusion and trains this architecture across multiple clients using locally held images from the breast histopathology image (BHI) dataset. The federated approach achieved competitive and state-of-the-art performance, generalized well to the BreakHis dataset, and outperformed existing methods in accuracy, F1 score, and AUC-ROC.