International Journal of Advanced Research in Computer and Communication Engineering

Abstract: Federated Learning (FL) has emerged as a powerful paradigm for privacy-preserving medical image analysis, enabling collaborative training across multiple pathology centers without sharing raw patient data. In breast cancer diagnostics, both histopathological image classification and segmentation are essential for identifying malignant regions and supporting early clinical decision-making. However, the development of high-performing deep learning models is challenged by institutional data silos, staining and scanner variability, annotation inconsistencies, and severe non-IID data distributions across clinical sites. This literature review synthesizes recent advances in FL for histopathology - including encrypted aggregation, differential privacy mechanisms, attention-guided architectures, parameter-efficient modality adapters, and segmentation-driven frameworks for whole-slide imaging. Special emphasis is placed on FedImp, an impurity-based optimization method that adaptively weights client updates using entropy-driven data-quality measures. While originally evaluated in classification scenarios, FedImp directly addresses critical limitations affecting segmentation tasks, such as uneven label availability, morphological heterogeneity, and client imbalance. By prioritizing informative updates and suppressing noisy or skewed contributions, FedImp enhances convergence stability, improves generalization, and reduces communication overhead in multi-center FL settings. Through a comparative review of ten influential studies, this work highlights existing methodological gaps and positions FedImp as a compelling foundation for future federated histopathological breast cancer segmentation pipelines, integrating privacy, robustness, and clinical scalability.

Keywords: Federated Learning, Histopathological Image Analysis, Breast Cancer Detection, Important Deep Neural Network Layers, Non-IID Data,FedImp.

Downloads: | DOI: 10.17148/IJARCCE.2025.1411142