Table 1.

Comparison of state-of-the-art studies on breast cancer detection and federated learning.

Ref.	Methodology	Dataset Used	Key Findings	Limitations
[7]	Auto-encoder for unsupervised breast cancer identification	Breast Cancer Wisconsin (Diagnostic) Dataset	Auto-encoder achieved 98.4% precision and recall	No comparison with other unsupervised methods, limited discussion on scalability
[30]	FL framework with CNN	Multiple cancer datasets	Achieved above 90% accuracy for five cancer types	Limited comparison with MLP and data imbalance not addressed
[31]	FL with transfer learning, SMOTE, and FeAvg-CNN + MobileNet model	Mammography datasets	Proposed method showed superior classification performance	No exploration of federated privacy concerns
[25]	FL for histopathology-based breast cancer classification	BreakHis dataset	FL model performed on par with centralized learning	Limited discussion on scalability to larger datasets
[32]	DL-based Xception model	Kaggle dataset	Model exhibited proficiency in breast cancer detection	Increased processing time not discussed
[33]	FL for predicting triple-negative breast cancer	Confidential patient data	FL enhanced the effectiveness of detecting triple cancer	Specific FL implementation details not provided, potential variability in data quality