Federated Breast Cancer Detection Enhanced by Synthetic Ultrasound Image Augmentation

2025-06-29

AI Summary

Federated learning (FL) for breast ultrasound diagnosis suffers from poor model generalization and performance degradation due to medical data silos, scarce annotated samples, and highly non-IID client data. Method: We propose a class-specific generative adversarial network (GAN) that synthesizes high-fidelity benign/malignant ultrasound lesion images locally; synthesized data—incorporated into FL training at controllable ratios—is integrated into both FedAvg and FedProx frameworks, enabling privacy-preserving multi-center collaboration on real-world clinical datasets. Contribution/Results: Experiments demonstrate that moderate incorporation of synthetic data improves AUC by 0.0031 for FedAvg and 0.0109 for FedProx, significantly enhancing model robustness and diagnostic accuracy. To our knowledge, this is the first work to systematically introduce class-aware generative data augmentation into medical FL, establishing a scalable new paradigm for cross-institutional, small-sample, non-IID collaborative modeling.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Security and Privacy: Data transparency and provenance

Abstract

Federated learning (FL) has emerged as a promising paradigm for collaboratively training deep learning models across institutions without exchanging sensitive medical data. However, its effectiveness is often hindered by limited data availability and non-independent, identically distributed data across participating clients, which can degrade model performance and generalization. To address these challenges, we propose a generative AI based data augmentation framework that integrates synthetic image sharing into the federated training process for breast cancer diagnosis via ultrasound images. Specifically, we train two simple class-specific Deep Convolutional Generative Adversarial Networks: one for benign and one for malignant lesions. We then simulate a realistic FL setting using three publicly available breast ultrasound image datasets: BUSI, BUS-BRA, and UDIAT. FedAvg and FedProx are adopted as baseline FL algorithms. Experimental results show that incorporating a suitable number of synthetic images improved the average AUC from 0.9206 to 0.9237 for FedAvg and from 0.9429 to 0.9538 for FedProx. We also note that excessive use of synthetic data reduced performance, underscoring the importance of maintaining a balanced ratio of real and synthetic samples. Our findings highlight the potential of generative AI based data augmentation to enhance FL results in the breast ultrasound image classification task.

Problem

• Enhancing federated learning for breast cancer detection
• Addressing limited data in federated medical imaging
• Balancing synthetic and real ultrasound image data

Innovation

• Generative AI for synthetic ultrasound image augmentation
• Federated learning with synthetic image sharing
• Class-specific DCGANs for benign and malignant lesions