Similarity Weighted Aggregation with Global Differential Privacy for Federated Brain Lesion Segmentation

Federated learning for medical image classification based on prototype alignment

Federated Learning-Enabled Image Processing for Privacy-Preserving Gastrointestinal Disease Screening

Background: Gastrointestinal (GI) diseases, including colorectal cancer, gastric cancer, polyps, and inflammatory bowel disease, account for over three million deaths annually worldwide. Automated deep learning-based screening from endoscopic images has demonstrated strong diagnostic potential; however, cross-institutional collaboration is severely impaired by patient data privacy regulations, yielding under-powered models trained on single-site data. We propose FedGI-Screen, a novel federated learning (FL) framework for privacy-preserving multi-institutional GI disease screening. The system integrates three original contributions: (i) a Heterogeneity-aware Federated Aggregation (HFA) algorithm that weights client contributions by data quality and distributional divergence, addressing the critical non-IID challenge in heterogeneous hospital data; (ii) an Adaptive Differential Privacy (DP) module (Adaptive DP-SGD) with dynamic gradient clipping calibrated per communication round via a Rényi accountant, achieving tighter privacy-utility trade-offs; and (iii) an EfficientNet-B4 + Lightweight Vision Transformer (ViT) hybrid backbone with multi-scale endoscopic image preprocessing and GradCAM-based explainability for clinical transparency. Evaluated across five publicly available GI endoscopy datasets (Kvasir, HyperKvasir, GastroVision, KvasirCapsule, EDD 2020; N = 76,884 images) simulated across 8 federated clients under non-IID conditions, FedGI-Screen achieves 94.8% accuracy, 94.7% F1-score, and an AUC of 0.976— surpassing FedAvg by 7.5 and centralised training-without-federation by 1.7 percentage points in F1. Under DP (ε = 6, δ = 10-5), performance degrades by only 0.8%, demonstrating a strong privacy-utility balance. FedGI-Screen demonstrates that privacy-preserving FL can match or exceed the performance of centralised models for GI disease screening, while maintaining rigorous data confidentiality compliance with GDPR and HIPAA. The proposed HFA and Adaptive DP-SGD provide novel, reviewer-validated contributions that advance the state of the art in both federated medical imaging and gastroenterological AI.

PPFedKD: Privacy‐Preserving Federated Learning and Adaptive Contrastive Distillation for Medical Image Diagnosis

Experimental results on diabetic retinopathy and breast cancer pathology datasets demonstrate that PPFedKD outperforms baseline methods in classification accuracy, privacy protection, and communication efficiency, providing a secure and effective solution for medical image classification.

A Divergence-Aware Personalized Federated Transformer Framework for Privacy-Preserving Multi-Center Medical Image Intelligence

In a non-IID medical imaging scenario, the problems that occur in federated learning (FL) include high data variability, data leakage, convergence instability, and suboptimal global aggregation. The number of medical imaging applications is vast, and Federated Learning (FL) has already been used in many of them; some main challenges are the large variability in data, the potential for leakage of privacy, convergence instability, and suboptimal global aggregation in non-IID scenarios. Current adaptive aggregation strategies are heuristic optimizer switching, do not take advantage of representation learning to transform the inputs, and are not able to personalize it in a divergence-aware way. This paper will present DAP-FedTrans, a Divergence-Aware Personalized Federated Transformer framework that will be used to conduct privacy-preserving multi-center medical image intelligence. This framework quantifies the statistical heterogeneity with Jensen-Shannon divergence, gradient similarity, and Wasserstein feature distance for the purpose of dynamically partitioning the clients and providing the aggregation per cluster. The classification heads are specialized for institutions, and the universal Vision Transformer encoder is used to encourage generalization. The privacy guarantees are augmented with secure aggregation and adaptive differential privacy. The accuracy is 97.84%; F1 is 0.968, and the loss of communication is 33% in the extreme non-IID case. The outcomes reveal increased convergence stability, equity, and scalability, making DAP-FedTrans a possible paradigm for the collaborative implementation of AI-based healthcare.

PriFedDD: Federated Discrete Diffusion for Medical Image Generation with a Utility-Privacy Trade-Off

Generative modeling for medical imaging requires accurate representation of anatomical structures while limiting direct exposure of patient data in cross-institutional collaboration. Although continuous diffusion models have shown effectiveness in natural image generation, their Gaussian corruption process is less aligned with discrete medical segmentation masks and may introduce artifacts in federated settings. To address this limitation, we propose PriFedDD, a privacy-aware federated discrete diffusion framework for medical image generation. PriFedDD formulates diffusion in a discrete state space to better model categorical segmentation data and incorporates an Information-Dense Encoding strategy that folds local spatial patterns into compact multi-channel tensors, reducing federated memory consumption and training time while preserving structural information. Experiments on standard benchmarks and medical datasets, including BraTS and Covid-Xray, demonstrate that PriFedDD achieves improved FID and KID for discrete segmentation generation compared to continuous baselines. Furthermore, privacy analysis based on PSNR and SSIM confirms the discrete formulation balances sample utility with reconstruction risk, and downstream augmentation experiments show the generated samples effectively improve federated learning performance on medical tasks.

Decentralized multimodal medical image segmentation via federated learning and segment anything model

Medical image segmentation faces critical challenges, including poor multi-center data generalization, limited multimodal handling, and high privacy risks. This study proposes a novel decentralized multimodal medical image segmentation method (FL-SAMMed) that integrates federated learning with the SAM-Med2D model. Firstly, the pre-trained SAM-Med2D model, optimized for feature learning across heterogeneous medical images, was leveraged to construct a multimodal segmentation model. A custom federated learning framework was then designed to enable privacy-preserving distributed training on multi-center data. Experimental results on multiple medical image datasets confirm that the proposed FL+SAM-Med2D method outperforms baseline approaches, achieving an average Dice score of 0.862 ± 0.028 and a mean HD95 of 6.68 ± 1.20 mm. These results highlight substantial improvements in segmentation accuracy and robustness, especially for cardiac MRI segmentation scenarios. Furthermore, the method mitigates data heterogeneity and ensures strict privacy compliance.