What question did this study set out to answer?

This study aims to improve medical image classification accuracy while ensuring data privacy through a novel federated learning framework.

May 14, 2026Scientific Reports2 citationsOpen Access

Multi-modal personalized federated learning with adaptive differential privacy for medical image classification and a privacy-preserving approach

AMAdhi Siva MVellore Institute of Technology University CCChiranji Lal ChowdharyVellore Institute of Technology University

Key Points

This study aims to improve medical image classification accuracy while ensuring data privacy through a novel federated learning framework.
Developed MM-PFL-ADP framework using Vision Transformer for multi-modal feature extraction.
Implemented adaptive differential privacy with per-parameter privacy budget allocation and personalization masks.
Evaluated performance on the MRI-MS dataset across 10 simulated medical institutions.
Achieved 97.3% accuracy on MRI-MS dataset (95% CI: 96.9–97.7%) with ε = 1.5, outperforming FedAvg (92.1%) and DP-FedAvg (87.3%) with p < 0.001.
Reduced communication by 47% and increased processing speed by 45%, completing in 47 rounds versus 85 rounds of FedAvg.
Maintained 95.2% accuracy under extreme data heterogeneity (α = 0.1) with membership inference attack probability dropping to 52.1%.

Abstract

Deep learning on medical images classification intervention needs to use large data on multi-institutional datasets but privacy laws inhibit sharing of data (GDPR, HIPAA). Federated Learning (FL) facilitates collaborative training without data transfer; until now, the known methods can only address privacy, personalisation, and accuracy not at the same time in a multi-modal environment. We present MM-PFL-ADP, a framework that combines Vision Transformer (ViT) based multi-modal feature extraction in four new elements: (i) privacy budget allocation (independent of number of samples): Fisher information-based adaptive per-parameter privacy budget allocation (₋₎₂₀₋ / ₒ₇₀ₑ₄₃ = 1. 5) ; (ii) personalisation masks: dynamic KL divergence based personalisation masks; (iii) respect The framework gives formal client-level (, ) -DP guarantees on transmitted gradient updates, in K = 10 simulated medical institutions. On the MRI-MS dataset, MM-PFL-ADP achieves 97. 3\% accuracy (95% CI: 96. 9–97. 7\%) at = 1. 5, outperforming FedAvg (92. 1\%) and DP-FedAvg (87. 3\%) by large margins (p < 0. 001). The framework is 45\% faster than FedAvg (47 vs. 85 rounds), has 47\% less total communication and keeps 95. 2\% accuracy in case of extreme heterogeneity in data (= 0. 1). The probability of membership inference attack has decreased to 52. 1 which was close to the random baseline (50\%). MM-PFL-ADP shows that the concepts of privacy, personalisation, and accuracy are synergistic, but not oppositional to federated medical AI. The single-system Fisher information framework greatly simplifies the hyperparameter tuning problem and can meet formal privacy criteria. Before being deployed, prospective validation against the performance of expert radiologists is desired.

Ask AI

Helpful

Bookmark

View Full Paper