Adaptive Federated Few-Shot Rare-Disease Diagnosis with Energy-Aware Secure Aggregation

TL;DR

Rare-disease diagnosis suffers from data scarcity, privacy constraints, and heterogeneous edge devices. The paper proposes AFFR, a modular framework that combines few-shot federated optimization, meta-learning adapters, energy-aware client scheduling, and calibrated secure aggregation with differential privacy to enable privacy-preserving, label-efficient collaboration across sites. Empirical results on simulated and pilot clinical data show up to a 10% accuracy improvement over baselines, more than a 50% reduction in client dropouts with energy-aware scheduling, and privacy-utility trade-offs that remain clinically acceptable. AFFR offers a practical blueprint for equitable, trustworthy federated diagnosis in rare diseases with extensible plug-ins for meta-learning, secure aggregation, and resource-aware deployment.

Abstract

Rare-disease diagnosis remains one of the most pressing challenges in digital health, hindered by extreme data scarcity, privacy concerns, and the limited resources of edge devices. This paper proposes the Adaptive Federated Few-Shot Rare-Disease Diagnosis (AFFR) framework, which integrates three pillars: (i) few-shot federated optimization with meta-learning to generalize from limited patient samples, (ii) energy-aware client scheduling to mitigate device dropouts and ensure balanced participation, and (iii) secure aggregation with calibrated differential privacy to safeguard sensitive model updates. Unlike prior work that addresses these aspects in isolation, AFFR unifies them into a modular pipeline deployable on real-world clinical networks. Experimental evaluation on simulated rare-disease detection datasets demonstrates up to 10% improvement in accuracy compared with baseline FL, while reducing client dropouts by over 50% without degrading convergence. Furthermore, privacy-utility trade-offs remain within clinically acceptable bounds. These findings highlight AFFR as a practical pathway for equitable and trustworthy federated diagnosis of rare conditions.

Figures (6)

• Figure 1: High-level overview of the proposed Adaptive Federated Few-Shot Rare-Disease Diagnosis (AFFR) framework.
• Figure 2: LR: global accuracy across rounds under Baseline, DP, EA, and DP+EA. Faster convergence was observed with EA; DP caused a minor absolute drop.
• Figure 3: MLP: accuracy--round trajectories. Higher terminal accuracy and faster stabilization were obtained relative to LR; DP reduced accuracy slightly while EA accelerated convergence.
• Figure 4: Participating clients per round (of 12). EA achieved consistently higher participation, especially in later rounds as battery depletion accumulated.
• Figure 5: Average dropout rate by scenario. EA reduced dropouts by roughly half relative to random selection, with minimal effect on utility.