Our paper on decentralized swarm learning for privacy-preserving medical AI on mobile edge devices has been accepted for publication at the 22nd International Conference on Mobile Web and Intelligent Information Systems (MobiWIS 2026), which will take place in Granada, Spain.
We are pleased to announce that our paper “Swarm Learning at the Edge: Privacy-Preserving Incremental Knowledge Transfer for Medical AI” has been accepted for publication at MobiWIS 2026.
The study, led by Burcu Selçuk and Tacha Serif, proposes a fully decentralized Swarm Learning (SL) framework designed for collaborative medical AI training on mobile edge devices without relying on centralized servers or persistent global models.
The proposed Android-based framework enables peer-to-peer model synchronization while preserving patient privacy by exchanging only lightweight classifier parameters instead of raw medical data. The system combines a frozen MobileNetV3 backbone with a trainable classification head, allowing efficient on-device incremental learning in resource-constrained environments.
The framework was evaluated using heterogeneous medical imaging datasets, including brain MRI and breast ultrasound images. Experimental results demonstrate that the merged decentralized model successfully preserves previously learned knowledge while integrating new medical imaging classes through peer-to-peer synchronization.
This research highlights the feasibility of privacy-preserving, decentralized, and resource-efficient collaborative AI systems for future mobile healthcare and edge computing applications.