Federated and Explainable Neuro Nail-SNN: A Privacy-Preserving, Fair, and Energy-Efficient Spiking Neural Framework for Nail Disease Diagnostics

Abstract

Automated nail disease diagnostics provide early indicators of systemic disorders such as diabetes and cardiovascular conditions. While NeuroNail-SNN achieved high accuracy and 65% energy reduction on edge devices , centralized training raises concerns regarding privacy, fairness, and interpretability. This paper proposes Federated and Explainable NeuroNail-SNN, integrating Federated Learning (FL), Explainable AI (XAI), fairness evaluation, and uncertainty quantification into a spiking neural framework. Federated training enables decentralized multi-hospital collaboration without sharing raw patient data. Spike saliency maps and temporal visualizations improve transparency. Fairness analysis ensures equitable performance across skin tone, gender, and age. Experimental results across five federated nodes demonstrate 97.85% accuracy with <0.3% drop compared to centralized SNN, while maintaining fairness gap < 2.5%. The proposed system bridges neuromorphic efficiency and trustworthy AI for real-world clinical deployment.