Early Parkinson's Disease Detection Using a Hybrid Graph-Spike Neural Network Architecture

Parkinson's disease (PD) affects millions worldwide, with early detection being crucial for effective treatment and quality of life maintenance. This paper presents a novel hybrid neural network architecture that combines Graph Neural Networks (GNNs) with Spiking Neural Networks (SNNs) for early-stage Parkinson's disease detection using gait analysis data. Our approach leverages the relational modeling capabilities of GNNs to capture patient similarity patterns while utilizing the neuromorphic properties of SNNs for biologically-inspired computation. The model achieves 100% accuracy on synthetic gait data from 50 patients (25 healthy, 25 Parkinson's), with prediction confidence exceeding 99.6% for positive cases. The system provides interpretable risk flagging and has been implemented as a deployable prototype using TorchScript. This work demonstrates the potential of hybrid neuromorphic architectures in healthcare applications and represents a significant step toward accessible, AI-augmented medical diagnostics developed through collaborative human-AI research methodology.