Bobcat-optimized hybrid quantum-classical spike-driven network for MRI-based Alzheimer's stage prediction

Alzheimer's disease (AD) is a progressive neurodegenerative disorder that severely affects memory, cognition, and behavioral functions, making early and accurate stage classification essential for timely clinical intervention and treatment planning. Conventional MRI-based diagnostic approaches are often limited by noise sensitivity, manual interpretation, and insufficient capability to model complex non-linear neuroimaging relationships across multiple disease stages. To address these limitations, this paper proposes a novel Hybrid Quantum-Classical Spike-Driven Network optimized with the Bobcat Optimization Algorithm(HQSDNet-BOA)for automated multistage AD classification using MRI data. The suggested framework introduces an integrated architecture that combines Square Root Sage-Husa Adaptive Robust Kalman Filtering(SRS-HARKF)for adaptive noise suppression and covariance stabilization, Graph-Enhanced Fuzzy Clustering (GEFC)for structurally consistent brain tissue segmentation, and a Hybrid Structural Graph Attention Network(HSGAN)for learning discriminative local-global anatomical representations. Moreover, Hybrid Quantum-Classical Spike-Driven Network (HQSDNet), which incorporates quantum convolutional learning, spike-driven transformers, and structural attention schemes, is designed to precisely identify nonlinear and spatial-temporal disease patterns. Bobcat Optimization Algorithm (BOA) is used for optimizing network parameters in real-time to ensure convergence efficiency and computational optimization. This study was performed using two datasets: ADNI (5,300 images representing five disease stages) and OASIS-3 (3,712 images representing three cognitive classes). The suggested HQSDNet-BOA yielded improved classification results with 98.8% accuracy, 98.0% precision, 98.5% recall, and 98.25% F1-score when applied to the ADNI dataset and outperformed the existing techniques in terms of computational speed. The obtained results confirm that the suggested framework provides a robust, efficient, and clinically relevant solution for accurate AD stage prediction and neuroimaging-based diagnostic support.