Amyotrophic Lateral Sclerosis Detection and Staging Through Hybrid Convolution Neural Network and Vision Transformer Framework
Year: 2026
Type: conference-paper
Abstract: Amyotrophic Lateral Sclerosis (ALS) is a progressive neurodegenerative disease characterized by motor neuron degeneration and multisystem involvement, often accompanied by subtle visual dysfunction during the early stages of the disease. However, current diagnostic approaches remain limited in their ability to simultaneously detect ALS, classify disease stages. This study proposes a multimodal deep learning framework for comprehensive ALS analysis through the integration of heterogeneous biomedical modalities, including Optical Coherence Tomography, Electroencephalography, Electrooculography, and Electrocardiography. The proposed framework addresses four major tasks: the vision task is defined as the representation learning objective for modality-specific feature enhancement, while the risk task is defined as the predictive estimation of disease progression. The proposed system addresses ALS detection, stage classification, and severity assessment using a hybrid convolution neural network-vision transformer architecture combining local spatial feature extraction and global contextual representation. Also, an attention-based multimodal fusion strategy is utilized to capture complex interactions among structural, neural, oculomotor, and physiological biomarkers. Experimental results demonstrate that the proposed model achieves 95% accuracy for ALS detection, outperforming unimodal baseline approaches 86%. Overall, the proposed approach provides clinically interpretable outputs for early diagnosis, disease staging, and proactive risk assessment, contributing to the advancement of precision medicine and early intervention strategies for ALS. Future prospect includes, but not limited to increasing the clinical database and improving the modal generalization.
Institutions: Beirut Arab University, Université Paris Cité
Language: English
Subfield: Clinical Neurology
Field: Medicine
Domain: Health Sciences
Open Access status: hybrid
OpenAlex ID: W7204165017