Explainable Multi-Modal Hybrid Deep and Machine Learning Framework for Alzheimer's Detection Using MRI and Clinical Data

Alzheimer's disease (AD) is a neurodegenerative disorder that progresses over time, and its early identification continues to pose a significant challenge in clinical settings. Conventional diagnostic approaches depending on neuroimaging or clinical assessment frequently do not fully reflect the disease's complex, multifaceted characteristics. This research showcases an interpretable multi-modal hybrid learning framework that combines deep learning models trained on magnetic resonance imaging (MRI) data with machine learning models applied to numerical clinical features. Convolutional and transformer networks capture distinguishing imaging characteristics, whereas gradient-boosting methods evaluate organized clinical factors. The integrated hybrid system shows enhanced classification capabilities at various disease stages, producing interpretable and balanced results that surpass single-modality approaches. Techniques for explainability like Gradient-weighted Class Activation Mapping (Grad-CAM++) and SHapley Additive exPlanations (SHAP) clarify model reasoning and highlight important biomarkers, creating a clear and trustworthy computational method for precision-focused Alzheimer's identification and clinical decision support.