A Detection Method for Alzheimer’s Disease Based on Multimodal Deep Learning

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder with no known cure. In China alone, more than 5 million individuals are affected, accounting for over one quarter of the global patient population, with approximately 300,000 new cases diagnosed annually. This paper proposes a multimodal deep learning framework for the automated diagnosis of AD based on brain imaging data, integrated with clinical information. Specifically, the model leverages a Transformer-based architecture to unify heterogeneous inputs, where embedding layers convert MRI scans, structured clinical records, and unstructured patient histories into a shared representation space. Through intra-modal and inter-modal attention mechanisms, the framework captures both local pathological patterns and cross-modal dependencies that are often overlooked by traditional methods. Extensive experiments are conducted on datasets derived from multiple imaging sources and clinical cohorts. Model performance is systematically evaluated using four key indicators—loss function, area under the receiver operating characteristic curve (AUC), accuracy, and F1-score—on both training and validation sets. Results demonstrate that the proposed method consistently outperforms conventional unimodal approaches, achieving improved diagnostic accuracy and robustness across diverse patient subgroups. Overall, the proposed multimodal deep learning framework provides a scalable and objective tool for supporting clinical decision-making in Alzheimer’s disease diagnosis. It not only reduces reliance on manual interpretation but also holds promise for early detection, risk stratification, and large-scale population screening in real-world healthcare settings.