Integrating multimodal neuroimaging genetics and deep learning to predict Alzheimer's disease: a systematic review

Elucidating the pathological mechanisms and enabling early, precise diagnosis of Alzheimer's disease (AD) are critical goals in clinical practice. Neuroimaging genetics provides a powerful framework for unraveling the complex genetic architecture and neural substrates of AD across macroscale brain structures and molecular pathways. This review systematically outlines state-of-the-art methodologies for computer-aided AD diagnosis using multimodal data. Specifically, we examine neuroimaging modalities including structural and functional magnetic resonance imaging (MRI), diffusion tensor imaging (DTI), and positron emission tomography (PET), alongside genetic data such as single-nucleotide polymorphisms (SNPs) and gene expression profiles. Correlation analyses are discussed for identifying preliminary associations between genetic variants and imaging phenotypes. Traditional machine learning approaches enable linear cross-modal data fusion, while deep learning methods capture complex nonlinear interactions between genetic and neuroimaging features. Accumulating evidence indicates that multimodal integration significantly improves the accuracy of both AD diagnosis and prognostic prediction. Future research should leverage large-scale longitudinal cohorts and foster interdisciplinary collaboration to enhance the interpretability and clinical applicability of deep learning models, thereby advancing AD research from associative observations toward mechanism-driven precision medicine.