دکتر جعفر زمانی

Alzheimer's disease is a complex and heterogeneous disorder in which pathological changes can begin years before clear clinical symptoms emerge. Advances in neuroimaging, fluid biomarkers, genetics, cognitive assessment, and electronic health data are providing increasingly rich sources of information, but integrating these diverse modalities into a coherent picture remains a major challenge. Multimodal artificial intelligence offers a promising approach for combining complementary data and identifying patterns that may improve the prediction of underlying Alzheimer's disease pathology.

This session explores how multimodal AI can integrate structural and functional imaging, molecular biomarkers, clinical variables, and other patient-level data to support earlier and more accurate characterization of Alzheimer's disease. It will examine the potential of these models for predicting pathological burden, identifying high-risk individuals, improving patient stratification, and supporting research and clinical decision making across the disease continuum.

The session will also address the key challenges involved in developing clinically meaningful multimodal models, including missing and heterogeneous data, limited generalizability, model interpretability, validation across populations, and the need to connect predictive performance with real-world clinical utility.