Multimodal Biomarker AI for Alzheimer's Diagnosis and Prognosis: A Systematic Review. 
Across 27 studies, multimodal biomarker AI consistently outperformed the best unimodal models for Alzheimer's/MCI classification (diagnostic AUC often ~0.85–0.95) and, more modestly, for MCI-to-AD prognosis (~0.75–0.85); limited external validation and interpretability still constrain clinical translation.