AI model enhances dementia diagnosis accuracy with multimodal data

An AI model accurately diagnoses various dementia types using multimodal data, aiding early and personalized treatment. The model achieved a microaveraged AUROC of 0.94 and AUPR of 0.90, outperforming CatBoost on ADNI and FHS datasets. It showed resilience to incomplete data and strong diagnostic ability across ten dementia etiologies, with an AUROC of 0.96. AI-augmented clinician assessments improved diagnostic performance, showcasing potential for enhanced clinical dementia diagnosis.