Predicting Longitudinal Cognitive Decline via Multimodal Deep Learning Fusing Brain MRI and Clinical Data

作者：Tamoghna Chattopadhyay, Pavithra Senthilkumar, Rahul H. Ankarath, Christopher Patterson, Emma J. Gleave, Sophia I. Thomopoulos, Heng Huang, Li Shen, Lei You, Degui Zhi, Paul M. Thompson · 年份：2025 · DOI：10.1109/sipaim67325.2025.11283404 · 被引用次数：2 · 研究领域：Dementia and Cognitive Impairment Research、Machine Learning in Healthcare、Functional Brain Connectivity Studies

Predicting clinical decline in aging individuals, especially those with mild cognitive impairment, Alzheimer’s disease, Parkinson’s disease, or vascular dementia, is critical for guiding treatment and optimizing trials. We compared two deep learning methods for forecasting 2-year changes in the Clinical Dementia Rating sum of boxes (sobCDR), a key dementia metric ranging from 0 (no impairment) to 18 (severe). Our hybrid model combined 3D T1-weighted brain MRI with clinical and demographic data (e.g., age, sex, BMI, baseline sobCDR) using a convolutional neural network. We benchmarked it against AutoGluon, an automated multimodal ML framework. Our results highlight the value of integrating imaging and tabular data for personalized prognostics in dementia.