Clinical trials for Alzheimer's disease (AD) often fail due to biased allocation of participants with varying cognitive decline rates. This study proposes a stratified randomization method using an AI model to predict cognitive decline, aiming to reduce allocation bias and improve trial efficiency. A multimodal deep learning model predicts CDR-SB changes using brain images and clinical data. Simulations using the ADNI dataset show that AI-based randomization reduces allocation bias by approximately 22%, leading to a nearly 37% reduction in sample size.
