arXiv AI recent: REVEAL++: Differentiable Phenotypic Grouping for Vision-Language Retinal Modeling of Alzheimer's Disease Risk
Researchers proposed REVEAL++, a vision-language modeling framework for predicting Alzheimer's disease risk using retinal fundus images and clinical risk narratives. The framework replace...
The model uses a soft-target contrastive objective to jointly learn cross-modal alignment and phenotypic structure. It was evaluated using UK Biobank retinal imaging data for incident Alzheimer's disease prediction, where it outperformed standard vision-language baselines and discrete group-based...