Explainable OCTA Phenotyping for Alzheimer's Disease Identification
September 7, 2026
A new pipeline integrates vessel segmentation and biomarker extraction with LLM-generated reporting for retinal OCTA imaging. Using 117 ROSE-1 images, the system achieved ROC-AUC values between 0.916 and 0.970 for identifying Alzheimer's disease biomarkers.
HOW THIS AFFECTS YOU
●
designerYou can leverage the measurement-grounded LLM reporting to create more interpretable diagnostic interfaces for clinicians.
●
healthThis offers a non-invasive, scalable way to screen for Alzheimer's via retinal microvasculature analysis.