ReMoE: Report-Guided Mixture-of-Experts for Multimodal OCT/OCTA Anomaly Detection

Researchers have developed ReMoE, a framework that uses medical reports to guide a mixture-of-experts model for identifying anomalies in retinal OCT and OCTA scans. By leveraging clinical text to assist image analysis, the approach aims to improve diagnostic accuracy in scenarios where training data for abnormal conditions is limited.