Pediatric MS Detection

This study investigates whether multimodal deep learning models using optical coherence tomography (OCT) alone can accurately diagnose pediatric-onset multiple sclerosis (POMS). Using 3D OCT images and segmented retinal features from children with POMS and non-inflammatory controls, an early-fusion multimodal model achieved the best performance with 90% accuracy and an AUC of 0.87. The results demonstrate that combining retinal imaging and structural features significantly improves diagnostic accuracy, highlighting OCT-based multimodal AI as a promising tool for early pediatric MS diagnosis.