Anatomy-Guided Vision-Language Learning with Angular Prototype Separation for Multi-Label Video Capsule Endoscopy Classification Under Class Imbalance
Researchers have developed a novel framework for multi-label video capsule endoscopy classification, specifically addressing the challenge of extreme class imbalance in medical datasets. Their approach integrates an Angular Separation Loss with a Biological State Machine temporal decoder, utilizing the BiomedCLIP foundation model. This method enhances the detection of transient pathological signals and conditions predictions on anatomical context, leading to a significant improvement in classification accuracy on a challenging test set. AI
IMPACT Introduces a novel AI methodology to improve diagnostic accuracy in medical imaging by addressing class imbalance.