ScribFormer: Hybrid CNN-Transformer for Medical Image Segmentation

This session introduces ScribFormer, a novel model for medical image segmentation addressing challenges in sparse scribble annotations through a triple-branch architecture combining CNN, Transformers, and ACAM for refined feature analysis. Attendees will gain insights into its design, components, and superior performance validated against state-of-the-art models. The discussion includes its adaptability across medical imaging contexts, leveraging pseudo-labeling and innovative loss functions...