Interactive Medical Image Segmentation
Overview
Interactive medical image segmentation studies how people and AI can work together to create and correct masks of anatomy, lesions, and surgical structures, making image analysis more accurate and annotation less laborious. Common approaches use promptable foundation models guided by clicks, boxes, scribbles, or language, alongside mask propagation, uncertainty-based review, and online adaptation from corrections. Methods are evaluated across 2D images, 3D CT and MRI volumes, ultrasound, and surgical video, often measuring both segmentation quality and the number or time cost of interactions. A key challenge is making systems reliable across modalities and clinical settings while ensuring that simulated-user results translate to efficient, trustworthy workflows with real experts.
Latest papers
52 papers4 papers in the last four weeks, level with the four weeks before. 0.0% of all new papers.