Ultrasound Image Segmentation
Overview
Ultrasound image segmentation identifies anatomical structures and lesions in noisy images, supporting measurements, diagnosis, and procedures. Common approaches include U-Net-style convolutional models, attention and transformer or state-space networks, and ultrasound-adapted foundation models; semi-supervised learning and pseudo-labeling aim to reduce the need for pixel-level annotations. Work also uses video consistency, anatomical or boundary constraints, and domain adaptation to improve results across frames, devices, and clinical sites. A central challenge is making masks reliable under domain shift and uncertainty, since high overlap scores can conceal boundary errors that affect clinical measurements.
Latest papers
73 papers6 papers in the last four weeks, down 33% on the four weeks before. 0.0% of all new papers.