Advances in Medical Image Segmentation

The field of medical image segmentation is moving towards leveraging innovative data augmentation techniques and vision-language models to improve performance and reduce reliance on extensive expert annotations. Recent developments have focused on exploring the potential of hard and soft mixing data augmentation methods, as well as integrating vision-language models into semi-supervised learning frameworks. These approaches have shown promise in addressing data scarcity issues and improving segmentation accuracy. Notable papers include HSMix, which proposes a novel data augmentation approach involving hard and soft mixing for medical semantic segmentation, and ZEUS, which introduces a zero-shot visual-language segmentation pipeline for whole-slide images. Another noteworthy paper is VESSA, which integrates vision-language models into semi-supervised medical image segmentation, and SPROUT, which presents a fully training- and annotation-free prompting framework for nuclear instance segmentation.

Sources

HSMix: Hard and Soft Mixing Data Augmentation for Medical Image Segmentation

Zero-shot segmentation of skin tumors in whole-slide images with vision-language foundation models

Vision--Language Enhanced Foundation Model for Semi-supervised Medical Image Segmentation

Supervise Less, See More: Training-free Nuclear Instance Segmentation with Prototype-Guided Prompting

Built with on top of