Co-Seg++: Mutual Prompt-Guided Collaborative Learning for Versatile Medical Segmentation.

Area of Science:

• Medical Image Analysis
• Computer Vision
• Biomedical Engineering

Background:

• Medical image segmentation is crucial but challenging, especially for joint organ/tissue segmentation and instance identification.
• Existing methods often isolate segmentation tasks, neglecting interdependencies and leading to suboptimal performance.

Purpose of the Study:

• To propose Co-Seg++, a novel framework for versatile medical image segmentation.
• To enhance segmentation performance by enabling semantic and instance segmentation tasks to mutually benefit each other.

Main Methods:

• Introduced a novel co-segmentation paradigm for integrated semantic and instance segmentation.
• Developed a spatio-sequential prompt encoder (SSP-Encoder) for spatial and sequential relationship encoding.
• Designed a multi-task collaborative decoder (MTC-Decoder) for joint mask computation and contextual consistency.

Main Results:

• Co-Seg++ demonstrated superior performance across diverse CT and histopathology datasets.
• Achieved state-of-the-art results in semantic, instance, and panoptic segmentation.
• Successfully segmented dental anatomical structures, histopathology tissues, and nuclei instances.

Conclusions:

• The proposed Co-Seg++ framework effectively addresses the limitations of isolated segmentation tasks.
• Co-Seg++ offers a versatile and high-performing solution for complex medical image segmentation challenges.
• The framework shows significant potential for advancing medical image understanding and analysis.