Related Experiment Video
Updated: May 14, 2026
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
Reliable Multi-Prototypical Contrastive Learning for Semi-Supervised Heterogeneous and Multi-Organ Medical Image
IEEE Transactions on Medical Imaging
|May 12, 2026
Summary
This study introduces TP-Net, a novel semi-supervised framework for accurate multi-organ segmentation in diverse medical images. It overcomes data scarcity and heterogeneity, improving surgical navigation systems.
Area of Science:
- Medical Imaging
- Computer Vision
- Machine Learning
Background:
- Accurate multi-organ segmentation is crucial for surgical navigation but hindered by limited annotated data.
- Existing semi-supervised methods struggle with inter-class feature ambiguity and multi-source data heterogeneity.
Purpose of the Study:
- To develop TP-Net, a semi-supervised framework addressing bottlenecks in medical image segmentation.
- To enhance multi-organ segmentation accuracy in heterogeneous medical datasets for clinical applications.
Main Methods:
- Proposed TP-Net framework integrating reliable multi-prototype contrastive learning.
- Developed a heterogeneous prototype dynamic evolution mechanism for adaptive intra-class distribution modeling.
- Implemented an uncertainty-aware cross-domain alignment strategy and contrastive separation to resolve feature ambiguity and prototype shift.
Main Results:
- TP-Net achieved state-of-the-art performance on public and in-house datasets.
- Clinical validation demonstrated the method's viability in a self-developed surgical navigation system.
Conclusions:
- TP-Net effectively addresses challenges in semi-supervised multi-organ segmentation.
- The framework shows significant potential for improving real-world surgical navigation systems.