Related Experiment Video
Updated: May 22, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
Uncertainty-guided model learning for trustworthy medical image segmentation.
1School of Electronic and Information Engineering, South China University of Technology, Guangzhou, 510640, China.
Medical & Biological Engineering & Computing
|May 21, 2026
Summary
This study introduces a novel trustworthy medical image segmentation method. It enhances reliability and uncertainty estimation using multi-scale features and subjective logic, improving clinical AI adoption.
Area of Science:
- Artificial Intelligence
- Medical Imaging
- Computer Vision
Background:
- Clinician skepticism exists regarding the reliability of current AI algorithms in medical imaging.
- Trustworthy AI is crucial for the clinical adoption of advanced medical image analysis tools.
Purpose of the Study:
- To develop a trustworthy medical image segmentation method.
- To provide reliable segmentation results and uncertainty estimations.
- To minimize the computational burden of AI in medical imaging.
Main Methods:
- Feature map fusion based on voxel-level uncertainty from multi-scale decoders.
- Modeling probability and uncertainty using subjective logic and Dirichlet distributions.
- Implementing an uncertainty-based adaptive threshold strategy for optimized segmentation.
Main Results:
- Demonstrated effectiveness of the proposed method across multiple public datasets.
- Achieved reliable segmentation and uncertainty estimation without significant computational overhead.
- Validated the approach's ability to leverage multi-scale semantic information.
Conclusions:
- The proposed method enhances trustworthiness in AI-driven medical image segmentation.
- Subjective logic and uncertainty-based strategies improve segmentation reliability.
- This approach offers a promising solution for overcoming clinical skepticism towards AI.