Related Experiment Video
Updated: Sep 11, 2025
08:05
Lung CT Segmentation to Identify Consolidations and Ground Glass Areas for Quantitative Assesment of SARS-CoV Pneumonia
Published on: December 19, 2020
14.3K
GRMA-Net: A novel two-stage 3D Semi-supervised Pneumonia Segmentation based on Dual Multiscale Uncertainty Estimation
Jianning Zang1, Yu Gu1, Lidong Yang1
1Inner Mongolia Key Laboratory of Pattern Recognition and Intelligent Image Processing, School of Digital and Intelligent Industry, Inner Mongolia University of Science and Technology, Baotou, 014010, China.
Current Medical Imaging
|August 13, 2025
Summary
This study introduces a novel semi-supervised framework for pneumonia segmentation, enhancing accuracy by effectively using unlabeled data and managing uncertainty. The method shows significant performance improvements in key segmentation metrics.
Area of Science:
- Medical Imaging
- Computer Vision
- Artificial Intelligence
Background:
- Accurate pneumonia lesion segmentation is crucial for diagnosis but challenged by limited labeled data and image uncertainties.
- Existing methods struggle with precise pixel-level labeling and effective utilization of abundant unlabeled data.
Purpose of the Study:
- To propose and evaluate a two-stage semi-supervised segmentation framework for pneumonia.
- To address challenges in high-precision segmentation using dual multiscale uncertainty estimation and graph reasoning.
- To leverage unlabeled data for improved pneumonia lesion segmentation accuracy.
Main Methods:
- Developed a guided supervised training strategy for modeling dual-scale aleatoric uncertainty (AU).
- Implemented a multi-scale noisy pseudo-label correction strategy to mitigate prediction bias.
- Integrated fused feature interaction graph reasoning (FIGR) with attention modules for enhanced feature capture, especially in small infected regions.
Main Results:
- The proposed framework demonstrated improved performance on the MosMedData dataset.
- Achieved performance gains of 1.25% (Dice), 1.03% (Jaccard), 2.98% (NSD), and 0.59% (ADB) over the baseline.
- The method effectively utilizes unlabeled data and uncertainty modeling for better segmentation outcomes.
Conclusions:
- The developed semi-supervised framework significantly enhances pneumonia segmentation by effectively leveraging unlabeled data and managing uncertainties.
- The approach offers potential clinical benefits for pneumonia diagnosis, though generalization and computational efficiency require further investigation.
- Future work will focus on GAN-based data synthesis and architecture optimization to address limitations.
