Related Experiment Video
Updated: Aug 31, 2025
06:48
Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
Published on: January 7, 2019
9.0K
An Efficient Optimization Approach for Glioma Tumor Segmentation in Brain MRI
Zeynab Barzegar1, Mansour Jamzad2
1Sharif University of Technology, Tehran, Iran.
Journal of Digital Imaging
|August 22, 2022
Summary
This study introduces a novel semi-supervised framework for precise glioma segmentation, achieving 94% accuracy. The method accurately identifies whole tumor, tumor core, and enhancing tumor regions, outperforming existing state-of-the-art techniques.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computational Biology
Background:
- Glioma segmentation is challenging due to tumor heterogeneity.
- Existing methods like multi-atlas and machine learning have limitations.
- Accurate segmentation is crucial for effective glioma treatment.
Purpose of the Study:
- To develop a semi-supervised framework for accurate multi-label glioma segmentation.
- To overcome limitations of existing atlas-based and machine learning approaches.
- To improve the precision of identifying distinct glioma regions.
Main Methods:
- Proposed a semi-supervised unified framework using Markov Random Field (MRF) energy optimization.
- Formulated segmentation as an optimization problem on a parametric graph.
- Applied the framework to publicly available BRATS datasets (low- and high-grade gliomas).
Main Results:
- Achieved competitive performance compared to state-of-the-art methods.
- Obtained the best Dice score for segmenting "whole tumor" (WT), "tumor core" (TC), and "enhancing active tumor" (ET) regions.
- Demonstrated high accuracy with a mean Dice score of 94%.
Conclusions:
- The proposed MRF-based framework offers precise and accurate glioma segmentation.
- The method effectively maps segmentation to a graphical optimization model.
- This approach provides a robust solution for challenging glioma segmentation tasks.
