Related Experiment Video
Updated: Apr 22, 2026
Quantitative Magnetic Resonance Imaging of Skeletal Muscle Disease
Published on: December 18, 2016
A modified method for MRF segmentation and bias correction of MR image with intensity inhomogeneity
Mei Xie1, Jingjing Gao, Chongjin Zhu
1University of Electronic Science and Technology of China, No. 2006, Xiyuan Ave, West Hi-Tech Zone, Chengdu, 611731, Sichuan, China.
Abstract:
Markov random field (MRF) model is an effective method for brain tissue classification, which has been applied in MR image segmentation for decades. However, it falls short of the expected classification in MR images with intensity inhomogeneity for the bias field is not considered in the formulation. In this paper, we propose an interleaved method joining a modified MRF classification and bias field estimation in an energy minimization framework, whose initial estimation is based on k-means algorithm in view of prior information on MRI. The proposed method has a salient advantage of overcoming the misclassifications from the non-interleaved MRF classification for the MR image with intensity inhomogeneity. In contrast to other baseline methods, experimental results also have demonstrated the effectiveness and advantages of our algorithm via its applications in the real and the synthetic MR images.
Insights
This study introduces an improved Markov random field (MRF) model for brain tissue classification in MRI scans. The novel method enhances accuracy by accounting for intensity inhomogeneity, improving MR image segmentation.
Area of Science:
- Medical Imaging
- Computational Neuroscience
- Image Processing
Background:
- Markov random field (MRF) models are widely used for brain tissue classification in MRI segmentation.
- Traditional MRF models struggle with intensity inhomogeneity caused by bias fields, leading to misclassifications.
- Existing methods often do not adequately address the bias field in MR image analysis.
Purpose of the Study:
- To develop an improved MRF model for brain tissue classification that effectively handles intensity inhomogeneity.
- To integrate bias field estimation directly into the MRF framework for more accurate segmentation.
- To enhance the robustness of MR image segmentation in the presence of bias fields.
Main Methods:
- An interleaved method combining a modified MRF classification and bias field estimation was developed.
- The framework utilizes an energy minimization approach.
- Initial estimations were derived using the k-means algorithm with prior MRI information.
Main Results:
- The proposed interleaved method significantly overcomes misclassifications common in non-interleaved MRF approaches for inhomogeneous MR images.
- Experimental results on both real and synthetic MR images demonstrate superior performance compared to baseline methods.
- The algorithm shows effectiveness in improving brain tissue classification accuracy.
Conclusions:
- The proposed interleaved MRF classification and bias field estimation method offers a robust solution for MR image segmentation.
- This approach effectively addresses intensity inhomogeneity, leading to more accurate brain tissue classification.
- The method presents a significant advancement over traditional MRF techniques for analyzing MR images.
