Related Experiment Video
Updated: Jun 27, 2026
06:48
Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
Published on: January 7, 2019
Automatic segmentation of magnetic resonance images using a decision tree with spatial information
Wen-Hung Chao1, You-Yin Chen, Sheng-Huang Lin
1Department of Electrical and Control Engineering, National Chiao Tung University, No. 1001, Ta-Hsueh Rd., Hsinchu 300, Taiwan, ROC.
Summary
This study introduces an automatic decision tree method for segmenting brain tissues in MRI scans. The technique accurately identifies anatomical structures, offering an easier approach to neuroimage analysis.
Area of Science:
- Neuroimaging
- Medical Image Analysis
- Computer Vision
Background:
- Accurate segmentation of brain tissues in Magnetic Resonance (MR) images is crucial for diagnosing neurological conditions.
- Existing methods may struggle with noise and inhomogeneities present in MR images.
- Automating this process can improve efficiency and consistency in neuroimage analysis.
Purpose of the Study:
- To develop and evaluate an automatic segmentation method for classifying brain tissues in MR images using a decision tree algorithm.
- To assess the impact of different spatial information features on segmentation accuracy.
- To compare the proposed method's performance against established metrics like Hausdorff distance, Computer to Observer Difference (COD), and Interobserver Difference (IOD).
Main Methods:
- An automatic decision tree algorithm was employed for brain tissue segmentation.
- Spatial information, including general gray level (G), spatial gray level (S), and 2D wavelet transform (W), was combined in Euclidean (x, y) and polar (r, theta) coordinate systems.
- Segmentation accuracy was evaluated using phantom and simulated brain MR images with varying noise levels and inhomogeneities.
- Boundary detection was performed using Hausdorff distance to compare with COD and IOD for gray matter (GM), white matter (WM), and all areas (ALL).
Main Results:
- The decision tree method achieved high segmentation accuracy rates, with phantom images reaching up to 0.9999 and simulated brain images reaching 0.9532.
- Segmentation accuracy was highest when general gray level (G) information was included, followed by spatial gray level (S), and then wavelet transform (W).
- The mean COD for gray matter segmentation was comparable to established interobserver differences, indicating reliable performance.
Conclusions:
- The proposed automatic segmentation method based on a decision tree algorithm offers an effective and straightforward approach for segmenting brain tissues in MR images.
- Incorporating general gray level information significantly enhances segmentation accuracy.
- This method shows promise for improving the identification of human anatomical structures in neuroimaging studies.
