White matter lesion extension to automatic brain tissue segmentation on MRI

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• Accurate segmentation of brain tissues and white matter lesions (WML) is crucial for neurological disorder diagnosis and research.
• Manual segmentation is time-consuming and prone to inter-observer variability.
• Existing automated methods often require manual adjustments or lack comprehensive lesion detection.

Purpose of the Study:

• To develop and validate a fully automated method for segmenting cerebrospinal fluid (CSF), gray matter (GM), and white matter (WM).
• To extend this method for accurate white matter lesion (WML) segmentation.
• To evaluate the accuracy and reliability of the automated segmentation against manual segmentation.

Main Methods:

• An atlas-based k-nearest neighbor classifier was employed for multi-modal magnetic resonance imaging (MRI) data.
• Brain atlases were registered to subject data for classifier training.
• White matter lesion segmentation utilized fluid-attenuated inversion recovery (FLAIR) scans, with a threshold determined by GM segmentation and false positives removed by spatial constraints.

Main Results:

• Visual validation on 209 subjects showed 98% accuracy for brain tissue segmentation and 97% for WML segmentation.
• Quantitative evaluation on a subset of subjects demonstrated that automated segmentation accuracy closely approximates inter-observer variability.
• The method successfully identified and excluded false positive lesions within the white matter.

Conclusions:

• The developed automated method provides accurate and reliable segmentation of brain tissues and white matter lesions.
• This technique significantly reduces the time and variability associated with manual segmentation.
• The high accuracy suggests its potential for clinical application in diagnosing and monitoring neurological conditions.