Automatic white matter lesion segmentation using contrast enhanced FLAIR intensity and Markov Random Field.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• White matter lesions (WMLs) are indicative of various neurological conditions.
• Accurate segmentation of WMLs is crucial for diagnosis and treatment monitoring.
• Manual segmentation is time-consuming and prone to inter-rater variability.

Purpose of the Study:

• To develop and validate a reliable automated method for segmenting white matter lesions (WMLs).
• To improve the accuracy and efficiency of WML detection in brain imaging.
• To provide a tool that assists neuroradiologists in clinical practice.

Main Methods:

• A novel filter was employed to enhance WML intensity.
• A random forest classifier was trained using enhanced intensity, anatomical, and spatial features for initial segmentation.
• A Markov Random Field (MRF) with an edge potential function was utilized for refining segmentation and removing false positives.

Main Results:

• The method achieved Dice similarity indices of 0.76 (severe), 0.73 (moderate), and 0.61 (mild) on the ENVISion dataset.
• Outperformed three state-of-the-art methods in segmentation accuracy on the MICCAI MS lesion challenge dataset.
• Demonstrated robust performance validated against expert manual segmentation.

Conclusions:

• The proposed automated method offers a reliable and accurate approach for WML segmentation.
• This technique has the potential to significantly assist neuroradiologists in clinical WML assessment.
• The findings support the integration of automated tools in neurological diagnostic workflows.