Semi-Supervised Learning in Medical MRI Segmentation: Brain Tissue with White Matter Hyperintensity Segmentation...

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Artificial Intelligence

Background:

• White-matter hyperintensity (WMH) is a key biomarker for cerebrovascular disease and Alzheimer's disease.
• FLAIR MRI is valuable for WMH detection but lacks normal tissue information, necessitating multi-modal analysis.
• Clinical settings often limit MRI availability to FLAIR sequences.

Purpose of the Study:

• To develop a semi-supervised learning method for full brain segmentation using only FLAIR MRI.
• To assess the reliability and accuracy of the proposed segmentation method for WMH analysis.

Main Methods:

• A semi-supervised learning approach was employed for brain segmentation.
• Segmentation results were compared against reference labels derived from FreeSurfer segmentation on T1w MRI.
• Relative volume differences and Dice similarity coefficients were used for evaluation.

Main Results:

• The proposed method demonstrated high reliability in segmenting brain structures using FLAIR MRI.
• Quantitative comparisons showed good agreement between the proposed method's results and reference segmentations.

Conclusions:

• The semi-supervised learning method offers a reliable approach for brain segmentation using FLAIR MRI, even without T1w data.
• This method has potential for broader applications in MRI-based brain tissue segmentation and analysis of neurological conditions.