NABS: non-local automatic brain hemisphere segmentation.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• Accurate segmentation of brain sub-regions is crucial for neurological research and clinical diagnosis.
• Existing methods for magnetic resonance image (MRI) segmentation face challenges in speed, accuracy, and robustness to pathologies.

Purpose of the Study:

• To develop an automated method for segmenting five main brain sub-regions from MRI scans.
• To improve segmentation accuracy, speed, and robustness compared to current state-of-the-art techniques.
• To demonstrate the clinical utility of the proposed method in quantifying brain asymmetry in Alzheimer's disease.

Main Methods:

• A novel multi-label block-wise label fusion strategy was employed.
• The method utilizes a library of pre-labeled brain images in stereotactic space.
• Segmentation was performed on main brain sub-volumes, optimizing computational efficiency.

Main Results:

• The proposed method achieved faster and more accurate segmentations than a state-of-the-art approach.
• Evidence suggests increased robustness against brain pathologies compared to the benchmark method.
• Clinical value demonstrated through improved asymmetry quantification in Alzheimer's disease patients.

Conclusions:

• The developed automatic segmentation method offers significant improvements in speed and accuracy for brain MRI analysis.
• The multi-label block-wise fusion strategy provides enhanced robustness, particularly in the presence of brain pathologies.
• This technique holds promise for clinical applications, including the assessment of neurological disorders like Alzheimer's disease.