Robust skull stripping using multiple MR image contrasts insensitive to pathology.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• Accurate brain extraction (skull-stripping) is crucial for neuroimage analysis.
• Existing methods often struggle with pathological brains and diverse MRI contrasts.
• T1-weighted MRI is common, but other sequences offer complementary tissue information.

Purpose of the Study:

• To develop and validate a novel multi-contrast brain extraction method (MONSTR).
• To improve skull-stripping accuracy, especially in the presence of brain pathologies.
• To demonstrate robustness across different MRI acquisition protocols and scanners.

Main Methods:

• Proposed a sparse patch-based Multi-cONtrast brain STRipping (MONSTR) method.
• Utilized non-local patch information from multiple MR contrasts (T1-w, T2-w) and atlases.
• Compared MONSTR against four state-of-the-art methods on six diverse datasets.

Main Results:

• MONSTR demonstrated significant improvements in skull-stripping accuracy over competing methods.
• The method showed robust performance on both healthy subjects and patients with various pathologies.
• Multi-contrast approach maintained accuracy across different acquisition types and scanners, even with normal brain atlases.

Conclusions:

• MONSTR offers a robust and accurate solution for brain extraction, particularly in challenging pathological cases.
• The multi-contrast framework enhances skull-stripping by integrating complementary MRI information.
• MONSTR is applicable even when pathological brain atlases are unavailable, broadening its utility.