Fast and accurate pseudo multispectral technique for whole-brain MRI tissue classification

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Biomedical Engineering

Background:

• Accurate brain tissue classification is crucial for neurological studies.
• Existing methods struggle with low-contrast regions and noise in MRI scans.
• Standard techniques include gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) segmentation.

Purpose of the Study:

• To introduce a novel Pseudo Multispectral Classification (PMC) technique for enhanced brain tissue segmentation.
• To improve accuracy in low-contrast and high-noise Magnetic Resonance Imaging (MRI) data.
• To compare PMC against established segmentation software (FSL, SPM8, K-means).

Main Methods:

• Converted grayscale T1-weighted MPRAGE images to multispectral CIE LAB space.
• Applied a novel contrast enhancement technique for improved tissue separation.
• Utilized an optimized iterative K-means clustering algorithm for classification.
• Validated the approach using simulated and in vivo human MRI data.

Main Results:

• PMC demonstrated superior performance in low signal-to-noise ratio (SNR) conditions compared to FSL, SPM8, and K-means.
• In simulated studies, PMC achieved a mean Jaccard Index (JI) of 0.74.
• In vivo human studies showed PMC yielding a mean JI of 0.92, outperforming other methods.

Conclusions:

• The proposed PMC technique offers a robust solution for automatic brain tissue classification.
• PMC maintains high accuracy even in the presence of significant image noise.
• This method holds significant potential for advancing neurological research and diagnostics.