Accurate cortical tissue classification on MRI by modeling cortical folding patterns.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• Accurate tissue classification is essential for Magnetic Resonance Imaging (MRI) morphometry.
• Existing automated methods struggle with regional intensity variations and artifacts.
• Disparities in tissue composition and cortical folding patterns challenge global histogram-based approaches.

Purpose of the Study:

• To develop and evaluate a novel anatomy-driven method for improved brain tissue classification in MRI.
• To address limitations of current automated methods in handling regional intensity variations and complex brain structures.
• To enhance the accuracy of white matter-gray matter (GM) interface delineation for morphometric studies.

Main Methods:

• A new anatomy-driven approach was developed, extracting brain parcels based on cortical folding.
• Nonparametric mean shift clustering was applied to classify each extracted parcel.
• The method was evaluated on manually labeled MRI datasets acquired at 3.0 Tesla (n=15) and 1.5 Tesla (n=20).

Main Results:

• The proposed method achieved high tissue classification accuracy (Dice index >97.6% at 3.0T, >89.2% at 1.5T).
• It outperformed established methods like SPM8, FSL-FAST, and a cube-based local classifier.
• Superior white matter-gray matter interface classification was observed, especially in central and occipital cortices.

Conclusions:

• The anatomy-driven local classification algorithm significantly improves cortical boundary definition.
• The method demonstrates excellent accuracy even without intensity inhomogeneity correction.
• This approach holds potential for advancing morphometric inference and biomarker discovery in neuroimaging.