Towards a Holistic Cortical Thickness Descriptor: Heat Kernel-Based Grey Matter Morphology Signatures

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Anatomy

Background:

• Accurate analysis of brain magnetic resonance imaging (MRI) is crucial for understanding neurological disorders.
• Existing methods may not fully exploit volumetric morphological information, limiting statistical power in detecting subtle brain changes.

Purpose of the Study:

• To propose a novel heat kernel-based regional shape descriptor for enhanced volumetric morphological analysis of brain MRI.
• To improve the statistical power and classification accuracy for neurological conditions like Alzheimer's disease.

Main Methods:

• Utilized graph spectrum and heat kernel theory on tetrahedral meshes to capture volumetric geometry.
• Employed volumetric Laplace-Beltrami operator for point pair correspondence between brain matter boundaries.
• Developed multi-scale grey matter morphology signatures based on random walk transition probabilities.
• Applied point distribution model and sparse linear discriminant analysis for feature selection and dimensionality reduction.

Main Results:

• The proposed descriptor outperformed FreeSurfer's cortical thickness in classifying Alzheimer's disease and mild cognitive impairment.
• Demonstrated improved classification accuracies using the selected morphology feature set.
• The physics-based volumetric structure features showed stronger statistical power compared to traditional methods.

Conclusions:

• The proposed heat kernel-based shape descriptor effectively captures volumetric grey matter morphology.
• This approach offers enhanced statistical power for MRI-based analysis, particularly for neurodegenerative diseases.
• The method shows promise for more accurate and sensitive detection of brain structural changes.