Detecting and visualizing differences in brain structures with SPHARM and functional data analysis.

Area of Science:

• Neuroscience
• Biostatistics
• Machine Learning

Background:

• Accurate classification of brain structures is crucial for understanding neurological processes.
• Existing methods may lack interpretability or optimal predictive power in complex neuroimaging analyses.

Purpose of the Study:

• To introduce a novel procedure for classifying brain structures using SPHARM (Superquadric Primitive-based Representation of Meshes).
• To evaluate the performance and interpretability of the proposed method against established techniques in a neuroeducation context.

Main Methods:

• Utilized functional principal component analysis (fPCA) or functional independent component analysis (fICA) for dimension reduction.
• Employed stepwise variable selection coupled with linear discriminant classification.
• Applied the procedure to analyze left and right putamen structures in 33 participants to study mathematical reversal errors.

Main Results:

• The proposed method demonstrated outstanding predictive performance compared to well-known classification techniques.
• The procedure yielded an interpretable linear discriminant function for analyzing 3D brain structures.

Conclusions:

• The new classification procedure offers a powerful and interpretable tool for neuroimaging research.
• This method advances the analysis of brain structure variations and their relation to cognitive functions like mathematical problem-solving.