High-resolution random mesh algorithms for creating a probabilistic 3D surface atlas of the human brain

Area of Science:

• Neuroimaging
• Computational Anatomy
• Medical Image Analysis

Background:

• Individual variations in brain geometry complicate standardized neuroanatomical atlases.
• Quantifying deviations from normal brain anatomy is challenging due to complex variations.

Purpose of the Study:

• To develop and implement a 3D statistical method for creating probabilistic surface atlases of the human brain.
• To automatically generate detailed probability maps of new subjects' anatomy, quantifying subtle deviations.

Main Methods:

• Utilized connected systems of parametric meshes to model internal sulcal structures in both brain hemispheres.
• Developed a probability space of random transformations based on Gaussian random field theory to model anatomical variability.
• Computed probability density functions to establish confidence limits on surface variation.

Main Results:

• Successfully modeled the internal course of key sulci, including the parieto-occipital, calcarine, cingulate, marginal, and supracallosal sulci.
• Generated a family of surface maps encoding statistical properties of local anatomical variation within individual sulci.
• Provided a quantitative framework for assessing intersubject variations in brain architecture.

Conclusions:

• The developed surface mapping and probabilistic techniques enable the generation of anatomical templates that preserve quantitative information on brain architecture variations.
• This approach facilitates the integration of functional and anatomical data across subjects and modalities.
• The method supports the development of expert diagnostic systems for neuroimaging analysis.