Atlas generation for subcortical and ventricular structures with its applications in shape analysis.

Area of Science:

• Neuroimaging
• Computational Anatomy
• Medical Image Analysis

Background:

• Atlas-driven morphometric analysis is crucial for studying anatomical shape variation in neuroimaging.
• Local coordinate representation aids in understanding anatomical observations across populations.

Purpose of the Study:

• To present a procedure for generating an atlas of subcortical and ventricular brain structures.
• To validate the atlas's representativeness and utility in detecting shape variations and classification.

Main Methods:

• Utilized the large deformation diffeomorphic metric atlas generation algorithm.
• Constructed the atlas from manually labeled brain volumes of 41 subjects from the Open Access Series of Imaging Studies (OASIS) database, including diverse age groups and dementia patients.
• Included subcortical structures: amygdala, hippocampus, caudate, putamen, globus pallidus, thalamus, and lateral ventricles.

Main Results:

• The generated atlas was shown to be representative of the population based on metric distance.
• Using the estimated atlas potentially increases statistical power for identifying group shape differences compared to a single-subject atlas.
• Metric distances to within-class atlases created a feature space for shape-based classification.

Conclusions:

• The developed atlas generation procedure is effective for subcortical and ventricular structures.
• The atlas improves statistical power in shape variation detection and facilitates shape-based classification in neuroimaging research.