Detection, visualization and animation of abnormal anatomic structure with a deformable probabilistic brain atlas...

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Anatomy

Background:

• Accurate quantification of anatomical variation in the human brain is crucial for understanding neurological diseases and development.
• Existing methods often lack the precision to capture subtle, distributed patterns of anatomical deviation.

Purpose of the Study:

• To develop and validate a comprehensive probabilistic atlas of the human brain using high-dimensional vector field transformations.
• To enable the detection and quantification of deviations from normal anatomy in individual 3D brain images.

Main Methods:

• Utilized high-dimensional vector field transformations to elastically deform individual brain scans into structural correspondence with a reference population.
• Developed a probability space based on anisotropic Gaussian random fields to model anatomical variability.
• Generated color-coded probability maps indicating the likelihood of anatomical points being unusually situated.

Main Results:

• The technique successfully generated detailed, color-coded probability maps for new subjects.
• Demonstrated the ability to detect and quantify subtle anatomical variations in 3D MRI and cryosection volumes from subjects with tumors and Alzheimer's disease.
• Simulated and animated the dynamic effects of anatomical deformations on regional probability maps.

Conclusions:

• The developed probabilistic atlas provides a powerful tool for analyzing anatomical variability in the human brain.
• Applications include disease detection, mapping functional and histological data, and studying developmental changes.
• This method offers a robust approach for quantifying local shape changes in 3D medical images.