Surface-based labeling of cortical anatomy using a deformable atlas

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Anatomy

Background:

• Accurate segmentation and labeling of the human brain's cortical surface are crucial for understanding neurological function and disease.
• Existing methods often require manual intervention, which is time-consuming and prone to inter-observer variability.
• Automated approaches are needed to improve efficiency and reproducibility in neuroimaging studies.

Purpose of the Study:

• To develop and validate a computerized method for automatic identification and labeling of the cortical surface in 3D magnetic resonance (MR) brain images.
• To leverage a deformable brain atlas to accurately map anatomical labels onto individual brain structures.
• To establish a robust pipeline for processing MR images and extracting detailed cortical surface information.

Main Methods:

• A pre-labeled brain atlas is modeled as an elastic object capable of deforming to match image data.
• Image preprocessing involves boundary detection and morphological operations to extract the brain and sulci, creating a smoothed surface representation.
• Energy-minimizing deformable models, parameterized by 3D bicubic B-spline surfaces, are employed to locate image features accurately.
• A conjugate gradient method is used to minimize an energy function that attracts model points to image-based cortical fissures (sulci) and the brain surface.
• Labels from the deformed atlas are propagated to the high-resolution brain surface.

Main Results:

• The developed method successfully automates the process of finding and labeling the cortical surface in 3D MR images.
• Deformable models accurately converge to the smoothed brain surface, guided by image features and an elastic atlas.
• The energy function effectively attracts model points to relevant anatomical landmarks like sulci.
• Labels are reliably propagated from the atlas to the target brain, enabling automated anatomical labeling.

Conclusions:

• The described computerized method offers an efficient and automated solution for cortical surface labeling in 3D MR brain images.
• The use of a deformable atlas and energy-minimizing models provides accurate convergence to brain structures.
• This approach has the potential to significantly enhance the speed and consistency of neuroimaging analysis.
• Automated cortical surface labeling facilitates large-scale studies of brain anatomy and its relation to function and disease.