Direct mapping of hippocampal surfaces with intrinsic shape context

Area of Science:

• Neuroimaging
• Computational Anatomy
• Medical Image Analysis

Background:

• Hippocampal (HC) shape analysis is crucial for understanding neurological conditions like Alzheimer's disease (AD).
• Establishing point correspondences across different HC surfaces is challenging but essential for localized shape analysis.
• Existing methods may lack automation or pose invariance, limiting their clinical applicability.

Purpose of the Study:

• To develop a novel, automatic, and pose-invariant method for mapping hippocampal surfaces.
• To enable localized shape analysis of the hippocampus for studying neurodegenerative diseases.
• To investigate temporal changes in hippocampal asymmetry in Alzheimer's disease.

Main Methods:

• Defined a novel geometric feature: intrinsic shape context, to capture global HC shape characteristics.
• Developed an automatic algorithm to detect stable landmark curves across populations.
• Solved direct mapping between source and target HC surfaces using harmonic energy minimization and partial differential equations on implicit surfaces.

Main Results:

• The direct mapping method is automatic and invariant to HC shape pose.
• Applied the method to study temporal changes in hippocampal asymmetry in 12 AD patients and 14 normal controls.
• Observed significantly different trends in temporal changes of hippocampal asymmetry between the AD group and controls.

Conclusions:

• The proposed direct mapping method effectively establishes correspondences on HC surfaces for shape analysis.
• The findings indicate altered temporal dynamics of hippocampal asymmetry in Alzheimer's disease.
• The method's flexibility was demonstrated through spherical mapping and confirmed by spherical harmonics (SPHARM) analysis.