Detecting global and local hippocampal shape changes in Alzheimer's disease using statistical shape models

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Anatomy

Background:

• Alzheimer's disease (AD) early affects the hippocampus.
• Structural magnetic resonance (MR) imaging can assess hippocampal morphology changes in AD.
• Statistical Shape Models (SSMs) describe population-level shape variations.

Purpose of the Study:

• To utilize SSMs for classifying Alzheimer's disease (AD) from normal controls (NC) using hippocampal shape features.
• To investigate if feature selection enhances the discriminative power of SSMs for AD detection.
• To correlate identified shape predictors with cognitive measures.

Main Methods:

• Applied SSMs to hippocampal shape data from AD and NC cases.
• Used Hotelling's T2 test for landmark subset selection.
• Employed principal component analysis (PCA) for variation mode extraction.
• Evaluated classification performance using bagged support vector machines (SVMs).

Main Results:

• Selecting landmarks with better separation between AD and NC significantly increased SSM discrimination power.
• Extracted predictors showed stronger correlations with memory assessments (Logical Memory, AVLT, ADAS).
• The refined SSM approach improved AD classification accuracy.

Conclusions:

• SSMs, particularly when focused on discriminative landmarks, are effective for classifying Alzheimer's disease.
• Hippocampal shape analysis provides valuable biomarkers for AD detection and progression monitoring.
• This method shows potential for early AD diagnosis and understanding disease-related memory decline.