Disease classification with hippocampal shape invariants.

Area of Science:

• Neuroimaging
• Medical Diagnostics
• Computational Anatomy

Background:

• Alzheimer's disease (AD) diagnosis relies on clinical assessments and neuroimaging, but early and accurate detection remains challenging.
• Hippocampal atrophy is a known biomarker for AD, necessitating advanced methods for its quantitative assessment.

Purpose of the Study:

• To develop and validate a novel method for Alzheimer's disease detection using global shape descriptors of hippocampal surface models.
• To assess the diagnostic performance of the proposed method in distinguishing AD patients from healthy elderly controls.

Main Methods:

• Utilized global shape descriptors to form a 'bag of features' from hippocampal surface models.
• Employed Support Vector Machine (SVM) classification for subject categorization.
• Implemented a leave-one-out cross-validation strategy for performance evaluation.
• Developed a rigid shape registration tool for visualizing shape variations as local displacement maps.

Main Results:

• Achieved 75.5% sensitivity, 87.3% specificity, and 82.1% overall accuracy in classifying 49 Alzheimer's disease (AD) patients and 63 elderly controls.
• Demonstrated that the global shape description provides complementary information beyond simpler shape measures.
• Successfully visualized anatomical meaning of descriptors through local displacement maps.

Conclusions:

• The proposed global shape description method offers a robust and accurate approach for Alzheimer's disease detection.
• This technique enhances diagnostic capabilities by providing novel insights into hippocampal morphological changes.
• The visualization tool aids in understanding the anatomical basis of the classification results.