Automatic discrimination of Alzheimer's disease from normal aging based on MRI hippocampal shape analysis

Objective: Based on the MRI hippocampal shape analysis, to study the regional pattern differences between Alzheimer's disease (AD) and normal aging, and build effective classifiers to assist the diagnosis of AD. Methods: Conventional MRI were performed in 19 AD patients and 20 age- and gender-matched healthy controls. Then hippocampal surface models were constructed and regional surface deformations were characterized by surface-based measures. Finally, effective classifiers were built to discriminate AD from normal aging. Results: The accuracy of automatic recognition were 82.1% and 92.3% by using leave-one-out cross-validation, and similarly the average accuracy of randomized 3-fold cross-validation by 100 times were 82.5% and 87.2% resulted by right and left hippocampus respectively. Conclusion: Hippocampal. shape analysis is effective for the automatic recognition of AD.