Discriminating Alzheimer's Patients from Cognitively Normal Older Adults Based on Hippocampal Volumes - Voxel-Based Morphometry with DARTEL and Standard Registration Versus Manual Volumetry

Hippocampal atrophy on Magnetic Resonance Imaging (MRI) has been increasingly used as a supportive feature for the diagnosis of Alzheimer's patients. We attempt to compare voxel-based morphometry (VBM) with DARTEL and standard registration versus manual hippocampal volumetry in discriminating Alzheimer's disease (AD) patients from cognitively normal older adults. Participants were 20 cognitively normal elderly subjects and 19 AD patients who met the criteria of probable AD according to NINCDS-ADRDA. High resolution T1-weighted (T1W) 3D volumetric Fast Field Echo sequence was obtained using 3.0-Tesla scanner. Bilateral manual hippocampal volumetry (ANALYZE 7.0 software) and VBM hippocampal region-of-interests (using hippocampal masks MARINA) were generated with DARTEL and standard registration (Statistical Parametric Mapping SPM5 software). Normalization with total intracranial volumes was employed for both methods. All normalized hippocampal measurements showed significant reduction (20–30%; p < 0.001) in AD compared to controls. Logistic regression analysis also showed significant effects (odds ratios ranged from 88.2% to 94.0%) of all normalized measurements in predicting AD incidence after adjusting for age, gender, and education. The overall prediction accuracies of manual RH and LH volumes, standard RH-ROI and LH-ROI VBM, DARTEL RH-ROI and LH-ROI VBM were 87.2%, 84.6%, 87.2%, 76.9%, 87.2%, and 87.2%, respectively. VBM with DARTEL and standard registration have similarly high efficacies as manual hippocampal volumetry in discriminating AD from cognitively normal elderly adults.