No abstract

We trained our approaches using T1-weighted brain MRIs from 30 subjects [10 normal elderly, 10 mild cognitive impairment (MCI), and 10 Alzheimer's disease (AD)], and tested on an independent set of 40 subjects (20 normal, 20 AD). Here we registered all of the brain images into the same stereotaxic space. Then we define a bounding box around the training hippocampal plus some neighborhood voxels). We compared two automated methods for hippocampal segmentation using different machine learning algorithms: 1) support vector machines (SVM) with manual feature selection, 2) hierarchical SVM with automated feature selection (Ada-SVM). After surface reconstruction, we computed significance maps, and overall corrected pvalues, for the 3-D profile of shape differences between AD and normal subjects. We assessed each approach's accuracy relative to manual segmentations, and its power to map AD effects. Our AdaBoost and Ada-SVM segmentations compared favorably with the manual segmentations. We also evaluated how segmentation accuracy depended on the size of the training set, providing practical information for future users of this technique.