Voxel-based morphometry from conventional T1-weighted images has proved effective to quantify Alzheimer's disease (AD) related brain atrophy and to enable accurate automated classification of AD patients, mild cognitive impaired (MCI) patients and elderly controls. Little is known, however, about the classification power of volume-based morphometry, where features of interest consist of a few brain structure volumes (e.g. hippocampi, lobes, ventricles…) as opposed to about a million voxel-wise gray matter (GM) concentrations. This work aims to experimentally compare voxel-based and volume-based brain morphometry for automated disease classification.