We determined predictors of conversion to Alzheimer's disease (AD) from mild cognitive impairment (MCI) with automated magnetic resonance imaging (MRI) regional cortical volume and thickness measures. One hundred amnestic MCI subjects, 118 AD patients, and 94 age-matched healthy controls were selected from AddNeuroMed study. Twenty-four regional cortical volumes and 34 cortical thicknesses were measured with automated image processing software at baseline. Twenty-one subjects converted from MCI to AD determined with the cognitive tests at baseline and 1 year later. The hippocampus, amygdala, and caudate volumes were significantly smaller in progressive MCI subjects than in controls and stable MCI subjects. The cortical volumes achieved higher predictive accuracy than did cognitive tests or cortical thickness. Combining the volumes, thicknesses, and cognitive tests did not improve the accuracy. The volume of amygdala and caudate were independent variables in predicting conversion from MCI to AD. We conclude that regional cortical volume measures are more powerful than those common cognitive tests we used in identifying AD patients at the very earliest stage of the disease.