We describe a method of automatically calculating hippocampal atrophy rates on T1-weighted MR images without manual delineation of hippocampi. This method was applied to a group of Alzheimer's disease (AD) (n=36) and control (n=19) subjects and compared with manual methods (manual segmentation of baseline and repeat-image hippocampi) and semi-automated methods (manual segmentation of baseline hippocampi only). In controls, mean (S.D.) atrophy rates for manual, semi-automated, and automated methods were 18.1 (53.5), 15.3 (50.2) and 11.3 (50.4)mm 3 loss per year, respectively. In AD patients these rates were 174.6 (106.5) 159.4 (101.2) and 172.1 (123.1)mm 3 loss per year, respectively. The automated method was a significant predictor of disease (p=0.001) and gave similar group discrimination compared with both semi-automated and manual methods. The automated hippocampal analysis in this small study took approximately 20min per hippocampal pair on a 3.4GHz Intel Xeon server, whereas manual delineation of each hippocampal pair took approximately 90min of operator-intensive labour. This method may be useful diagnostically or in studies where analysis of many scans may be required.