Automatic calculation of hippocampal atrophy rates using a hippocampal template and the boundary shift integral

Abstract

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) mm3 loss per year, respectively. In AD patients these rates were 174.6 (106.5) 159.4 (101.2) and 172.1 (123.1) mm3 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 20 min per hippocampal pair on a 3.4 GHz Intel Xeon server, whereas manual delineation of each hippocampal pair took approximately 90 min of operator-intensive labour. This method may be useful diagnostically or in studies where analysis of many scans may be required.