Fast and robust extraction of hippocampus from MR images for diagnostics of Alzheimer’s disease


J. Lötjönen, R. Wolz, J. Koikkalainen, V. Julkunen, L. Thurfjell, R. Lundqvist, G. Waldemar, H. Soininen, D. Rueckert, The Alzheimer’s Disease Neuroimaging Initiative

Rapid, robust definition of hippocampus volume is helpful for clinical decision making in Alzheimer’s disease. In this study, a method for automatic computation of hippocampus volume using a modified multi-atlas segmentation framework was introduced and then assessed in 1000 cases from three cohorts. The rate of correct classification of Alzheimer’s disease and cognitively normal cases was 80%, while classification of stable and progressive mild cognitive impairment cases achieved an accuracy of 65%. The method took ~2 minutes to compute on a standard laptop computer.