Reliability of in vivo volume measures of hippocampus and other brain structures using MRI

Abstract

Volume reductions of the hippocampus are associated with Alzheimer's disease, schizophrenia, and epilepsy. We used clinically available MRI methods (2D acquisition; inversion recovery and calculated T2 images; 3 mm contiguous slices) that optimize image contrast, quality, and resolution and standardized positioning protocols to maximize the in vivo accuracy (test-retest reliability) of brain volume measurements in volunteers who were scanned two or three times. Volunteers were scanned in the same MRI instrument (intrascanner reliability) as well as in two different instruments (interscanner reliability). A single rater obtained brain volume measures of seven contiguous slices centered on the anterior commissure. The in vivo intrascanner reliability for measures of anterior hippocampus and ventricular volumes was very good, with reliability coefficients [intraclass r (rxx)] ranging between .855 and .997, and a median coefficient of variation (CV) of 6.4%. Reliability was good for amygdala (rxx of .740 and .764) and for total frontal and temporal lobe volumes and white matter volume measures (rxx ranging between .640 and .823, median coefficient of variation was 3.2%). Overall, interscanner reliability was also good. We discuss the implications of our results relative to the possible clinical utility of hippocampal quantification and the feasibility of prospective studies aimed at quantifying progressive neurodegeneration.