Measuring the volumes and thickness of hippocampal subfields in vivo using automatic segmentation of T2-weighted MRI: A pilot evaluation study

Purpose: We demonstrate that meaningful cross-sectional differences in the subfields of the hippocampal formation (HF) can be detected in patients with mild cognitive impairment (MCI) using automatic segmentation of focal T2-weighted MRI. Boundaries between the layers and subfields of the HF are virtually indistinguishable in routine 1 mm T1-weighted MRI at 1.5 and 3 Tesla. However, mounting evidence suggests that focal fast spin echo MRI tailored for HF imaging (HF-MRI for short) provides sufficient contrast between the HF layers to allow reliable measurement of change in HF subfields. Cross-sectional differences in HF subfield volumes measured in HF-MRI have recently been shown to be consistent with the topography of neuropathology in several brain disorders, including AD and MCI [1,2,3]. However, until now, analysis of HF subfields in HF-MRI required prohibitively expensive manual segmentation (3-4 hours per image). Our group developed an automatic method for subfield segmentation in HF-MRI. Its accuracy for the larger subfields (CA1, dentate gyrus) is comparable to the best results reported for whole HF segmentation. In the present work, we evaluate the ability of this method to detect subfield-specific differences between MCI patients and matched healthy controls. To our knowledge, these are the first reported results of subfield-specific HF morphometry using automatic analysis of HF-MRI, a modality recognized as superior for HF imaging.