Volumetry of Hippocampus and Amygdala with High-resolution MRI and Three-dimensional Analysis Software: Minimizing the Discrepancies between Laboratories

Abstract

Within the medial temporal lobe, both the hippocampus and amygdala are frequently targeted by researchers and clinicians for volumetric analysis based on magnetic resonance imaging (MRI). However, different data acquisition techniques, analysis software and anatomical boundaries have in the past made it difficult to compare results of MRI studies from different laboratories. In order to reduce these differences, a segmentation protocol was established with 40 healthy normal control subjects recently scanned in our laboratory. Data acquisition was performed with a three-dimensional gradient echo technique, and scans were corrected for non-uniformity and registered into standard stereotaxic space prior to segmentation. Volumetric analysis was performed manually using three-dimensional software that allows simultaneous analysis of sagittal, coronal and horizontal images. Intra-and inter-rater coefficients yielded correlation coefficients comparable with other protocols. The hippocampal volume was larger in the right hemisphere (3324 versus 3208 mm 3 ), while no interhemispheric differences for the amygdala (1154 versus 1160 mm 3 ) could be observed. Most importantly, results from recent segmentation protocols for hippocampus and amygdala seem to approach each other with regard to mean volumes and interhemispheric differences. This indicates that the advances in scanning technique, volume preparation and segmentation protocols allow a more precise definition of medial temporal lobe structures with MRI, and that results for mean volumes for hippocampus and amygdala from different laboratories will eventually become comparable.

MR-based in vivo hippocampal volumetrics: 1. Review of methodologies currently employed

“…The protocols may differ in a number of factors related to image acquisition, image processing, and anatomical guidelines, which are important for accurate hippocampal volume determination, namely image acquisition parameters, magnetic field strength, the number of slices assessed and the thickness of slices, hippocampal orientation correction, volumetric correction, software used, inter-rater reliability, and anatomical boundaries of the hippocampus. [43][44][45] These differences are discussed in greater detail below.…”

Patch-based segmentation using expert priors: Application to hippocampus and ventricle segmentation

“…The MR images were manually segmented by an expert directly into stereotaxic space. For each subject, the HC label was manually defined using the protocol described by Pruessner et al (2000). The resulting segmentations obtained an intraclass reliability coefficient (ICC) of 0.900 for inter-rater reliability (4 raters) and 0.925 for intra-rater reliability (5 repeats).…”

“…As usual in quantitative MR analysis, manual segmentation is considered the gold standard (Pruessner et al, 2000). For both datasets, the impact of the patch size, search volume size, and number of training subjects was studied.…”

Patterns of neuropsychological impairment in MCI patients with small subcortical infarcts or hippocampal atrophy

“…Hippocampal volumetry was evaluated by means of the segmentation protocol proposed by Pruessner et al (2000) in a study conducted on 40 normal adults. This method is based on a manual segmentation of the hippocampus and on the use of a three-dimensional software that allows simultaneous analysis of sagittal, coronal and horizontal images.…”

“…The images were analyzed in stereotactic space. Because subjects studied by Pruessner et al (2000) were younger than our patients, and hippocampal volume decreases with advancing age in healthy adults, we adopted the cutoff points derived by a recent study of Lupien et al (2007). In this study, several age ranges were considered.…”