Parcellation of the human amygdala using 7T structural MRI

Introduction The amygdala is located deep within the temporal lobe. Mainly constituted of grey matter, it has low diffusion anisotropy and relatively little MRI contrast, and is thus challenging to study in vivo. MRI scans at normal field strengths show a region of almost featureless grey matter. Histologically, the amygdala is divided into several subnuclei. Each has multiple connections to other brain areas, including higher cognitive circuits. The in vivo study of these subdivisions could facilitate the neuroscientific understanding of human emotion and several diseases. The higher sensitivity available at 7T, and the combination of different MRI contrasts, may allow the discrimination of amygdaloid subregions. Methods 9 healthy subjects underwent structural scanning on a 7 Tesla Magnetom MRI system (Siemens, Erlangen) with an 8-channel head array coil (RAPID Biomedical, Rimpar). The whole brain was acquired with an MP-RAGE sequence (TR=3000 ms, TE=2.95 ms, TI=1100 ms, voxel size: (1.2 mm)3). Additionally 30 coronal slices (FOV 230×230 mm2, matrix 192×192, thickness 1.2 mm, no gap) were obtained using a fully flow compensated spoiled GRE (TR=2000 ms, TE=20.2 ms) and a TSE sequence (TR=9130 ms, TE=22 ms). To assess replicability of the results, sequences were repeated at each session and subjects were scanned twice. Acquisition time was 40 minutes for each session. Correlations of voxel intensities within the amygdala between all acquisitions were computed. Obtaining subfields of the amygdala An amygdala mask was manually created for each subject [1] to exclude extraneous tissue. Each contrast gives particular tissue information. Spectral clustering [2] was applied, based on multivariate feature vectors (GRE, TSE) (Fig.1). To assess consistency of the clustering, we compared results on each subject using Cramer's V, which has values in the interval [0,1]. This was applied within each session and across scanning sessions which took place one week apart. 6 out of 9 subjects showed values higher than V=0.50 across weeks when 3 clusters were chosen. The remaining subjects had higher consistency when 5 clusters were applied. Results Results and discussion. All subjects showed high correlation within sequences and within both sessions. Multiple amygdaloid subregions have been previously studied in animal studies and human cadaver brain [3]. How many nuclei should be considered as independent entities remains an open question. This study shows in vivo reproducible heterogeneity within the amygdala (Fig.2) which could correspond to several subdivisions. The use of 7T MRI may contribute significantly to the understanding of the anatomy and function of the amygdala