Parcellation of human amygdala in vivo using ultra high field structural MRI

Abstract

Histological studies show that human amygdala is subdivided into several nuclei with specific connections to other brain areas. One such study has been recently used as the basis of a probabilistic amygdala map, to enable in vivo identification of specifically located functions within the amygdala and connections to it. The involvement of the amygdala in cognition, emotion and action, which may underlie several psychiatric disorders, points to a need for discrimination of these nuclei in living human brains using different techniques. Structural MRI scans of the human amygdala at standard field strengths (≤ 3 T) have shown a region of generally featureless gray matter. Apparently homogeneous regions may reveal internal structure, however, when improved imaging strategies and better SNR are available. The goal of this study is the in vivo anatomical segmentation of the amygdala using high resolution structural MR data. The use of different MRI tissue contrast mechanisms at high field strengths has been little explored so far. Combining two different contrasts, and using cutting-edge image analysis, the following study provides a robust clustering of three amygdala components in vivo using 7 T structural imaging.

Introduction

One important objective of magnetic resonance imaging is to distinguish human brain areas in vivo. In an attempt to understand the functions of the amygdala complex, a structure located deep within the temporal lobe, the segmentation of its components has always been of wide interest (Alheid et al., 1995, Amunts et al., 2005, Aylward et al., 1999, Johnston, 1923, Price et al., 1987). However, standard radiological MRI techniques do not discriminate intra-amygdala structures, and the amygdala generally appears as a relatively uniform large gray matter region.

The amygdala is an anatomically complex region of about 1500 mm³ which suggests a substantial functional diversity also manifested by its role in multiple disorders (Amunts et al., 2005, Ball et al., 2009, Boccardi et al., 2002, Floresco and Ghods-Sharifi, 2007, Garcia-Marti et al., 2008, Hayman et al., 1998, Lanteaume et al., 2007, LeDoux et al., 1988, Sah et al., 2003, Schiller et al., 2009, Sharot et al., 2007, Walker and Davis, 2002, Wiest et al., 2006). Pioneering studies of amygdala segmentation were performed by Johnston (1923). He dissociated two amygdala compartments on the basis of cell migration behavior: the primitive group and the new group, each of which consisted of several amygdala subnuclei (Johnston, 1923).

More recent cytoarchitectonic studies support a segmentation comprising three main groups of amygdala nuclei: superficial, basolateral and centromedial, together with a recently added independent area known as the extended amygdala (Alheid et al., 1995). These studies have provided maps of the amygdala nuclei in a sample of cadaver brains (Amunts et al., 2005) from which a probabilistic atlas has been derived. This has been increasingly used as a reference frame for the localization of particular functions and connections in the amygdala complex (Ball et al., 2009, Onur et al., 2009). This allows a better understanding of the relation between amygdala tissue and its function.

A recent study has shown a two compartment segmentation by utilizing diffusion tensor imaging data at a 3 T scanner (Solano-Castiella et al., 2010). Structural MRI studies at 2 T and 3 T have attempted volumetric delineation of the amygdala as a whole (Bonilha et al., 2004, Morey et al., 2009, Pruessner et al., 2000), resulting in an improved in vivo identification of the entire area and its boundaries. Due to the low contrast variations within this structure, the amygdala nuclei have not been previously anatomically delineated in MR structural images. However, the neighboring hippocampus, which has more easily observable internal structure, has been successfully parcellated at 7 T (Thomas et al., 2008).

Currently, attention regarding in vivo segmentation has been drawn to the evident higher sensitivity available at higher field strengths such as 7 T. The signal-to-noise ratio (SNR) quantifies the mean overall intensity of the MR image, as compared with background noise, which is only thermal noise in ideal conditions. CNR or contrast to noise-ratio-compares with background noise the signal difference between tissue components of interest, typically gray matter and white matter. The SNR is at least proportional to the strength of the static magnetic field, (Collins and Smith, 2001, Edelstein et al., 1986, Triantafyllou et al., 2005) resulting in more than twice the SNR compared to 3 T. Hence, images show an excellent anatomical detail due to the high contrast-to-noise ratio (CNR) which increases dramatically with the magnetic field (Li et al., 2006, Thomas et al., 2008).

Besides utilizing a 7 T scanner, in the present study, T2 and T2*-weighting were combined. Previous MR contrast combination has already shown great success in structural identification of different brain regions (Aubert-Broche et al., 2009, Dijkhuizen and Nicolay, 2003, Helms et al., 2006, Lalys et al., 2010, Ordidge et al., 1994). Thus, this study attempts to differentiate regions within the amygdala on the basis of in vivo ultra high field MR structural imaging.