Connectivity-based parcellation of the amygdala and identification of its main white matter connections

Abstract

The amygdala plays a role in emotion, learning, and memory and has been implicated in behavioral disorders. Better understanding of the amygdala circuitry is crucial to develop new therapies for these disorders. We used data from 200 healthy-subjects from the human connectome project. Using probabilistic tractography, we created population statistical maps of amygdala connectivity to brain regions involved in limbic, associative, memory, and reward circuits. Based on the amygdala connectivity with these regions, we applied k-means clustering to parcellate the amygdala into three clusters. The resultant clusters were averaged across all subjects and the main white-matter pathways of the amygdala from each averaged cluster were generated. Amygdala parcellation into three clusters showed a medial-to-lateral pattern. The medial cluster corresponded with the centromedial and cortical nuclei, the basal cluster with the basal nuclei and the lateral cluster with the lateral nuclei. The connectivity analysis revealed different white-matter pathways consistent with the anatomy of the amygdala circuit. This in vivo connectivity-based parcellation of the amygdala delineates three clusters of the amygdala in a mediolateral pattern based on its connectivity with brain areas involved in cognition, memory, emotion, and reward. The human amygdala circuit presented in this work provides the first step for personalized amygdala circuit mapping for patients with behavioral disorders.

Figure 1

Cortical (D, E, F, H) and subcortical (B, C, G) masks of the target brain areas and the amygdala (A, center) from a single subject. (A): Amygdala. (B): Brainstem (BS). (C): Nucleus accumbens (NAc) (D): Dorsolateral prefrontal cortex (DLPFC). (E): Orbitofrontal cortex (OFC). (F): Rostral anterior cingulate cortex (rACC). (G): Hippocampus (Hippo). (H): Insula. (I) Probability of connectivity from the amygdala to each target region. The connection probability of each voxel of the amygdala to each of the seven target regions was averaged over all amygdala voxels and this value then averaged across all subjects. Note that the highest probability of connectivity is to the hippocampus (48%). Thus, 48% of all streamlines from the amygdala to the above targets terminated in the hippocampus.

Figure 2

Individual results were normalized to a common MNI space. The clusters were separated and summed across all 168 subjects such that each voxel value in the final cluster represented the number of subjects with the cluster in that location. These group probability maps of the clusters were then thresholded at 50% to optimize the overlap between each other.

Figure 3

Qualitative comparison of our connectivity-based parcellation with other amygdala atlases^,–. (A) Our results show three clusters in a mediolateral pattern that is consistent with the gross structural configuration of the amygdala described in previous studies (B through E). L, lateral. (B) Basal. M, medial. CM, centromedial.

Figure 4

Connectivity (measured by probability of connectivity 0–1, y axis) of the seven target regions to each amygdala cluster. DLPFC (dorsolateral prefrontal cortex), Hippo (hippocampus), NAc (nucleus accumbens), OFC (orbitofrontal cortex), rACC (rostral anterior cingulate cortex). *= p ≤ 0.05, **= p ≤ 0.01, ***= p ≤ 0.001.

Figure 5

Population connectivity maps of the amygdala clusters. (A) The lateral cluster also showed stronger connectivity with sensory areas including parietal, occipital, and posterior temporal areas (arrows) (B) The basal and lateral clusters also showed stronger connectivity through the parahippocampal radiation of the cingulate bundle (white arrow in sagittal view), with the temporal pole (black arrow in sagittal view), and through the amygdalotectal pathways with the colliculi. (C) The lateral cluster had stronger connectivity through the subcaudate white matter and ventromedial part of the uncinate fascicle with the PFC (arrow).

Figure 6

Population connectivity maps of the amygdala clusters. (A) The basal cluster had stronger connectivity with the medial thalamus through the inferior thalamic peduncle (arrow in sagittal), and through the anterior limb of the internal capsule with the PFC (arrow in axial view). (B) with the HC through the amygdalohippocampal bundle and (arrow). (C) with the ventral striatum (arrow in C).

Figure 7

Population connectivity maps of the amygdala clusters. (A) The medial and basal clusters showed stronger connectivity through the stria terminalis (arrow in sagittal view) and ventral amygdalofugal pathway (arrow in axial view). These two pathways converged in the BNST, hypothalamus, and septal area. The medial cluster showed unique connectivity with the ventral tegmental area through the medial forebrain bundle (black arrow in axial). (B) and (C) The medial and lateral cluster showed stronger connectivity with ventral and dorsal areas of the NAc respectively.

Figure 8

Functional model of the amygdala circuit. (A) The basolateral complex (L, B) receives and integrates information [from visual (V), sensory areas, hippocampus (H), and prefrontal cortex]. (B) The centromedial complex (CM) translates this information to behavior through the connections with the hypothalamus (Hth) and the brainstem (Bst).