Response properties of human amygdala subregions: evidence based on functional MRI combined with probabilistic anatomical maps

Abstract

The human amygdala is thought to play a pivotal role in the processing of emotionally significant sensory information. The major subdivisions of the human amygdala-the laterobasal group (LB), the superficial group (SF), and the centromedial group (CM)-have been anatomically delineated, but the functional response properties of these amygdala subregions in humans are still unclear. We combined functional MRI with cyto-architectonically defined probabilistic maps to analyze the response characteristics of amygdala subregions in subjects presented with auditory stimuli. We found positive auditory stimulation-related signal changes predominantly in probabilistically defined LB, and negative responses predominantly in SF and CM. In the left amygdala, mean response magnitude in the core area of LB with 90-100% assignment probability was significantly larger than in the core areas of SF and CM. These differences were observed for pleasant and unpleasant stimuli. Our findings reveal that the probabilistically defined anatomical subregions of the human amygdala show distinctive fMRI response patterns. The stronger auditory responses in LB as compared with SF and CM may reflect a predominance of auditory inputs to human LB, similar to many animal species in which the majority of sensory, including auditory, afferents project to this subdivision of the amygdala. Our study indicates that the intrinsic functional differentiation of the human amygdala may be probed using fMRI combined with probabilistic anatomical maps.

Figure 1. EPI data quality.

A representative example of raw EPI (echoplanar imaging) data normalized to MNI space is shown in a coronar (MNI y = −7 mm) and axial (MNI z = −20 mm) section. The axial section corresponds approx. to the sections shown in . The example represents the average across all EPI volumes acquired during the experimental session of one subject. The outline of the amygdala is shown (red line), enclosing the area with at least 50% probability of belonging to the amygdala (according to the probabilistic maps from [26]). The extent of the laterobasal (LB) group of the amygdala (LB, > = 80% probability) is shown in blue, the superficial group (SF) in green, and the centromedial group (CM) in magenta. In addition, the outline of the segmentation mask enclosing the area with sufficient signal for application of point spread function (PSF) based EPI distortion correction (see reference for further details) is shown in yellow. In all subjects, the whole analyzed extent of the amygdala was within the segmentation mask and therefore distortion correction was possible in this region. Good EPI signal quality in the amygdala region was achieved in all subjects investigated.

Figure 2. Simulated effect of probabilistic region of interest (ROI) definition on robustness against localization errors.

In (a) and (b), axial sections through the laterobasal part (LB) of the left amygdala are shown at z = −29 (MNI coordinate system). Probabilities to belong to LB are color coded with black = 10% and pure red = 100%. In (a), the probabilistic ROI is defined as the area with> = 50% probability to belong to LB (solid yellow line). As a consequence of a simulated localization error which consists of a linear shift of approximately 4 mm (indicated by the white arrow) the true measured volume is localized as indicated by the dashed yellow line, containing voxels with little or no probability to belong to LB. In contrast, when restricting the ROI to voxels with 90% or 100% LB probability (b) and given an equally large localization error, the true measured volume is still entirely located within the area of LB. In this way, the robustness against localization errors could be increased, counteracting potential localization errors while still maintaining excellent anatomical specificity.

Figure 3. Music-related amygdala responses.

The extent of the amygdala subregions LB, SF, and CM defined by maximum-probability maps (MPMs, [28]) are rendered in blue, green, and magenta, respectively, on a coronar section. Regions with significant (p<0.05, corrected for multiple comparisons in the search volume) music-related responses are shown in red. The BOLD percentage signal change (PSC) of each of the three activation sites in the four music conditions (1: dissonant slow, 2: consonant slow, 3: dissonant fast, 4: consonant fast) is presented as mean and standard error. The fifth PSC value refers to the time period during which subjects evaluated the emotional effect evoked by the preceding musical piece, compared with rest. In the left amygdala, a cluster with positive responses to all music conditions was mainly located in LB and a cluster with negative responses was mainly found in SF and CM (see also Tab. 2). A smaller cluster with consistently positive responses was also found in the right amygdala approximately in the mirror position to the positively responding left cluster.

Figure 4. Human amygdala responses to music in probabilistically defined anatomical regions.

The volumes with at least 50% and 80% probability to belong to the three major amygdala subregions: LB = laterobasal group (blue), SF = superficial group (green), and CM = centromedial group (magenta) are rendered with high (for 50% probability) and low (for 80% probability) transparency. Voxels with a significant increase in BOLD signal to music presentation are displayed as red or yellow squares (p<0.05, corrected for multiple comparisons, see material and methods for further details). Voxels with a significant signal decrease are displayed as black or yellow triangles. Anatomical probabilities for red and black responses were 50% to 60%, for yellow responses 70% or above. The majority of significant effects was found in the left amygdala. Positive effects were found in both right and left LB and SF (c.f. Tab. 2), negative effects were localized in left LB, SF and CM. For visualization, probability maps were smoothed using a spatial filter with a 5 mm isometric Gaussian convolution kernel.

Figure 5. Lateralization patterns in the major subdivisions of the human amygdala.

Histograms on the left show the distribution of BOLD percentage signal change (PSC) for voxels in: LB = laterobasal group, SF = superficial group, and CM = centromedial group. Values for the right (black) and left (light grey) amygdala are shown in a stacked way. For each PSC bin, a lateralization index (LI) was calculated. LIs of 100, −100, and 0 indicate purely right-sided, purely left-sided, and symmetrically distributed effects, respectively. LB and SF showed a similar lateralization pattern of both extremely positive and negative PSC predominating in the left amygdala. CM showed a different lateralization pattern with negative PSC predominating in the left and positive PSC in the right amygdala.

Figure 6. Response differences between amygdala subregions.

We performed a region of interest (ROI) analysis to statistically evaluate response differences between the core regions (with 90–100% assignment probability) of the three amygdala subregions (see Material and Methods for further details). Bars in the above figure show the result of three averaging steps: first across all voxels of a given area, then either across all four music conditions, the dissonant conditions, or the consonant conditions, and finally across the 14 subjects. Results are given for left LB, SF, and CM in blue, green, and magenta. Error bars indicate the standard error of the mean from of the average across subjects. In each case, the differences between left LB and left SF and CM were significant at p<0.005. There were no corresponding significant differences in the right amygdala.