fMRI measurements of amygdala activation are confounded by stimulus correlated signal fluctuation in nearby veins draining distant brain regions

Abstract

Imaging the amygdala with functional MRI is confounded by multiple averse factors, notably signal dropouts due to magnetic inhomogeneity and low signal-to-noise ratio, making it difficult to obtain consistent activation patterns in this region. However, even when consistent signal changes are identified, they are likely to be due to nearby vessels, most notably the basal vein of rosenthal (BVR). Using an accelerated fMRI sequence with a high temporal resolution (TR = 333 ms) combined with susceptibility-weighted imaging, we show how signal changes in the amygdala region can be related to a venous origin. This finding is confirmed here in both a conventional fMRI dataset (TR = 2000 ms) as well as in information of meta-analyses, implying that "amygdala activations" reported in typical fMRI studies are likely confounded by signals originating in the BVR rather than in the amygdala itself, thus raising concerns about many conclusions on the functioning of the amygdala that rely on fMRI evidence alone.

Figure 1

Average functional activation from 16 subjects measured with a low-TR (333 ms) multiband EPI sequence for the contrast between fearful faces (top) and threatening IAPS pictures (bottom) compared to geometric forms in a block-design matching task. Note the activation pattern in the amygdala area following the typical course of the basal vein of Rosenthal (BVR) around the brainstem until no longer distinguishable from the activation cluster in the occipital lobe. The values depicted are beta coefficients for the linear models averaged across subjects and can be interpreted as percent signal change between the faces/IAPS blocks on one hand and the geometric figures on the other. Arrows indicate locations of activation foci in the meta-analysis by Sabatinelli et al., the arrows in the axial slice at z = −16 and in the coronal slices at y = 4 and y = 6 pointing to the amygdala foci, the arrows in the axial slices at z = −8 and z = −4 pointing to the parahippocampal gyrus foci, see table 1. Note the close proximity in particular in the amygdala region between the meta-analytic activation foci and the BVR activations.

Figure 2

Example single-subject susceptibility-weighted images and results from the GLM of the matching paradigm (using the contrast ‘Faces – Forms’). Subject 2 exhibits a clear posterior drainage of the BVR (green and white arrows) around the brain stem in the left hemisphere, subject 8 in the right hemisphere, and subject 18 in both. Despite some geometric distortions between the two images, leading to the vein not always appearing on the same slice in the two, a correspondence between the activations and the veins in the SWI images (black) is visible. Approximate amygdala positions for each subject encircled in red.

Figure 3

Mean SWI. Though most smaller brain vessels vanish when averaging SWI images across subjects, the typical course of the BVR is clearly visible as a dark line from the uncus close to the amygdala, around the brain stem to the vein of Galen. The course of the BVR is highlighted with green arrows. Approximate amygdala position encircled in red.

Figure 4

Resting-state functional connectivity from 16 subjects measured with a low-TR (333 ms) multiband EPI sequence using each subjects 100 most strongly activated voxels in the amygdala region as a seed at single-subject level, r-to-z transformed and averaged across subjects. Signals in these voxels are most strongly correlated with voxels containing large vessels, as around the brain stem and in the lateral fissure.

Figure 5

InstaCorr functional connectivity maps for 12 voxels at the amygdala/BVR border for one exemplary subject from the low-TR (333 ms) dataset, with colors representing correlation coefficients. The location of the 12 seed voxels are shown in the bottom left corner in a zoomed-in view of the raw EPI slice. Seeds in the BVR, e.g. at (14.6, 7.7, 15.5), have a characteristic connectivity pattern involving strong correlations to other voxels in the vasculature, whereas seeds more clearly within the amygdala itself, e.g. (18, 7.7, 15.5), display no correlation with these voxels. The top right seed voxel at coordinates (19.7, 6, 15.5) corresponds to the left amygdala activation focus in Sabatinelli et al. and shows an intermediary pattern in this particular subject, with apparently some contamination by venous signal.

Figure 6

Average functional activation from a dataset of 134 subjects measured with a high-TR (2000 ms) EPI sequence for the contrast between fearful faces (top) and threatening IAPS pictures (bottom) compared to geometric forms in a block-design matching task. As in figure 1, the activation pattern in the amygdala region follows the typical course of the BVR around the brainstem until no longer distinguishable from the activation cluster in the occipital lobe. The values depicted are beta coefficients for the linear models, averaged across subjects, and can be interpreted as percent signal change between the faces/IAPS blocks on one hand and the geometric figures on the other. Arrows depict activation foci from the meta-analysis by Sabatinelli et al., the arrows in the axial slice at z = −16 and in the coronal slices at y = 4 and y = 6 pointing to the amygdala foci, the arrows in the axial slices at z = −8 and z = −4 pointing to the parahippocampal gyrus foci, see table 1. Note the proximity of these foci to the BVR activations for both the amygdala and the parahippocampal gyrus activation foci from the meta-analysis.

Figure 7

Average functional activation from a subset of 30 subjects of the high-TR (2000 ms) dataset for the contrast between fearful faces (top) and threatening IAPS pictures (bottom) compared to geometric forms in a block-design matching task, similar to figure 6. Activations closely resemble those of that figure despite the smaller sample size. Note that activations in BVR voxels between the amygdala and regions affected by signal loss in proximity of the nasal cavities (characterized by noisy activation patterns) exist as in the complete dataset, but are more difficult to distinguish from the noise near the nasal cavities in the smaller subset.