Multimodal MRI reveals consistent basal ganglia and limbic system alterations in COVID-19 survivors

Abstract

The long-term impact of COVID-19 on the brain is multifaceted, encompassing structural and functional disruptions. A cohesive theory of the underlying mechanisms of the Post-COVID Syndrome (PCS) remains unknown, primarily due to high variability in findings across independent studies. Here, we present a multimodal, cross-sectional MRI analysis of brain morphology (T1-MRI), tissue microstructure (diffusion-MRI), functional connectivity (functional-MRI), and cerebral blood flow (arterial spin labeling MRI) in COVID-recovered patients (CRPs, N=76) and healthy controls (HCs, N = 51). Although the global brain volumes did not differ between the two groups, CRPs showed focal atrophy in the right basal ganglia and limbic structures, along with cortical thinning in paralimbic regions (prefrontal cortex, insula) (p < 0.05). Diffusion MRI analysis revealed reduced fractional anisotropy and elevated radial diffusivity in the uncinate fasciculus and cingulum. No differences were observed in resting-state functional connectivity (RSFC) and cerebral blood flow between HCs and CRPs (p > 0.05). We further investigated the effect of infection severity by stratifying the CRPs into hospitalized (HP; N = 21) and non-hospitalized (NHP; N = 46) groups. The microstructural damage was linked to infection severity, more pronounced in the HPs (p < 0.05). In HPs, RSFC was diminished between components of the default mode network and the insula and caudate as compared with HCs and NHPs (p < 0.05). Results suggest COVID-19 is associated with selective structural and functional alterations in basal ganglia-limbic-cortical circuits, with stronger effects in severe cases. Overall, our findings both validate previously reported neuroimaging biomarkers and reveal new ones associated with the post-COVID syndrome, motivating future hypothesis-driven studies on behavioral correlates and therapeutic interventions.

Keywords: T1-weighted MRI; arterial spin labeling (ASL); diffusion-weighted MRI (dMRI); post-COVID syndrome (PCS); resting-state functional MRI (rs-fMRI).

Fig. 1.

Fourteen resting-state networks (RSN) were identified by comparing the independent components (ICs) with network maps from the Yeo Atlas (https://surfer.nmr.mgh.harvard.edu/fswiki/CorticalParcellation_Yeo2011). The identified ICs are presented here with the best-matched RSN from the atlas. The values of corresponding correlation coefficients are provided in parentheses. Key: N/w = Network.

Fig. 2.

(A) Proportions of symptoms during SARS-CoV-2 infection in COVID-19 recovered patients, total CRPs = 67. (B) Proportions of post-COVID symptoms in COVID-19 recovered patients. We also requested CRPs to share their post-COVID experience and any persistent symptoms since recovery. Among 59 (30.85 $\pm$ 10.52 years, 15F) consenting CRPs, fatigue (40/59) was the most commonly reported post-COVID symptom, followed by unrefreshing sleep (27/59) and lack of attention (26/59). Other symptoms reported by CRPs were achy muscles (26/59), achy joints (23/59), and headache (24/59). A summary of the post-COVID symptoms is presented in (B).

Fig. 3.

Distribution of volume of regions that showed significant differences between the COVID-19 recovered patients (CRPs) and healthy controls (HCs) ($p_{FDR}<0.05$ ).

Fig. 4.

Significant clusters obtained upon comparison of surface-based morphological metrics: (A) Cortical volume differences between healthy controls (HCs) and COVID-19 recovered patients (CRPs) where cluster was observed in the right superior temporal gyrus, insular cortex ($p_{FDR}<0.05$ , HC>CRP). Significant differences in (B) cortical volume and (C) cortical thickness were observed between HCs and hospitalized patients (HP) wherein the cluster was found left medial prefrontal cortex ($p_{FDR}<0.05$ , HC>HP).

Fig. 5.

Distribution of diffusion metrics of tracts that showed a significant difference between healthy controls (HCs) and COVID-19 recovered patients (CRPs) ($p_{FDR}<0.05$ ). Key: L. = left, R. = right, FA = fractional anisotropy, MD = mean diffusivity, AD = axial diffusivity, RD = radial diffusivity.

Fig. 6.

Distribution of volume of regions that showed a significant effect of infection severity among healthy controls (HCs), non-hospitalized patients (NHPs), and hospitalized patients (HPs) ($p_{FDR}<0.05$ ).

Fig. 7.

Distribution of diffusion metrics of tracts that showed a significant effect of infection severity among healthy controls (HCs), non-hospitalized patients (NHPs), and hospitalized patients (HPs) ($p_{FDR}<0.05$ ). Key: L. = left, R. = right, FA = fractional anisotropy, MD = mean diffusivity, AD = axial diffusivity, RD = radial diffusivity, FDR = false detection rate.

Fig. 8.

Significant clusters that highlight the effect of infection severity on the functional connectivity (FC) of the anterior default mode network (DMN-A) with the brain among healthy controls (HCs), Non-hospitalized patients (NHPs), and hospitalized patients (HPs) ($p_{corr}<0.05$ ). (A) Post hoc comparison of HCs and HPs showed a significant cluster in the left posterior insular cortex (HC>HP). (B) Post hoc comparison of HPs and NHPs also highlighted a cluster in the left posterior insular cortex (NHP>HP). (C) Post hoc comparison of NHPs and HCs showed a significant cluster in the dorsal visual association cortex (NHP>HC).

Fig. 9.

Significant clusters that highlight the effect of infection severity on the functional connectivity (FC) of the posterior default mode network (DMN-C) with the brain among healthy controls (HCs), non-hospitalized patients (NHPs), and hospitalized patients (HPs) ($p_{corr}<0.05$ ). (A) Post hoc comparison of NHPs and HPs highlighted clusters in the precuneus, posterior cingulate, and the inferior frontal cortex (NHP>HP), and (B) post hoc comparison of HCs and HPs showed a significant cluster in the right caudate and the left thalamus (HC>HP).

Fig. 10.

Significant clusters that highlight the effect of infection severity on the functional connectivity (FC) of the somatomotor network (SMN) with the brain among healthy controls (HCs), non-hospitalized patients (NHPs), and hospitalized patients (HPs) ($p_{corr}<0.05$ ). (A) Post hoc comparison of the FC of the SMN-A in the NHPs and HPs highlighted a significant cluster in the left postcentral gyrus (NHP>HP) and (B) post hoc comparison of the FC of the SMN-B in the NHPs and HPs highlighted a significant cluster in the left and right caudate, the right Heschl’s and supramarginal gyri (NHP>HP).