Brain microstructure and connectivity in COVID-19 patients with olfactory or cognitive impairment

Abstract

Introduction: The COVID-19 pandemic has affected millions worldwide, causing mortality and multi-organ morbidity. Neurological complications have been recognized. This study aimed to assess brain structural, microstructural, and connectivity alterations in patients with COVID-19-related olfactory or cognitive impairment using post-acute (time from onset: 264[208-313] days) multi-directional diffusion-weighted MRI (DW-MRI).

Methods: The study included 16 COVID-19 patients with cognitive impairment (COVID-CM), 35 COVID-19 patients with olfactory disorder (COVID-OD), and 14 controls. A state-of-the-art processing pipeline was developed for DW-MRI pre-processing, mean diffusivity and fractional anisotropy computation, fiber density and cross-section analysis, and tractography of white-matter bundles. Brain parcellation required for probing network connectivity, region-specific microstructure and volume, and cortical thickness was based on T1-weighted scans and anatomical atlases.

Results: Compared to controls, COVID-CM patients showed overall gray matter atrophy (age and sex corrected p = 0.004), and both COVID-19 patient groups showed regional atrophy and cortical thinning. Both groups presented an increase in gray matter mean diffusivity (corrected p = 0.001), decrease in white matter fiber density and cross-section (corrected p < 0.05), , and COVID-CM patients also displayed an overall increased diffusivity (p = 0.022) and decreased anisotropy (corrected p = 0.038) in white matter. Graph-based analysis revealed reduced network modularity, with an extensive pattern of connectivity increase, in conjunction with a localized reduction in a few connections, mainly located in the left hemisphere. The left cingulate, anterior cingulate, and insula were primarily involved.

Conclusion: Expanding upon previous findings, this study further investigated significant alterations in brain morphology, microstructure, and connectivity in COVID-19 patients with olfactory or cognitive disfunction. These findings suggest underlying neurodegeneration, neuroinflammation, and concomitant compensatory mechanisms. Future longitudinal studies are required to monitor the alterations over time and assess their transient or permanent nature.

Keywords: Brain Connectivity; Brain Microstructure; COVID-19; Diffusion-weighted MRI; Neurodegeneration; Neuroinflammation.

Graphical abstract

Fig. 1

Schematic representation of the MRI processing pipeline. Both the T1-weighted and the 32-direction diffusion-weighted MRI (DW-MRI) scans first undergo pre-processing to correct for possible noise, artifacts, and distortions. The DW-MRI scan is then processed to derive the tensor image with the pertinent Mean Diffusivity (MD) and Fractional Anisotropy (FA) maps. Constrained spherical deconvolution is applied to estimate the Fiber-orientation Distribution Image (FOD), subsequently normalized. A probabilistic algorithm generates the streamlines' tractography from the FOD, and the reconstructed fibers are filtered using a model that best fits the diffusion signal. FOD images are also used as the basis for the fixel-based analysis of the Fiber Density and Cross-section (FDC). The pre-processed T1 scan, after registration to the DWI-b0 scan, undergoes gray matter (GM) and white matter (WM) segmentation and is then used in combination with GM and WM anatomical atlases to parcellate the brain into tissues and regions. DW-MRI and T1 processing outputs are finally combined to investigate connectivity and microstructure, with the latter examined by both ROI analysis and tractometry.

Fig. 2

Volume (left column), Fractional Anisotropy (FA, middle column), and Mean Diffusivity (MD, right column) distributions in the white and gray matter in COVID-19 patients with cognitive deficit (COVID-CM, n = 16; panel A, top rows, in orange) or olfactory disorder (COVID-OD, n = 35; panel B, bottom rows, in yellow) as compared with normal controls (CTRL, n = 14; green boxplots). P-values were computed using the Wilcoxon-Mann-Whitney test. T-test p-values of the group variable in the regression model accounting also for sex, age, and brain volume of the subjects are reported when significant (p < 0.05). The examples of the segmented tissue on the right are taken from a representative COVID-19 patient, specifically a 54-year-old female.

Fig. 3

Fiber Density and Cross-Section (FDC) alterations in COVID-19 patients with cognitive deficit (COVID-CM, panels A and B, top row) or olfactory disorder (COVID-OD, panels C and D, bottom row) as compared with normal controls. All four panels show fixels with significant reduced FDC in patients (p < 0.05). In the left column (panels A and C), the color represents altered fixels orientation, and in the right column, the color scale represents statistical significance. T-test p-values were computed using non-parametric permutation and connectivity-based fixel enhancement to correct for multiple comparisons. Sex and age were set as covariates in the general linear model.

Fig. 4

Network modularity and connectome in COVID-19 patients with cognitive deficit (COVID-CM, n = 16, top) or olfactory disorder (COVID-OD, n = 35, bottom) as compared with normal controls (CTRL, n = 14). The figure's left side (boxplots A and C) displays the network modularity distribution, computed via the community detection Louvain method. The network comprises 58 gray matter regions derived from the AAL3 atlas, excluding the cerebellum, vermis, supplementary motor areas, and paracentral lobules. Wilcoxon-Mann-Whitney p-value is reported along with the result from t-test the regression model that also accounts for the sex, age, and brain volume of the subjects. The connectomes on the right side of the figure (panels B and D) show the connections with significantly altered strength (p < 0.01) and the direction of the alteration. Blue indicates the connections weakened in the COVID-19 patients. Connection-specific P-values were computed using the Wilcoxon-Mann-Whitney test.

Fig. 5

Brain connection impairment in COVID-19 patients with cognitive (COVID-CM, n = 16, top) or olfactory disorder (COVID-OD, n = 35, bottom) as compared with normal controls (CTRL, n = 14). The left side of the figure shows the pattern and list of significantly impaired connections (A, C). The right side of the figure (boxplots B and D) shows the results of the connectivity analysis, and in particular, the distribution of the density, number of modules, modularity, and global efficiency of the gray matter subnetwork affected by the impaired connections. The significantly altered connections were identified and corrected for multiple comparisons using the Network-Based Statistics (NBS) tool t-tests and while taking into account the age and sex of the subjects. P-values for connectivity measures were computed using the Wilcoxon-Mann-Whitney test. T-test p-values of the group variable in the linear regression model accounting also for sex, age, and brain volume of the subjects are reported when significant (p < 0.05). Abbreviations: ACC: anterior cingulate cortex; N Acc: nucleus accumbens.