Neuroimaging biomarkers of post-acute sequelae of Coronavirus Disease 2019

Abstract

COVID-19, caused by SARS-CoV-2, has led to the condition known as Long COVID or post-acute sequelae of COVID-19 (PASC), where individuals experience persistent debilitating symptoms long after the initial infection. We provide here a comprehensive review of findings in the central nervous system associated with PASC. Neuroimaging has been instrumental in identifying brain changes associated with PASC. Structural MRI studies consistently reveal grey matter volume reductions in the frontal and temporal lobes and white matter hyperintensities, particularly in the periventricular regions. Studies especially found these changes to correlate strongly with cognitive deficits. Diffusion tensor imaging has shown increased tissue damage and oedema in the brain’s white matter tracts, particularly in the sagittal stratum and thalamic radiation. Resting-state functional MRI studies indicate altered brain connectivity in PASC patients, especially in those with post-traumatic stress symptoms. Reduced connectivity within and between critical networks, such as the default mode network and the executive control network, has been observed. These changes correlate with cognitive impairments, such as attention and memory deficits. Dynamic functional connectivity analyses further reveal that PASC patients spend less time in states with rich inter-regional connectivity, and transitions between connectivity states were linked to post-traumatic stress disorder symptoms. Positron emission tomography scans have shown hypometabolism in the frontal and temporal lobes, particularly in regions associated with memory and executive functions. Hypometabolism in the hippocampus and thalamus is linked to symptoms like anosmia and fatigue. Despite the heterogeneity in clinical presentations and diagnostic criteria, these neuroimaging findings underscore the significant impact of COVID-19 on brain structure and function. Continued research using advanced imaging techniques is essential for a deeper understanding of PASC’s neurological effects. This will aid in developing targeted interventions and improving outcomes for those affected by Long COVID and inform studies investigating downstream effects of viral infections on the brain.

Introduction

COVID-19, caused by the novel coronavirus, has had a profound global impact since its emergence in late 2019.1 This highly contagious respiratory infection has resulted in widespread illness, overwhelmed healthcare systems, and led to a significant loss of life and morbidity. Long COVID, also known as post-acute sequela of COVID (PASC), refers to the condition where individuals experience persistent and often debilitating symptoms for weeks to months following COVID-19.^2,3 While most people recover from the virus within a few weeks, those with PASC continue to experience a wide range of symptoms such as fatigue, breathlessness, joint pain, brain fog, and anxiety, among others in the weeks to months following acute infection.^4,5 The exact cause of PASC is still not fully understood, and it can affect individuals regardless of the severity of their initial infection. This condition has significant implications for public health, highlighting the need for further research, support, and clinical interventions to help individuals manage and overcome the long-term effects of COVID-19. Among other diagnostic tools, imaging may play an important part in defining and prognosticating those with long-term clinical findings.

Pathophysiological mechanisms of COVID-19

COVID-19, which is caused by the severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), primarily affects the respiratory system.^6,7 The virus most avidly binds to tissue in the lung, allowing it to replicate, propagate, and destroy infected cells, releasing new virus particles.⁸ This triggers an immune response, leading to the release of proinflammatory cytokines and chemokines.^9–11 The surge of cytokines (cytokine storm) released causes endothelial damage and compromises microvascular permeability, resulting in systemic coagulopathy and multi-organ failure. COVID-19 can also affect other organ systems, such as the heart (leading to myocarditis and arrhythmias), and the neurological system (leading to encephalitis and other neurological complications).

Pathophysiological mechanisms of PASC

The pathophysiological mechanisms of PASC remain poorly understood. Several potential factors likely contribute to the persistence of symptoms in individuals with PASC: (1) It is hypothesized that in some cases, the SARS-CoV-2 virus or viral particles may persist in specific tissues or organs even after the acute phase of the infection has resolved. This persistent viral presence could contribute to ongoing inflammation and tissue damage, leading to prolonged symptoms¹²; (2) PASC may arise from an autoimmune response resultant from autoantibody production. This autoimmune response can lead to persistent inflammation and damage to multiple organs¹³; (3) SARS-CoV-2 also affects the endothelial cells that line blood vessels, leading to endothelial dysfunction that can result in impaired blood flow, blood clot formation, and organ damage, which may contribute to the symptoms experienced in PASC.^14,15 Identifying biomarkers of PASC may improve our understanding of the underlying mechanism and potential therapeutic targets.¹⁶ Here, we conducted a systematic literature review to identify neuroimaging biomarkers of PASC.

Methods

Identifying relevant publications

Studies that used imaging (MRI, CT, PET, or SPECT) to evaluate PASC were included in this review. To include as many relevant publications as possible and a wide range of clinical, neurological, radiological, medical, and scientific databases, we devised a PubMed search strategy using the words COVID OR SARS-COV OR PASC OR Post-acute sequelae of COVID OR Long COVID OR Brain Imaging OR Imaging OR Cardiac OR Cardiovascular OR Pulmonary OR Magnetic Resonance Imaging OR Computed Tomography OR Positron Emission Tomography OR Single Photon Emission Computed Tomography OR Radiology OR Biomarker OR MRI OR CT OR X-Ray OR PET OR SPECT. We then added a Long COVID filter to our PubMed Search. Exclusion criteria were: (1) case reports with < 5 cases, (2) review articles, (3) undocumented criteria for PASC. Four authors (S.R.L., P.P., C.Z., M.M.B.) independently reviewed and screened the studies and discrepancies were resolved by consensus.

This search yielded 796 publications between March 16, 2020, and May 23, 2023. Of the 796 publications, articles that were reviews, based on 1-3 subjects or case reports were excluded. Thirty-four peer-reviewed manuscripts that leverage imaging (MRI, CT, PET, SPECT) to understand whether there were sustained cognitive deficits after a COVID-19 were identified. A review of findings from these is described below. Figure 1 summarizes all the findings below. Table 1 lists all the reviewed studies outlining the title, authors, inclusion/exclusion criteria, age, sex distribution, imaging modalities, and major findings.

Figure 1.

(A) Studies demonstrate grey matter (GM) volume reduction most consistently in the inferior and temporal regions. Some studies showed a laterality predilection for left inferior frontal and right inferior temporal atrophy. Volume reduction has also been reported involving parahippocampal gyri and cerebellum, as well as prefrontal and lateral temporal regions (not pictured). Some FDG-PET studies show multifocal hypometabolism, primarily in the frontal and temporal lobes. Hypometabolism in parietal lobe, hippocampus, and thalamus has also been reported (not pictured). (B) In addition to cortical GM volume reduction in structures such as insula, lateral and inferior temporal lobe gyri, and parahippocampal gyri, studies also demonstrate volume loss in deep GM structures, including caudate, putamen, amygdala, and hippocampus. Deep white matter changes are also observed. (C) Diffusion studies demonstrate mean and radial diffusivity changes within white matter tracts, including increased mean diffusivity in anterior thalamic radiations, genu of corpus callosum, forceps minor, and sagittal stratum. Increased radial diffusivity has been reported in uncinate fasciculus, anterior corona radiata, and genu of corpus callosum. Of note, other studies have reported decreased diffusivity in these and other regions/tracts, underlining disease heterogeneity and possible changes in findings across time and evolution of the disease. Reductions in fractional anisotropy were not observed.

Table 1.

List of all peer-reviewed articles summarized in this work.

Author	Title	N	Inclusion/Exclusion criteria	Age range	Sex	Imaging	Findings
Abrams RMC et al	MRI-negative myelopathy post mild SARS-CoV-2 infection: vasculopathy or inflammatory myelitis?	5	Lower limb symptoms possibly due to COVID-related myelopathy	58-69	2F	MRI (brain and cervical spine)	Normal brain MRI, spinal atrophy/compression was not remarkable
Ajčević M et al	Cerebral hypoperfusion in post-COVID-19 cognitively impaired subjects revealed by arterial spin labelling MRI	24	Subjects admitted to the post-COVID clinic, inclusion: persistent or ex-novo cognitive impairment at least after 4 weeks from polymerase chain reaction (PCR) test-confirmed SARS-CoV-2 infection. Exclusion criteria were: age < 18 or > 65 years, previous neurological or psychiatric diseases, major vascular alterations, previous history of cognitive deficits, consumption of agents affecting the nervous system, patients positive to SARS-CoV-2 with clinical and radiographic evidence of lower respiratory tract disease or hospitalized for COVID-19	38-68	15F	T1, ASL	Significant hypoperfusion in the frontal cortex, as well as the parietal and temporal cortex
Benedetti F et al	Brain correlates of depression, post-traumatic distress, and inflammatory biomarkers in COVID-19 survivors: A multimodal magnetic resonance imaging study	42	42 COVID-19 survivors without brain lesions, at 91 ± 55 days after COVID	47-72	13F	T1, DTI, rsfMRI	Severity of depressive psychopathology is associated with decreasing GM volumes in the ACC, while post-traumatic symptoms are associated with decreasing GM volumes both, in the ACC and in bilateral insular cortex, and both associate with WM microstructure. Inflammation during acute COVID-19 predicted both, WM microstructure and current psychopathology several weeks after clearance of the virus. Moreover, resting-state FC associated both, with inflammation and psychopathology, confirming the brain functional implication of the structural effect
Campabadal A et al	Structural brain changes in post-acute COVID-19 patients with persistent olfactory dysfunction	48	Inclusion criteria: COVID‐19 diagnosis, based on biological diagnosis or medical report; ≥8 weeks after the onset of COVID‐19 symptoms; fluent in Spanish and having at least 6 years of formal education; exclusion criteria: previous diagnosis of any neurological, psychiatric, or medical condition that could affect the baseline cognitive performance; any severe medical condition or visual or auditory problems that prevent the completion of the cognitive assessment; imaging findings on MRI other than mild WM hyperintensities in the FLAIR sequence, and MRI artifacts	35-65	10F	T1, DTI	Olfactory dysfunction group had lower GM volume in the amygdala, insular cortex, parahippocampal gyrus, frontal superior and inferior orbital gyri, gyrus rectus, olfactory cortex, caudate, and putamen. This group also showed higher MD values in the genu of the corpus callosum, the orbitofrontal area, the anterior thalamic radiation, and the forceps minor; and higher RD values in the anterior corona radiata, the genu of the corpus callosum, and uncinate fasciculus compared with the normal olfaction group
Cecchetti G et al	Cognitive, EEG, and MRI features of COVID-19 survivors: a 10-month study	49				T1, FLAIR	No significant differences were found between COVID-19 patients and healthy subjects in terms of total brain, GM, and WM volumes, especially after 10 months. COVID-19 patients showed interrelated cognitive, EEG, and MRI abnormalities 2 months after hospital discharge, but the differences were minimized at 10 months. Dysgeusia and hyposmia during acute COVID-19 were related with increased vulnerability in memory functions over time
de Paula JJ et al	Selective visuoconstructional impairment following mild COVID-19 with inflammatory and neuroimaging correlation findings	135	Inclusion: COVID-19 positive on a Rt-PCR study. Exclusion: prior mental disorder	28-48	95F	T1, FDG PET	No structural changes were observed on the T1 MRI. However, Rey-Osterrieth Complex Figure Test performance was negatively associated with WM volume such as callosum, cingulum, and frontal regions as well as FDG metabolism
Díez-Cirarda M et al	Multimodal neuroimaging in post-COVID syndrome and correlation with cognition	122 (36 controls)	Inclusion: diagnosis of COVID-19 with RT-PCR with 3 months of subjective cognitive complaints, exclusion: prior cognitive complaints, alcohol abuse, sensory disorders, known cerebrovascular disease	49-91	81F	T1, FLAIR, rsfMRI, DWI	Reduced functional connectivity and grey matter volume between left and right parahippocampal gyrus, frontal orbital cortex, cerebellum compared to controls. No differences between WMH burden were detected. Reduced MD and axial diffusivity were also observed in corpus callosum, longitudinal fasciculus, uncinate tract, fronto-occipital fasciculus, reduced whole-brain mean FA, MD, RD, and axial diffusivity
Du Y et al	Two-year follow-up of brain structural changes in patients who recovered from COVID-19: a prospective study	49, 27 controls	Inclusion: COVID-19 patients still had neuropsychiatric symptoms, cognitive impairment, 27 HCs matched by age, sex, and education were recruited	44-64	20F	T1	The grey matter volume in the cerebellum and vermis were reduced 1- and 2-year post COVID, positively correlated with lymphocyte count, and negatively correlated with neutrophil count, neutrophil/lymphocyte ratio after 1 year, and systemic immune-inflammation index after 2 years. The decreased GM volume in the left middle frontal gyrus, inferior frontal gyrus of the operculum, right middle temporal gyrus, and inferior temporal gyrus at year 1 returned to normal by year 2
Du Y et al	Survivors of COVID-19 exhibit altered amplitudes of low frequency fluctuation in the brain: a resting-state functional magnetic resonance imaging study at 1-year follow-up	51 (29 controls)	Inclusion: Individuals hospitalized and recovered from COVID-19, healthy volunteers who had never tested positive or been hospitalized with COVID-19, and who matched the COVID-19 survivors in age, sex, and education level, normal sight, hearing, right handedness, no history of brain abnormalities or mental/psychiatric illnesses, exclusion: patients with implants contraindicated by MRI, hyperthermia, claustrophobia	46-62	11F	rsfMRI	COVID-19 survivors showed significantly increased amplitude of low-frequency fluctuation (ALFF) values in several brain regions, including the left precentral gyrus, middle frontal gyrus, inferior frontal gyrus (opercular and triangular parts), insula, hippocampus, parahippocampal gyrus, fusiform gyrus, postcentral gyrus, inferior parietal angular gyrus, supramarginal gyrus, thalamus, middle and inferior temporal gyri, caudate, and putamen. Furthermore, ALFF values in the left caudate were positively correlated with Athens Insomnia Scale scores, while those in the left precentral gyrus were positively correlated with neutrophil counts during hospitalization
Ferrucci R et al	Brain positron emission tomography (PET) and cognitive abnormalities 1 year after COVID-19	7	Inclusion: participants hospitalized for COVID with neuropsychological deficits for 12 months after hospital discharge, exclusion: no existing cognitive deficits prior to COVID-19, no diabetes, or metabolic disorders	44-68	3F	FDG and 18Famy PET	Patients showed hypometabolism in frontal and temporal regions as well as significant amyloid deposition
Fu Z et al	Dynamic functional network connectivity associated with post-traumatic stress symptoms in COVID-19 survivors	93, (43 controls)	Inclusion: 50 hospitalized COVID-19 survivors (ie, discharged between February and March 2020 from hospitals in Wuhan, China, about 6 months after discharging at the time of the testing) and 43 age- and sex-matched non-COVID-19 controls	41-63	63F	rsfMRI	Patients scanned 4 months after hospitalization showed dynamic heterogeneous patterns between sensorimotor and visual networks. The occurrence of these patterns was associated with PTSD symptoms
Goehringer F et al	Brain (18)F-FDG PET imaging in outpatients with post-COVID-19 conditions: findings and associations with clinical characteristics	56 (28 controls)	Inclusion: asthenia, anosmia/ageusia, neurological symptoms, and/or dysautonomia following COVID infection	36-54	42F	18F-FDG PET	Outpatients with post-COVID-19 conditions exhibited diffuse hypometabolism predominantly involving right frontal and temporal lobes including the orbito-frontal cortex and internal temporal areas
Griffanti L et al	Adapting the UK Biobank brain imaging protocol and analysis pipeline for the C-MORE multi-organ study of COVID-19 survivors	67 (25 controls)	Inclusion: patients hospitalized for COVID-19 and discharged 2-3 months later, controls were age-sex, BMI, and risk factor (smoking, diabetes, hypertension) matched	40-64	32F	T1, DWI, SWI, ASL, FLAIR	Patients had reduced GM volume, thickness, density in the left hippocampus, lateral occipital cortex, middle temporal regions, parietal region, cingulate, higher MD in the thalamus, T2* differences in thalamus and hippocampus. No perfusion differences were observed
Jegatheeswaran V et al	Neuroimaging findings of hospitalized COVID-19 patients: a Canadian retrospective observational study	102	Inclusion: non-ICU and ICU patients admitted in the hospital with COVID-19	59-87	47F	CT, MRI	177 patients got CT, 21 got MRI, most commonly, ischaemia of large and small vessels was detected, followed by micro and macrohaemorrhages, siderosis, white matter abnormalities
Kiatkittikul P et al	Abnormality pattern of F-18 FDG PET whole body with functional MRI brain in post-acute COVID-19	13	Inclusion: patients with a history of COVID-19 with persistent abnormal symptoms (fatigue, shortness of breath) more than 28 days after infection, exclusion: <20 years, disabilities, history of cancer, pregnant	35-54	7F	18F-FDG PET, rsfMRI	Most patients showed hypometabolism in all lobes and was consistent with observed lower functional connectivity
Morand A et al	Similar patterns of [(18)F]-FDG brain PET hypometabolism in pediatric and adult patients with Long COVID: a pediatric case series	107, 7 cases, 21 ped control, 35 adults with Long COVID, 44 control adults	Inclusion: persistent symptoms for more than 4 weeks following the initial acute COVID-19 symptoms, without symptom-free interval	10-13	6F	18F-FDG PET	The 7 pediatric patients had a similar brain hypometabolic pattern as that found in adult Long COVID patients, involving bilateral medial temporal lobes, brainstem, and cerebellum
Muccioli L et al	Cognitive and functional connectivity impairment in post-COVID-19 olfactory dysfunction	49 (26 controls)	Inclusion: participants confirmed to have COVID-19 with antigen testing and persistent olfactory dysfunction. Exclusion: Pre-existing olfactory issues, chronic rhinosinusitis, cognitive impairment	23-51	24F	fMRI, T1	No T1 changes, less than 20% patients had hyposmia and cognitive impairment. fMRI shows altered olfactory network (insula, accumbens, thalamus, cingulate) functional connectivity and hyperconnectivity in the thalamus
Paolini et al	Brain correlates of subjective cognitive complaints in COVID-19 survivors: A multimodal magnetic resonance imaging study	58	Inclusion: diagnosis of COVID-19 infection via PCR testing during hospital admission, new cognitive symptoms since infection that persisted 1-3 months later. Exclusion: pre-existing cognitive conditions, prior alcohol abuse, major neurological disorders, pregnancy	18-70	17F	fMRI, DTI	Mean diffusivity (MD) and radial diffusivity (RD) were greater in inferior fronto-occipital fasciculus, uncinate fasciculus, corona radiata as well as several sections of corpus callosum in individuals with cognitive complaints compared to those without. Increased FC in subjects with cognitive deficits in the bilateral insular cortex, bilateral supramarginal and opercular cortex, anterior cingulate gyrus, bilateral precentral gyrus, and inferior lateral occipital cortex; reduced FC in bilateral superior occipital cortex, posterior cingulate gyrus, left middle temporal gyrus and right cerebellum
Tassignon B et al	Longitudinal changes in global structural brain connectivity and cognitive performance in former hospitalized COVID-19 survivors: an exploratory study	19, 6 non-ICU patients as controls)	Inclusion: COVID-19 patients hospitalized and followed up for 4 months	38-69	4F	T1, rsfMRI	No differences in T1 or fMRI-based graph theoretic connectivity measures were detected between patients admitted and not admitted to the ICU due to COVID
Tu Y et al	Post-traumatic stress symptoms in COVID-19 survivors: a self-report and brain imaging follow-up study	252, 126 controls	Inclusion: patients discharged from hospital for about 3-6 months	38-65	176F	T1, rsfMRI	COVID-19 survivors had more PTSD symptoms, greater functional activity in bilateral hippocampi and amygdala compared to controls. Grey matter volumes were higher but negatively correlated with severity of PTSD symptoms
Voruz P et al	Functional connectivity underlying cognitive and psychiatric symptoms in post-COVID-19 syndrome: is anosognosia a key determinant?	102, 45 controls	Inclusion: anosognosic patients with PASC, ie, memory impairment	43-70	37F	T1, rsfMRI	No structural differences were observed. Anosognosic patients showed hypoconnectivity between the left lateral prefrontal cortex and the default mode network, right somatomotor network and cerebellum (right Lobule V), and right precuneus and cerebellum (right Lobule V)

Abbreviations: ACC = anterior cingulate cortex; ALFF = amplitude of low-frequency fluctuation; ASL = arterial spin labelling; COVID = coronavirus disease of 2019; DWI = diffusion-weighted imaging; DTI = diffusion tensor imaging; FDG = fluorodeoxyglucose; FLAIR = fluid-attenuated inversion recovery; GM = grey matter; MD = mean diffusivity; OCT = optical coherence tomography; PASC = post-acute sequelae of COVID-19; PCR = polymerase chain reaction; PTSD = post-traumatic stress disorder; rsfMRI = resting-state functional MRI; RD = radial diffusivity; SWI = susceptibility-weighted imaging; tDCS = transcranial direct current stimulation.

Results

Neuroimaging findings in PASC

Structural MRI shows consistent grey matter (GM) volume reduction in the inferior frontal and temporal regions and deep white matter changes: Primary structural MRI measures included cortical and subcortical structure volumes using T1 MRI, white matter hyperintensity (WMH) volumes using T2-weighted fluid-attenuated inversion recovery (FLAIR) images, and diffusion tensor imaging or DTI.^17-19 The UK Biobank imaging study was launched to improve disease diagnosis using MRI with over 40 000 patients in their database. Once the COVID-19 pandemic struck, a parallel study, C-MORE (Capturing Multi-Organ Effects of COVID), was launched to study the effect of how SARS-CoV-2 infection manifested in the UK Biobank data.²⁰ C-MORE included a shorter version of the MRI protocol but with imaging parameters similar to the UK Biobank.^21–24 A total of 58 participants were studied. With T1-weighted MRI, the study found lower cortical GM density in the frontal and temporal lobe regions. Atrophy was lateralized to the left frontal and right temporal regions. Although lateralization of abnormalities is unexpected, a prior study of individuals within the UK Biobank with and without COVID-19 showed similar outcomes. Participants with COVID-19 were scanned almost 3 months after infection and showed greater cognitive decline compared to their pre-COVID assessment than those who had not yet been infected by COVID-19. FLAIR images showed that the PASC individuals had a more significant periventricular WMH burden. Further, the hyperintensity burden was associated with a higher concentration of inflammatory markers in the blood. In the same study, DTI showed higher mean diffusivity, indicating loss of tissue integrity and increased oedema in the sagittal stratum and right thalamic radiation. Fractional anisotropy was not different between participants with and without COVID-19.

Another study by Campabadal et al of about 25 participants with PASC dominated by olfactory symptoms, more than 8 weeks after infection, found lower GM volume in the left amygdala, insular cortex, parahippocampal gyrus, frontal superior and inferior orbital gyri, gyrus rectus, olfactory cortex, caudate, and putamen.²⁵ With DTI, the study also showed higher mean diffusivity values in the genu of the corpus callosum, the orbitofrontal area, the anterior thalamic radiation, and the forceps minor and higher radial diffusivity values in the anterior corona radiata, the genu of the corpus callosum, and uncinate fasciculus compared with the normal olfaction group. Increased mean diffusivity is associated with increased neuronal injury, oedema, and inflammation. Lack of smell is one of the common symptoms of COVID-19. This study showed persistent symptoms to be associated with change in the olfactory cortex along with other striatal and frontal regions.

Du et al explored the cross-sectional and longitudinal GM volume changes in patients with COVID-19 1 and 2 years after discharge.²⁶ Twenty individuals completed follow-ups 1 year after discharge, and 18 completed follow-ups 2 years after discharge. In the longitudinal analysis, 2 patients’ average GM images could not be segmented; finally, 16 patients were included in the longitudinal study. Additionally, 27 controls matched by age, sex, and education were recruited for comparison. Compared to the controls, after 1 year, PASC patients showed significantly reduced GM volume in the bilateral cerebellum and vermis, left middle frontal gyrus and inferior frontal gyrus of the operculum, and right middle temporal gyrus and inferior temporal gyrus. After 2 years, they showed significantly reduced GM volume in the bilateral cerebellum and vermis, left middle temporal, and fusiform gyrus. Díez-Cirarda conducted a multi-modal neuroimaging study of 86 PASC and 36 healthy individuals using DTI, FLAIR, and resting-state functional MRI.²⁷ Exploratory analysis showed reduced GM volume in PASC in parahippocampal gyri, inferior frontal gyrus, anterior cerebellar, occipital, and bilateral superior temporal lobes (Figure 1A and B). Only this study showed decreased mean diffusivity and axial diffusivity in the corpus callosum, forceps minor, medial longitudinal fasciculus, uncinate tract, and fronto-occipital fasciculus compared to controls with DTI (P<.05, Figure 1C). Importantly, this study showed that the GM volume reductions were significantly correlated with cognitive dysfunction, specifically attention, working memory, and processing speed in PASC. This was attributed to reduced perfusion and increased inflammation instead of degenerative processes. Cytotoxic oedema, following inflammation, results in cell swelling and could reduce extracellular diffusion, thereby causing a reduction in diffusivity metrics. Discrepancies in findings across studies could also arise based on the variable timing of imaging relative to PASC or COVID diagnosis, wherein the cytotoxic oedema may not have developed or may have been fully resolved but led into degenerative processes.

Muccioli et al included 23 patients with persistent (2-19 months) COVID-19-related olfactory dysfunction and 26 sex- and age-matched healthy controls without prior olfactory dysfunction.²⁸ Using T1-weighted MRI, 13 bilateral cortical and 3 subcortical regions of interest (ROIs) were identified within the functional olfactory network: anterior cingulate cortex, insular cortex (central, short, anterior, inferior, and superior circular), orbital cortex (inferior frontal, lateral, medial olfactory, H-shaped), rectus cortex, parahippocampal cortex, hippocampus, amygdala, and thalamus. Like the studies above, comparison between groups revealed that the volumes of the left lateral orbital cortex, bilateral hippocampi, and amygdala were reduced in patients with PASC. However, a few studies found no GM volume differences with COVID-19 or PASC. For instance, in a survey by Cecchetti et al, examining 49 individuals 2 months after hospital discharge and 33 (of the 49) individuals 10 months after discharge and comparing with sex- and age-matched healthy controls, no significant differences were found between PASC (new onset of cognitive symptoms at least 1 month after hospital discharge from hospital stay due to COVID-19) and healthy subjects in terms of total brain, GM, and white matter volumes.²⁹ Instead, the PASC patients showed a greater WMH volume than controls, correlating with worse memory function at 2 months post-discharge and with the total number of cardiovascular risk factors.^30–32 No association with acute illness severity indices was detected. No significant cognitive deterioration (global cognition, executive functions, memory, visuospatial functions, language) was observed at the 10 month-timepoint compared to the 2 month-timepoint. Most investigations utilized structural MRI followed by resting-state functional MRI measure to assess COVID-19-related changes in brain function and cognition.^33,34

Graph theoretical analysis of structural data shows network disruption early in PASC which resolves in most participants. Tassignon et al compared whole-brain structural connectivity using T1-weighted scans and DTI between hospitalized COVID-19 patients admitted to the intensive care unit and hence believed to suffer from PASC with those not admitted.³⁵ Participants were followed over 2 months after discharge from the hospital. Network analysis and graph theory measures were applied to the DTI connectomes. Characteristic path length (minimum number of fiber tracts traversed between 2 regions), global efficiency (a measure of parallel processing and information integration that is inversely proportional to the characteristic path length), clustering coefficient (number of connections between adjacent brain regions), and local efficiency (communication efficiency between neighbouring brain regions and is related to the clustering coefficient) were evaluated. The study used a computerized test battery, “Cognition,” to assess memory, sensorimotor, and visuospatial function. It showed that only characteristic path length was greater at discharge, but it eventually decreased by 3 months in all participants. No group differences were observed, likely due to the small sample size. No other metrics show any significant differences. However, the digit symbol substitution and visual object learning tests were associated with the clustering coefficient. A lower digit symbol substitution test score was associated with a lower clustering coefficient in both groups at discharge and at their 2-month follow-up. A lower visual object learning test score was associated with a lower clustering coefficient at discharge, but this relationship was not observed at the 2-month follow-up. Using graph theory, the same study by Muccioli et al, described earlier, found increased global modularity (division of a network into sub-networks or modules with high modularity indicating dense connectivity within modules and sparse connectivity across modules) of the olfactory network in the COVID-19 group.²⁸ This increased modularity was related to the severity of hyposmia. The study also found that increased right thalamic connectivity strength was associated with poor memory function on verbal recall tests.

Resting-state functional MRI studies show cortical and subcortical connectivity changes with PASC, especially in individuals experiencing post-traumatic stress symptoms. Functional connectivity was assessed and compared with a database of 300 healthy individuals.³⁶ Lower connectivity strength was observed within all the lobes, with the left temporal lobe showing the most abnormal connectivity. Very few studies have assessed functional connectivity changes in PASC. One study tested whether anosognosia (lack of conscious awareness of cognitive deficits) is a feature of PASC.³⁷ Over 100 individuals with mild (at home), moderate (hospitalization), and severe (hospitalization in intensive care units with mechanical ventilation) COVID-19 underwent a cognitive battery a little over 6 months after infection. Based on this cognitive assessment, the participants were divided into 2 groups: with and without anosognosia. The study found that functional connectivity was significantly decreased in the anosognosia group within and between the left default mode, the bilateral somatosensory motor, the right executive control, the right salient ventral attention, and the bilateral dorsal attention networks, as well as the right Lobules IV and V of the cerebellum.

Fu et al investigated dynamic functional connectivity in 50 COVID-19 patients and its association with post-traumatic stress symptoms.³⁸ Participants were tested 6 months after discharge from the hospital. Forty-three age and sex-matched controls with no history of COVID-19 identified using antibody testing were used from the community. Typically, studies utilize static resting-state functional connectivity analyses, which assume that brain regions and their connections to each other throughout an entire scan remain unchanged.³⁹ However, it has been demonstrated that connectivity between brain regions varies over time at the level of minutes to seconds in functional MRI, which can be assessed using advanced dynamic functional connectivity analyses.^40,41 This study identified 3 distinct brain connectivity states, ie, different connection patterns between the different resting-state network regions. The brains spent less time in 2 of these states, which differed primarily in the connectivity between the subcortical and visual networks. PASC patients transitioned to one of these states more often than controls and the transition frequency was related to post-traumatic stress disorder checklist for DSM-5 (PCL-5). The third state was the predominant and sparsely connected brain state with weak inter-network connectivity. Within this state, connectivity of the thalamus and inferior parietal lobe was greater in PASC patients than controls and was also associated with PCL-5 scores.

Another study of post-traumatic stress due to COVID-19 assessed 42 survivors without brain lesions about 3 months after infection.⁴² The severity of inflammation, measured in the plasma based on peripheral lymphocyte, neutrophil, and platelet counts, was negatively associated with connectivity in the default mode network (DMN), language, and frontoparietal networks and positively associated with functional connectivity in the dorsal attentional, salience, and sensorimotor networks. A positive association was observed with cerebellar connectivity. Greater levels of C-reactive protein, a marker of inflammation, were associated with functional connectivity between the right caudate and the language network. The severity of post-traumatic distress was negatively associated with connectivity to the medial prefrontal cortex and positively associated with connectivity to the bilateral posterior cingulate cortex/precuneus in the DMN, superior temporal cortex in sensorimotor network, and right middle frontal gyrus. Tu et al studied post-traumatic stress symptoms 3 and 6 months after hospital discharge of COVID-19 patients.⁴³ They found severe post-traumatic stress symptoms after COVID-19, with severity increasing from 3 to 6 months. The total score of post-traumatic stress disorder (PTSD) checklist for DSM-5 (PCL-5) was greater in PASC participants and associated with lower hippocampal and amygdala volumes and greater connectivity between these structures than controls (without SARS-CoV-2 infection or cognitive symptoms).

Similar to MRI, FDG PET findings show abnormalities, specifically hypometabolism, primarily in the frontal and temporal structures: Kiatkittikul et al used F-18 fluorodeoxyglucose (FDG) PET/CT for whole body imaging and a brain PET/MRI to evaluate 13 patients with PASC.³⁶ Brain FDG PET showed hypometabolism in the parietal and temporal lobes in most patients and the frontal and occipital lobes in less than 50% of the patients (Figure 1A). They also performed functional connectivity in the same participants and compared it with a database of 300 healthy individuals. Like FDG PET, lower connectivity strength was observed within all the lobes, with the left temporal showing the most abnormal connectivity. Another study by Ferrucci et al, using PET/CT, studied 7 patients with PASC.^44,45 About 50% of the patients had abnormal metabolism in the temporal lobe, pons, prefrontal, and middle frontal cortex. In a study of 13 adult PASC patients compared to 26 melanoma patients, FDG PET showed specific hypometabolism in the hippocampus and the thalamus. The regional hypometabolism was associated with symptoms of anosmia and fatigue. In a study of 28 participants who presented with post-COVID-19 conditions and satisfied the WHO definition of PASC and 28 age- and sex-matched healthy controls, FDG PET showed a hypometabolic imaging pattern implicating the right frontotemporal lobes.⁴⁶ Independent of the imaging, a more significant number of clinical symptoms during the initial phase of infection and a longer duration of symptoms were risk factors for persistent asthenia, anosmia, and arthromyalgia.

Discussions

Many common trends in neuroimaging findings have been outlined above. All MRI and PET findings indicate alterations in the temporal and frontal regions of the brain, specifically in the hippocampal and middle temporal regions and the middle frontal region.^47–54 Structural imaging shows atrophy in these regions while functional imaging tends to show a breakdown of typical connectivity patterns especially in the olfactory, hippocampal, and amygdala regions. However, the use of neuroimaging as a biomarker of PASC is difficult given the clinical heterogeneity of this condition as well as the prevalence of individuals infected with COVID-19. Therefore, several considerations should be highlighted.

First and foremost, the clinical diagnosis of PASC has evolved, and studies reviewed in this work did not all identify PASC using identical criteria or include individuals with PASC.55–57 Most commonly, individuals with extended hospital stays for COVID-19 or post-traumatic stress symptoms were studied and were most likely to suffer from symptoms that mirror those experienced by patients with PASC. Secondly, PASC patients were studied at varying intervals ranging from a few weeks to about 2 years after SARS-CoV-2 infection. PASC symptoms may vary in severity following acute infection and fluctuate over the course of 2 years. Thirdly, the role of comorbidities, such as a sedentary lifestyle or cardiovascular conditions prior to infection, needs to be clarified, and their prevalence is not necessarily detailed in the study data. Fourth, almost all studies show a relationship between PASC symptoms and imaging findings. However, association with pathological markers such as inflammation, endothelial damage, oxidative stress as mediators of the neurological findings in PASC is understudied. All these factors can confound imaging findings and mask the signature of PASC as defined in the studies above. The most recent definition of PASC is post-acute sequelae of SARS-CoV-2 infection (PASC), also known as Long COVID and defined as ongoing, relapsing, or new symptoms or conditions present 30 or more days after infection.⁵⁸

Structural MRIs with T1- and T2-weighted sequences have consistently shown temporal and frontal cortical thinning in PASC. However, other structural modalities, such as FLAIR and DTI, have yet to be studied in detail in PASC. FLAIR can be used to assess WMHs and have been leveraged in a few studies to date. WMHs measure cerebrovascular disease burden and are associated with hypertension, dyslipidemia, diabetes, carotid atherosclerosis, and heart failure. Recent studies suggest that COVID-19 increases systolic and diastolic BP and may cause new-onset hypertension.⁵⁹ Chung et al showed that SARS-CoV-2, directly and indirectly, results in immune response mechanisms that impact the cardiovascular system.⁶⁰ Additionally, acute cardiac injury secondary to myocarditis or ischaemia is a common extrapulmonary manifestation of COVID-19. Although the pandemic has subsided, the National Center for Health Statistics found that 6.9% of adults (∼18 million) have suffered from PASC and 3.4% of adults (∼10 million) currently exhibit PASC in the United States.

Despite these challenges, the ongoing research underscores the importance of neuroimaging in advancing our understanding of PASC. It is necessary to better understand the underlying mechanisms of PASC, including the potential causes, risk factors, and how it affects the brain. By understanding who is more likely to experience PASC, healthcare providers can better tailor their care and support for patients at higher risk. This can lead to targeted interventions, such as earlier and more focused treatment, rehabilitation programs, and supportive care, ultimately improving outcomes for individuals with PASC. Future research should focus on continuing the standardization of diagnostic criteria for PASC, exploring the long-term trajectories of neuroimaging changes, and investigating the interplay between COVID-19, comorbidities, and neuroinflammation. Additionally, expanding the use of advanced imaging techniques, such as FLAIR, DTI, and arterial spin labelling (ASL),⁶¹ can provide deeper insights into the alterations in the brains of individuals with PASC. Similarly, more accessible and cost-effective techniques such as optical coherence tomography could be used to study retinal and vascular damage in PASC and could be indicative of vascular or immune-related damage in the brain.⁶²

Conclusion

In summary, while the neuroimaging landscape of PASC is still being charted, the evidence to date highlights the profound impact of COVID-19 on the brain and underscores the need for continued research and clinical vigilance to address the long-term effects of this pandemic on neurological health. Information gained will not only help develop targeted interventions and improve outcomes for those affected by Long COVID-19 but also inform studies investigating downstream effects of known and new viral infections on the brain.

Funding

This work was funded by grant titled, ‘Chronic Post COVID-19 Infection Neuroimaging and Cerebrovascular Imaging’ awared to authors and PI, MMB and Co-PI, SRL, from Bayer Healthcare LLC and by grant titled, ‘Long terms effects of COVID-19’ awarded to authors and PI MMB and Co-PI SRL.

Conflicts of interest

None declared.