Neuroimaging Findings of Hospitalized Covid-19 Patients: A Canadian Retrospective Observational Study

Purpose: Coronavirus disease (COVID-19) has been associated with neurologic sequelae and neuroimaging abnormalities in several case series previously. In this study, the neuroimaging findings and clinical course of adult patients admitted with COVID-19 to a tertiary care hospital network in Canada were characterized. Methods: This is a retrospective observational study conducted at a tertiary hospital network in Ontario, Canada. All adult patients with PCR-confirmed COVID-19 admitted from February 1, 2020 to July 22, 2020 who received neuroimaging related to their COVID-19 admission were included. CT and MR images were reviewed and categorized by fellowship-trained neuroradiologists. Demographic and clinical data were collected and correlated with imaging findings. Results: We identified 422 patients admitted with COVID-19 during the study period. 103 (24.4%) met the inclusion criteria and were included: 30 ICU patients (29.1%) and 73 non-ICU patients (70.9%). A total of 198 neuroimaging studies were performed: 177 CTs and 21 MRIs. 17 out of 103 imaged patients (16.8%) had acute abnormalities on neuroimaging: 10 had macrohemorrhages (58.8%), 9 had acute ischemia (52.9%), 4 had SWI abnormalities (23.5%), and 1 had asymmetric sulcal effacement suggesting possible focal encephalitis (5.8%). ICU patients were more likely to have positive neuroimaging findings, more specifically acute ischemia and macrohemorrhages (P < 0.05). Macrohemorrhages were associated with increased mortality (P < 0.05). Conclusion: Macrohemorrhages, acute ischemia and SWI abnormalities were the main neuroimaging abnormalities in our cohort of hospitalized COVID-19 patients. Acute ischemia and hemorrhage were associated with worse clinical status.


Introduction
Coronavirus disease 2019  caused by the SARS-CoV-2 virus is primarily a respiratory illness presenting most commonly with fever, dyspnea and cough. In addition to respiratory symptoms, there is growing evidence of CNS involvement with a range of reported neurologic manifestations including encephalitis, meningitis, and hemorrhagic necrotizing encephalopathy. [1][2][3] In a cohort of 214 patients with COVID-19 in Wuhan, China, 4 36.4% of patients had neurological symptoms ranging from dizziness and headache to dysgeusia and anosmia. Spinal sequelae including transverse myelitis and Guillain-Barre syndrome have also been reported. 5,6 A number of case reports and retrospective studies have been published characterizing brain imaging findings in patients with COVID-19. Acute infarction and hemorrhage are often the most commonly reported abnormalities in an early Italian series, followed by cerebral venous thrombosis, hemorrhagic and non-hemorrhagic encephalopathy, posterior reversible encephalopathy, and Guillain-Barre syndrome. 7 Extensive petechial hemorrhages in the juxtacortical white matter appear to be a common feature. [8][9][10][11][12][13] Additional brain MRI findings describe signal abnormalities of the medial temporal lobe, 8 variably enhancing multifocal white matter hyperintensities, [8][9][10] diffuse leukoencephalopathy, 13 basal ganglia abnormalities, 10 hemorrhagic and non-hemorrhagic encephalopathy, 2,7 and cytotoxic lesions of the corpus callosum. 9,10 The purpose of this study is to systematically characterize the neuroimaging findings in consecutive adult hospitalized COVID-19 patients from a large tertiary hospital network in Canada, and delineate any relationship between these findings and clinical factors.

Study Design
This study is a retrospective, observational study conducted at a tertiary hospital network comprising 3 hospitals in Ontario, Canada. Ethics approval was received from the institutional Research Ethics Board. The study period was between February 1, 2020 and July 22, 2020.
Inclusion criteria were: age >18, patients admitted to a medical ward and/or the intensive care unit (ICU), and a positive diagnosis of SARS-CoV-2 infection using reverse transcriptase polymerase chain reaction (RT-PCR) detection of viral RNA in nasopharyngeal swab, oropharyngeal swab, or bronchoscopy washings. Patients were excluded from the study if they had neuroimaging during the study period but outside of their COVID-19 admission period, for investigation of an alternative pre-existing diagnosis (e.g. brain tumor follow-up), or during the same admission but prior to symptom-onset and/or positive swab in cases of nosocomial COVID-19 infection.

Data Collection
Patient demographics (age, sex), past medical history, neurologic symptoms at and/or during admission, indications for scans, imaging reports, admission location (medical ward vs. ICU), and mortality data were collected from the hospital system's electronic medical records.

Image Acquisition and Analysis
CT scans were performed on one of the following scanner models: Siemens SOMATOM Drive, Flash, and Force (Siemens Healthineers, Erlangen, Germany). MRI was performed on one of the following scanner models using standard clinical coils: Siemens MAGNETOM Sola 1.5 T, Aera 1.5 T, and Avanto 1.5 T (Siemens Healthineers, Erlangen, Germany). Protocols included sagittal T1-weighted, axial DWI, axial T2weighted, axial FLAIR, axial 3D SWI, and time-of-flight MRA (in cases of suspected stroke) sequences. Post-gadolinium sequences, if acquired, included axial 3D T1-weighted MPRAGE with gadolinium contrast, and coronal spin-echo T1-weighted with gadolinium. Images were independently reviewed by 2 fellowshiptrained staff neuroradiologists (YC and MC). In cases of disagreement, images were reviewed and a consensus was reached. Challenging cases were also reviewed at multidisciplinary rounds.

Clinical Data Analysis
Primary outcome measures were abnormalities on neuroimaging and clinical outcome. We defined clinical outcomes using 3 different variables: location of admission (medical ward vs ICU), length of admission (days), and disposition (discharge vs death).

Statistical Analysis
Results are expressed as proportions/percentages, means and standard deviations or minimum and maximum values, as appropriate. Chi-squared tests and Fisher exact tests were used to measure associations between neuroimaging findings and clinical parameters. A P-value of less than 0.05 was considered statistically significant.

Imaging Findings
Out of 103 patients with neuroimaging, a total of 198 neuroimaging studies were performed. This includes 177 CTs (159 CT Head, 5 CT spine, and 13 CTA head/neck), and 21 MRIs (18 MRI brain, 3 MRI spine). 17 patients received at least 1 MRI brain examination. Imaging findings are summarized in Table 2. 17 patients had acute findings, including 9 patients with ischemic changes. 6 out of 9 had large territory infarctions including 2 with hemorrhagic transformation. One patient had global anoxia with hemorrhagic transformation (Figure 2).

Figure 3.
A 58-year-old COVID-19 positive male with a history of type II diabetes, hypertension, hyperlipidemia and end-stage renal disease on dialysis presented with a 3-day history of dyspnea and dry cough. His clinical course required intubation and ICU admission. On day 12 of his admission, he was found to be non-responsive with a GCS of 3. MRI brain (A, B) showed multifocal restricting and FLAIR/T2 hyperintense lesions in the supratentorial and infratentorial brain, including deep white matter, corpus callosum, brainstem, and bilateral lentiform nuclei. Small discrete lesions were even noted in the cervicomedullary junction and upper cervical cord (not shown). The patient was treated with pulse steroid without improvement and unfortunately died. Two patients had imaging findings of small vessel ischemia on MRI (Figures 3 and 4). 10 patients had macrohemorrhages. 7 had a probable or possible underlying cause, including 3 hemorrhagic complications of ischemia, 3 subdural hemorrhage (presumed traumatic), and 1 aneurysmal subarachnoid hemorrhage. The remaining 3 cases had no clear underlying cause, including 1 case of subarachnoid hemorrhage and 2 cases of parenchymal hemorrhage. All 3 of these cases demonstrated elevated d-dimer levels.
4 out of 17 patients who underwent MRI brain demonstrated SWI abnormalities. 3 patients had numerous microhemorrhages involving the juxtacortical white matter, as well as the  corpus callosum ( Figure 5). One patient concurrently had cortical superficial siderosis. All 4 patients with SWI abnormalities were on prolonged mechanical ventilation (>20 days), including one treated with extracorporeal membrane oxygenation (ECMO).
Finally, the MRI of one patient presenting with encephalopathic symptoms demonstrated asymmetric sulcal effacement in the left parietal lobe without enhancement ( Figure 6). Focal encephalitis is possible although the patient's CSF was negative for SARS-CoV-2.

Imaging-Clinical Correlations
In our cohort of hospitalized COVID-19 patients who underwent neuroimaging, those admitted to the ICU were more likely to have acute neuroimaging abnormalities (P < 0.05). ICU admitted patients are also more likely to have acute ischemia and macrohemorrhages (P < 0.05). Patients with macrohemorrhages also had increased mortality (P < 0.05). There was no statistically significant difference in length of admission between patients with and without acute neuroimaging abnormalities (P > 0.05).

Discussion
We present here the neuroimaging findings of 103 patients out of 422 (24.4%) consecutive hospitalized patients with PCRconfirmed COVID-19 admitted to a tertiary care hospital network in Ontario, Canada. Acute neuroimaging findings were found in 16.5% of study patients (17 out of 103 patients with neuroimaging) which were most commonly macrohemorrhages (58.8%), acute ischemic lesions (52.9%), and SWI abnormalities (23.5%, out of a subset of 17 patients with MRI brain). Acute macrohemorrhage was seen in 10 out of 17 (58.8%) patients with positive neuroimaging studies. Macrohemorrhage without clear cause (non-ischemic, non-traumatic, non-aneurysmal) was only seen in 3 out of 17 (17.6%) patients. All 3 patients had elevated d-dimer and severe infection requiring ICU admission. Intracranial hemorrhagic complications in COVID-19 patients have been reported in previous studies, 7,14 but are not as common as thrombotic complications. A multicenter retrospective study by Al-Samkari et al. demonstrated an overall bleeding rate of 4.8% in hospitalized COVID-19 patients. Interestingly, they found an association between elevated d-dimer and both bleeding and thrombosis risk. 15 The patient in Figure 5 had both subarachnoid hemorrhage and focal large vessel thrombosis. Similar to the phenomenon of sepsis-induced coagulopathy, 16 patients with severe COVID-19 infections have been reported to have elevated d-dimer and fibrinogen, as well as platelet reduction and disseminated intravascular coagulation. [17][18][19] Intravascular thrombosis with platelet consumption, coupled with diffuse viral-induced or inflammatory endothelial injury could then lead to multifocal petechial microhemorrhage and eventual macrohemorrhage.
Acute ischemia was seen in 9 out of 17 (52.9%) patients with positive neuroimaging studies, and present in 1.4% of all hospitalized COVID-19 patients. Acute ischemic stroke was found to be one of the most common neuroimaging findings in hospitalized COVID-19 patients in several previous studies, ranging between 23-93% of patients with positive neuroimaging studies. 7,10,14 COVID-19 patients have also been shown to be at higher risk of cerebral ischemic infarcts compared to patients with influenza. 20 Helms et al. compared patients with severe COVID-19 infections with non-COVID-19 matched controls, and found the former at increased risk of intracranial and extracranial thrombosis despite anticoagulation. 21 Two cases of acute ischemia in our cohort demonstrated multifocal small vessel ischemic lesions, with differing MRI appearance. Both were ICU patients with severe respiratory failure and prolonged mechanical ventilation. The patient presented in Figure 3 has innumerable well-defined small infarctions including the bilateral basal ganglia. Bilateral basal ganglia involvement suggests the presence of prolonged hypoxia (i.e. similar to carbon monoxide poisoning and global anoxic injury), 22 and may predispose to the multifocal endvessel infarcts in this case. In a recent French series, symmetric basal ganglia abnormalities were seen in 5% of COVID-19 patients, all with prolonged mechanical ventilation and late awakening after sedation withdrawal. 10 In contrast, the white matter lesions in the patient presented in Figure 4 may not be true infarcts, but rather resemble the described white matter abnormalities from recently published French and Swedish series. [8][9][10] The exact mechanism of this leukoencephalopathy with internal restricted diffusion is unclear, but has been postulated to be secondary to a COVID-19-related vasculitic phenomenon and/or inflammatory disorder with angiocentric involvement. [8][9][10] The underlying cause of large and small vessel cerebrovascular thrombosis in COVID-19 is not well understood. Possible etiologies include a hypercoagulable state, direct viral-induced endothelial damage, and infection-related septic and cardiogenic shock. 23 Patients with severe infections may develop a cytokine storm leading to a state of hyperinflammation causing endothelial injury. 24 Elevated antiphospholipid antibodies have also been implicated. 25 Pathologic findings of autopsies and biopsies of COVID-19 patients showed not only microvascular thrombosis, 17 but both complement-mediated 26 and direct viral-induced endothelial injury. 27 SWI abnormalities have been described as a key finding in several published cohorts of hospitalized COVID-19 patients. The foci of susceptibility were predominantly seen in the corpus callosum and juxtacortical regions, 8,9,11,12 similar to our study. Interestingly, Klironomos et al. found that some foci demonstrated ovoid or tubular appearance suggestive of a microvascular pathology. 9 All 4 of our patients with SWI abnormalities were on prolonged ventilation in the ICU, one of whom on ECMO. Fitsiori et al. found that critically-ill COVID-19 patients often presented with multifocal microhemorrhages with propensity for the corpus callosum. 28 This finding may be more related to prolonged hypoxemia from severe COVID-19 pneumonia, leading to blood brain barrier disruption and resultant cerebral microhemorrhages. 23 Intracerebral microhemorrhages have been described previously in patients with sepsis, critical illness and prolonged respiratory failure without or with ECMO therapy. [29][30][31][32] Finally, study patients admitted to the ICU were more likely to have acute ischemia and macrohemorrhages (P < 0.05).
Patients with macrohemorrhages had increased mortality (P < 0.05). This is in keeping with prior studies showing an association between severity of respiratory symptoms with presence of neurological symptoms. 4,33 This study had several limitations. First, it is a retrospective observational study with a relatively small sample size, which limits its power and external validity. Second, not every patient in this study underwent MRI. Since MRI is more sensitive to microhemorrhage and ischemia than CT, the prevalence of microhemorrhages and small vessel ischemic changes determined for this study population is likely an underestimate. Third, a lack of a control group limits the attribution of the neuroimaging abnormalities reported here solely to COVID-19 infection. Nevertheless, the imaging findings described here are similar to other published studies, which supports the generalizability of our results. Finally, information on long term outcomes and evolution of findings are limited because of a relatively short follow-up duration.

Conclusion
We present the neuroimaging findings in hospitalized adult COVID-19 positive patients in a large Canadian hospital network. The most common findings in our study include macrohemorrhages and acute ischemia (in patients with CT and/or MRI), and SWI abnormalities (in a subset of patients with MRI). ICU patients were more likely to have positive neuroimaging findings overall, as well as more specifically acute ischemia and macrohemorrhages. Macrohemorrhages were associated with increased mortality. To our knowledge this is the first retrospective observational study within the Canadian experience, and adds to the rapidly evolving body of literature in elucidating the neuroimaging spectrum of COVID-19 and its association with clinical outcomes. The presented abnormalities should be kept in mind when ordering and reviewing neuroimaging for COVID-19 patients with neurological manifestations.

Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.