Introduction

Since December 2019, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is causing a worldwide outbreak with substantial morbidity and mortality. Neuropsychiatric complications of COVID-19 including encephalopathy, encephalitis and anxio-depressive disorders have been increasingly described [1], but little is known about their neuronal substrates and pathogenic mechanisms. These manifestations could be caused by direct effects of the virus on the nervous system, para-infectious or post-infectious immune-mediated disease, or neurological complications of the systemic effects of COVID-19 and critical illness, these hypotheses being not mutually exclusive. Several specific patterns in structural magnetic resonance imaging (MRI) have been described but it remains often non-contributive [2] and functional imaging data is still limited. In line with published cases [3, 4], a previous case series of four patients suggested the involvement of the prefrontal cortex and cerebellum in acute COVID-19-related encephalopathy using 18F-fluorodeoxyglucose (18F-FDG) PET/CT analysed visually [5]. However, the interpretation of these results was limited due to the qualitative nature of the data. Here, we report a series of seven patients with acute encephalopathy in the context of COVID-19, who were investigated during the acute phase and longitudinally up to 6 months after COVID-19 onset with 18F-FDG-PET/CT analysed quantitatively in comparison with healthy controls using regions-of-interest and voxel-based approaches.

Methods

Patients

Seven consecutive cases of COVID-19-related encephalopathy (median age 63 y/o, range 50–72 y/o; 4 men) hospitalized in Neurology or Psychiatry wards were explored with 18F-FDG-PET/CT as part of their clinical investigation, from March to June 2020 in the Pitié-Salpêtrière Hospital in Paris, France.

The diagnosis of COVID-19-related encephalopathy was based on new-onset cognitive impairment with central focal neurological signs or seizures in the context of COVID-19, in the absence of another cause of encephalopathy. The diagnosis of COVID-19 was confirmed in all patients by a positive reverse-transcriptase polymerase chain reaction (RT-PCR) assay from a nasopharyngeal swab sample. All patients underwent cerebrospinal fluid (CSF) analysis, electroencephalogram (EEG), brain MRI and brain 18F-FDG-PET/CT. Other infectious and autoimmune disorders were excluded by extensive diagnostic workup. The evolution of their neuropsychiatric symptoms, including cognitive, behavioural and emotional status was assessed within the first 2 months, and 3 to 6 months after COVID-19 onset by a multidisciplinary team in our memory clinic. A comprehensive battery of validated cognitive tests was used to test global cognitive performances, working memory, executive functions, attention, verbal and visual episodic memory, praxis, visuo-constructive abilities and social cognition. The speech therapist tests battery assessed the global language abilities, and more specifically fluencies, denomination and semantic abilities. Emotional aspects were assessed by an interview conducted by a psychologist.

Patients received information and agreed to the use of their medical data in accordance with French regulations. The study was approved by the Sorbonne University Ethics Committee (CER-202028 on 24/04/2020).

Brain PET imaging

PET imaging was performed at the acute phase, 1 month later and 6 months after COVID-19 onset, following the same protocol: a 10-min acquisition was performed with a hybrid PET/CT system (Biograph mCT Flow, Siemens Healthcare) 30 min after the injection of 18F-FDG (2 MBq/kg). Images were reconstructed using an iterative algorithm and corrected for attenuation and scatter.

PET images were spatially normalized into the Montreal Neurological Institute (MNI) space with SPM8 (https://www.fil.ion.ucl.ac.uk/spm/), smoothed and normalized in intensity using the pons as reference region. Two-sample t tests were used to identify glucose metabolism changes in the whole patient group and in each patient individually in comparison with 32 healthy subjects (age 56.3 ± 8.7 y/o; range 40–72 y/o; 19 men). The same PET/CT system and imaging protocol were used for patients and controls. Age was set as a confounding variable. SPM(T) maps were thresholded with the primary defining threshold set at p < 0.05 at the peak level, corrected with the family-wise error rate (FWE) method except for one contrast (patients > controls). For the latter analysis, the threshold was decreased at p < 0.001 uncorrected at the peak level. A region-of-interest (ROI) analysis was performed with Scenium software (Siemens Healthcare).

Results

Clinical and imaging features during acute encephalopathy

Patients’ clinical presentation, main biological and imaging features are given in Table 1. A detailed description of four of the patients is available in a previous paper [5]. The patients had varied COVID-19 severity: three patients required mechanical ventilation (#1, 2, 7) and three required nasal oxygen (#4, 5, 6). Patient #3 was hospitalized but did not need oxygen. They had wide-ranging neurological presentations and variable degrees of cognitive dysfunction. They all experienced severe executive deficit and frontal behaviour changes with apathy. A cerebellar syndrome was observed in two patients (#3, 5), myoclonus in two (#1, 5), psychiatric manifestations in five (#1, 3, 4, 6, 7) and oculomotor disturbances in two (#1, 2). One patient had status epilepticus (#7) and one patient had a focal motor seizure (#4). One patient had delayed awakening (#1) and one patient had delirium (#2) following sedation stopping in the intensive care unit. Blood IL-6 levels were increased in 5/5 patients and CSF IL-6 levels in 3/4 patients. One patient had CSF leukocytes > 5 mm3 (#5), and two patient had CSF protein > 0.4 g/L (#2, 7). SARS-CoV-2 RT-PCR in the CSF was negative in all patients.

Table 1 Summary of the main clinical, biological and imaging findings in the patients with COVID-19-related encephalopathy
Full size table

The median delay between the onset of COVID-19, neurological symptoms and 18F-FDG-PET/CT was 7 days (range 0–29) and 33 days (range 15–89), respectively.

In the whole group of patients vs. controls, voxel- and ROI-based analyses showed a prominent hypometabolism of the prefrontal cortex prevailing on the right side, bilateral insula (14% decrease vs. controls), anterior cingulate (− 20%), right (− 33%) and left (− 26%) caudate nucleus (p < 0.05 corrected, Fig. 1, Tables 2 and 3). The SPM-based analysis showed mild hypermetabolism in the vermis, dentate nucleus and pons (p < 0.05, Table 2).

Fig. 1
figure 1

Brain metabolism changes in the whole group of patients (n = 7) vs. controls (n = 32) in the acute phase, 1 month later and 6 months after COVID-19 onset. Hot and cool color scales show regions with hypermetabolism and hypometabolism in patients vs. controls, respectively. The SPM maps are projected onto a surface rendering and onto axial views of the customized MRI template. The axial slices are shown using neurological conventions (right is right). R, right; L, left. At baseline, hypometabolism was found in the bilateral prefrontal cortex with right predominance, insula, anterior cingulate and caudate (p < 0.05 corrected). The analysis showed a mild hypermetabolism in the vermis, dentate nucleus and pons (p < 0.05). One month later, hypometabolism was limited to the mediofrontal, right dorsolateral areas, olfactory/rectus gyrus, bilateral insula, right caudate nucleus and cerebellum (p < 0.001). Six months after COVID-19 onset, metabolism decrease was observed in the same regions but they were less extended (p < 0.001)

Full size image
Table 2 18F-FDG-PET findings in the whole group vs. 32 healthy controls with the SPM approach, at the acute phase, 1 month later and 6 months after COVID-19 onset
Full size table
Table 3 18F-FDG-PET findings in the whole group vs. healthy controls: results of ROI analysis in acute phase, one month later and 6 month after COVID-19 onset
Full size table

Individual analyses confirmed a severe, bilateral prefrontal hypometabolism in 7/7 patients, including the rectus gyrus (p < 0.05 FWE corrected, Fig. 2, Table 4). In every patient, hypometabolism expanded bilaterally to the anterior cingulate and insula (p < 0.05 FWE corrected, Table 4). A mild to moderate parietotemporal associative hypometabolism was observed in 4/7 patients, whereas two others had a severe, spread parietotemporooccipital hypometabolism (#6, 7). Hypermetabolism was found in the visual cortex in patient #1 due to visual stimulation after 18F-FDG injection and in the left frontal cortex in patient #4 due to focal seizure. Notably, bilateral mesiotemporal hypermetabolism was found in one patient (#6) without T2-MRI hyperintensity, whereas hypometabolism was observed in patient #3 with known hippocampal sclerosis and in patient #5 having hippocampal atrophy. Metabolism was decreased in the caudate nucleus in five patients (#2–5, 7) and thalamus in two (#2, 3) (p < 0.05 corrected). Metabolism was decreased in the bilateral cerebellar hemispheres in all patients except one (#4) but increased in the vermis in 5/7 patients (p < 0.05).

Fig. 2
figure 2

Brain metabolism changes in each patient vs. controls (n = 32) in the acute phase, 1 month later and 6 months after COVID-19 onset. SPM T maps generated from patient #1 to patient #7 are displayed with a hot color scale for brain hypermetabolism (p < 0.05) and with a cool color scale for brain hypometabolism (p < 0.05 corrected), in axial orthogonal views in neurological convention (right is right). R, right; L, left. 3D rendering shows for all patients the right hemisphere except in patient #4 who presented hypometabolism in the left frontal cortex due to focal seizure

Full size image
Table 4 SPM findings in each patient vs. healthy controls, in acute COVID-19-related encephalopathy
Full size table

Structural MRI showed no cerebrovascular disease or specific COVID-related abnormalities except for one patient #1 presented with typical white matter enhanced lesions, as previously described [2]. Metabolism changes were not associated with morphological abnormalities except in patient #7 previously described [6]. This patient with status epilepticus presented with T2-MRI hyperintensities in the right prefrontal cortex and caudate nucleus, possibly due to peri-ictal diffusion abnormalities. Arterial spin-labelling perfusion imaging revealed frontal hypoperfusion and cerebellar hyperperfusion in 2/7 and 1/7 patients, respectively (Table 1).

Clinical and brain PET follow-up

All patients clinically improved within 2 months of disease onset, with remaining symptoms of variable severity (Table 1). Follow-up at a median delay of 5.1 months after COVID-19 onset (range: 3.6–8.3) confirmed that all patients, except one (#4) had improved with the normal physical neurological examination and had recovered normal autonomy in daily living. However, they all still had an abnormal cognitive evaluation with at least an attention/executive deficit. They also presented with depression and anxiety of varying severity: two patients (#2, 7) had well-controlled symptoms under psychotropic treatments, and one patient (#5) developed symptoms associated with post-traumatic stress disorder. After having partially recovered, patient #4 subsequently relapsed with severe depression associated with psychotic symptoms, hallucinations, worsening cognitive impairment and extrapyramidal syndrome, 2.6 months after disease onset.

In agreement with the clinical course, brain metabolism improved significantly 1 month after the first examination (median delay 28 days, range: 21–56) with less marked cortical hypometabolism and no more hypermetabolic areas (Fig. 1). Six months after COVID-19 onset (median delay 5.9 months, range 4.9–6.5), brain metabolism measured on the whole patient group vs. controls, returned almost normal with mild residual hypometabolism in the rectus/olfactory gyrus (with 9% decrease vs. − 19% at baseline), right insula (10% vs. 14%), anterior cingulate (− 13 vs − 20%) and right caudate nucleus (− 21% vs. − 33%) (Fig. 1, Tables 2 and 3). However, when assessed individually, follow-up revealed some degrees of heterogeneity across patients. Metabolism returned to almost normal in three patients (#1, 5, 6, Fig. 2, Table 4). In patients #2, 3 and 7, the improvement was moderate with prefrontal hypometabolism persisting 6 months after disease onset. In agreement with clinical course, patient #4 first improved and got worse again with severe hypometabolism appearing in the bilateral occipital cortex, and prefrontal and cerebellar hypermetabolisms (Fig. 2, Table 4).

Discussion

A pattern of 18F-FDG-PET/CT hypometabolism in a widespread cerebral network including the prefrontal cortex, anterior cingulate, insula and caudate nucleus was observed in seven patients with acute COVID-19-related encephalopathy. Longitudinal data suggest that this network remains mildly to severely impaired 6 months after disease onset.

Despite varied clinical symptomatology and variable degrees of cognitive impairment in acute COVID-19-related encephalopathy, a pattern of orbitofrontal, dorsolateral and mesiofrontal hypometabolism was found in all patients and was consistent with patients’ prominent clinical frontal lobe syndrome. These results reinforce the hypothesis of major frontal lobe impairment in COVID-19-related encephalopathy, in line with [3,4,5]. These regions-of-interest and voxel-based 18F-FDG-PET comparisons of a larger number of patients with healthy controls also provided further, unexpected insights, supporting the hypothesis of a large network dysfunction. Interestingly, every patient had severe hypometabolism within a network of functionally related areas, namely the insula, anterior cingulate and caudate nucleus. The insula has a key role in processing subjective awareness and integrating homeostatic information from the body. Insular disturbances could be implicated in the deregulation of respiratory failure perception which has been reported in patients with COVID-19 [7]. The anterior insula has a role in higher order consciousness and is highly connected with regions involved in social cognition, decision-making, motivational and affective processes, such as the orbitofrontal, prefrontal dorsolateral, anterior cingulate cortices, amygdala and hippocampus [8]. The anterior insula is also connected with the caudate nucleus which is involved in many associative, executive, motivational and affective processes. Moreover, in our series, the cerebellum was concerned by metabolism modifications, while this structure is implicated in the modulation of cognition (as executive function, attention, language) and, in affects via functional connexions as limbic structures or fronto-parietal cortex [9]. Thus, the hypometabolism observed in this network could be the neural substrate of motivational and emotional disturbances observed in our patients. In line with [4], our findings also suggest an impairment of the thalamus and pons, which could contribute to respiratory control impairment in some patients in the acute phase.

Clinical follow-up with longitudinal PET studies in COVID-19-related encephalopathy are still lacking in the literature. In our series, clinical, neurological examination and brain metabolism improved early in the course of the disease. However, attention/executive deficit with prefrontal hypometabolism was persisting in all patients 6 months after COVD-19 onset. Lasting anxio-depressive symptoms were also observed in agreement with persistent hypometabolism in the right insula which is involved in post-traumatic stress disorder [10]. Our data also reveals heterogeneity in patient outcomes. While some patients almost returned to normal metabolism (even those with initial widespread decrease), others partially improved or worsened. After having first improved, patient #4 subsequently worsened with clinical symptoms and a new pattern of brain hypometabolism which raises the question of an associated neurodegenerative disorder.

Underlying mechanisms of COVID-19-related encephalopathy are still debated. Our findings, such as the increased IL-6 levels in the blood and CSF in some patients may support a para-infectious cytokine release, post-infectious antibody- or cell-mediated immune mechanism. Cytokines can pass the blood-brain barrier, induce central inflammatory responses and influence neurotransmitter metabolism and neural plasticity. They can induce dysfunction in areas implicated in emotional and behavioural regulation and cognition (such as the prefrontal cortex, basal ganglia) and fear and anxiety-related regions (such as the amygdala, insula and anterior cingulate cortex) [11]. The mesiotemporal and/or cerebellar hypermetabolism in the acute phase, disappearing during the course of the disease with clinical recovery, also supports an immune mechanism [12]. No patient had SARS-CoV-2 in the CSF and/or meningitis nor 18F-FDG-PET anomalies limited to the olfactory gyrus that could corroborate a direct viral neuro-invasion.

Considering that 18F-FDG-PET findings are more strongly associated with clinical symptoms, disease course and status than MRI (in the exception of cerebrovascular events) [12, 13], it should be considered for the initial workup and for monitoring treatment in COVID-related encephalopathy. It could also help to further decipher the neuronal substrates of COVID-19-associated encephalopathy. Further longitudinal 18F-FDG-PET/CT studies with a larger sample size are needed to understand the pathophysiological bases of metabolism changes.