Hypometabolism of the left middle/medial frontal lobe on FDG‐PET in anti–NMDA receptor encephalitis: Comparison with MRI and EEG findings

Abstract Objectives To investigate changes in brain‐glucose metabolism in anti‐N‐methyl‐D‐aspartate receptor (NMDAR) encephalitis, and compare results with MRI and electroencephalography (EEG) findings at different disease stages. Methods The clinical data of 18 patients (median age, 35 years; 11 men) were retrospectively collected. Patients were divided into groups based on the time of symptom onset to examination, (≤1 month, >1 but ≤3 months, >3 months). Two‐sample t‐test results were compared with age and sex‐paired healthy controls using statistical parametric mapping and verified using a NeuroQ software normal database with a discriminating z‐score of 2. Results Abnormal patterns on FDG‐PET differed over time (T = 3.21–8.74, Z = 2.68–4.23, p < 0.005). Regional analysis showed hypometabolic left middle or medial frontal cortex in 4/5, 5/7, and 5/6 patients, respectively. Time‐subgroup analysis revealed hypermetabolic supertemporal cortex in 4/5, 5/7, and 2/6, patients, respectively. MRI and EEG abnormalities in any region and stage occurred in 10/18 and 10/16 patients, respectively. MRI and EEG time‐subgroup analysis showed abnormalities in 5/9, 4/5, and 1/4, and 1/3, 6/7, and 3/6 patients, respectively. Abnormal temporal lobes were detected most frequently in MRI analyses and occurred in 3/10 patients. Conclusions Decreased left middle/medial frontal metabolism could be common to all stages. Metabolism in other regions, MRI, and EEG results were associated with the progression of anti‐NMDAR encephalitis. The sensitivity rate of FDG‐PET was superior to that of MRI and EEG.


| INTRODUC TI ON
Anti-N-methyl-D-aspartate receptor-autoimmune encephalitis (NMDAR-AE) is the most common type of AE. Fortunately, NDMAR-AE can be treated with immune-related therapies that significantly improve prognosis. 1 This disease presents various complex neural and mental symptoms that often lead to a delay in diagnosis. 1,2 The diagnostic criteria mainly depend on detecting antibodies in serum and cerebrospinal fluid (CSF), and monitoring immune response to treatment. 1,3 However, these methods are invasive and are often not applicable during early evaluation and management.
FDG-PET and MRI examinations are important and recommended for auxiliary diagnosis of NMDAR-AE. 1,[3][4][5][6][7][8][9][10][11][12] Previous studies showed that FDG-PET provided earlier detection and had a higher diagnostic sensitivity rate than MRI. [4][5][6][13][14][15][16][17] However, there is no consensus on the abnormal patterns of FDG-PET, including hypo or hypermetabolism of the frontal, temporal, and occipital lobes, and the basal ganglia and cerebellum. To the best of our knowledge, most studies have been case reports or included fewer than five patients per group with subgroup analysis, 4-9,12-20 causing confusion regarding the accurate diagnosis of the disease. Electroencephalography (EEG) is useful for auxiliary diagnosis and evaluation of prognosis. [21][22][23][24] This study focused on the changes in FDG-PET results with disease progression in anti-NMDAR encephalitis and compared the clinical auxiliary diagnosis with MRI and EEG results at different stages of the disease.

| Clinical data
Patients who had undergone hospitalization for suspected encephalitis and 18  This retrospective cohort study was approved by our institutional review board, and the requirement for obtaining informed consent was waived.
Clinical data considered potentially relevant to AE were collected from patient medical records. Data included age at FDG-PET/CT imaging, sex, clinical history, CSF and serum antibodies, brain MRI scans, EEG, FDG-PET/CT treatment, and tumor histology within at least the previous 5 years from disease onset. The subgroup analysis was based on the timing of PET scans after disease onset three groups: ≤1 month, >1 but ≤3 months, >3 months, which represent acute, subacute, and chronic phase. All patients were stable at the time of imaging without the use of sedatives or tranquilizers.

| Neuronal antibody measurement
All patients with NMDAR-AE underwent serum and CSF antibody tests. Serum and CSF immunoglobulin titers were measured using cell-based and dot immunobinding assays. The reported routine-awakening EEG data were identified as normal or abnormal based on patient medical records.

| Visual assessment of FDG-PET/CT and MRI
Visual assessment of PET images was performed separately by two experienced nuclear-medicine physicians who were aware that the patients had suspected encephalitis, but did not have any information from other examinations such as MRI and EEG in advance.
Regions with metabolic changes compared to the contralateral were visually identified as abnormal, and bilateral lobe metabolism changes were compared to the cerebral cortex using maximum intensity projection. The cerebellar metabolism is slightly lower than that of the cerebral cortex, and abnormal hypermetabolism occurs when the cerebellar metabolism is equal to or exceeds that of the cerebral cortex.
Two well-trained radiologists who were blinded to the PET information were involved, and identified hyperintense lesions on T2, FLAIR, or DWI images, with the exclusion of other nonspecific causes such as age-related white-matter hyperintensities, lacunar infarction, and brain atrophy.
Electroencephalography data were retrieved from the medical records.

| Software analysis of FDG-PET/CT images
Considering age and sex differences in brain FDG metabolism, we initially used a fully-paired control to create a healthy normal database that determined possibly-related brain regions by statistical parametric mapping (SPM) analysis to make a voxel-wise comparison of anisotropy in the whole brain. The voxel-wise analysis for glucose metabolism was analyzed by counting the (18) Before statistical analysis, all images were spatially normalized into a standard brain space by global grand-mean scaling and smoothed using SPM-8 in a MATLAB R2018b environment (MathWorks, Inc.). Following validation, FDG-PET processing was performed based on an optimized semiquantitative procedure (details described above). 25,26 Next, images were tested for relative whole-brain hypo and hypermetabolism using a two-sample t-test implemented in SPM that compared the AE group with a group of age and sex machine-paired healthy individuals without any neurologic or psychiatric diseases. Because only three patients were scanned with a different apparatus, we used machine-matched healthy individuals as controls, and the different PET scans were not entered as a nuisance covariate. The resulting SPM-t maps, with a threshold at p < 0.005 (k = 50), represented the basis for the assessment of regional hypo and hypermetabolism.
NeuroQ implanted in Philips IntelliSpace Portal software v7.0 (Philips Healthcare) was used to process 18 F-FDG-PET/CT brain images. After automatic rigid deformation, the entire brain was divided into 47 regions. The metabolism of each region was calculated by comparison with the default software database. Areas of hypo or hypermetabolism in the brain differing from the mean were recorded as abnormal using a cutoff z-score value of 2.0.

| Patient characteristics
A total of 53 patients with suspected encephalitis who underwent PET-CT at our hospital between January 2015 and March 2021 were retrospectively identified. Eighteen patients were included in the analysis ( Figure S1 shows the flow diagram

| Analysis on 18 F-FDG-PET imaging and by visual assessment on MRI
Abnormal PET metabolic patterns were found in 18/18 (100%) patients, and 10/18 (55.56%) and 10/16 (62.5%) patients using MRI and EEG, respectively. Therefore, the sensitivity rate of PET/CT for discriminating brain lesions in patients with NMDAR-AE was higher than that of MRI and EEG using visual assessment (

| Regional analysis by software on 18 F-FDG-PET imaging for all patients
We found significant hypometabolism in all patients in the frontal lobe, parietal lobe, cingulate gyrus, and occipital lobe (t = 2.75-  Table 3 and are similar to the SPM results. A typical example (pat.6) is shown in Figure S3. Figure

| Subgroup analysis of changes on MRI and EEG
Considering that patients did not undergo all same examination at the three stages, we listed the specific times for each examination in Table 2.
Data analyses were also grouped by MRI examination time although this grouping may have contributed to some study limitations.
We found 5/9 (55.56%) patients with abnormal MRI findings in the first month, 4/5 (80%) between 2 and 3 months, and only 1/4 (25%) at >3 months. Single-region MRI analysis showed that the unilateral temporal lobe (the most involved region) was abnormal in 3/10 (30.0%) patients at all stages, and the highest abnormal rate was found in only 2/5 (40.0%) patients between 2 and 3 months in the left temporal lobe.

| DISCUSS ION
Abnormal FDG-PET brain scans have been reported in anti-NMDAR encephalitis patients but metabolic patterns have been inconsistent across different regions. 4-20, 25, 27-30 In this study, we investigated the application of FDG-PET during different stages of anti-NMDAR encephalitis disease progression. First, notably (but unsurprisingly) we found that consistent with other diseases, the brain-glucose metabolic patterns changed during different stages (p < 0.05). 10,31,32 Second, hypometabolism of the middle/medial frontal cortex was common to all stages in our study (14/18 and 13/18 patients, respectively), although not all of them. The results were consistent with other studies on MRI, 33-37 which reported decreased gray matter volume and reduced cerebral regional connectivity in left middle/medial frontal gyrus, they may have significantly correlated with memory performance and verbal inhibition F I G U R E 1 A typical case of the PET scan time is <1 month from symptom onset (pat. 6 had mental status changes and seizures for half a month, accompanied by cognitive impairment and speech disorders, and was diagnosed with ovarian teratoma). FDG-PET in 21 days revealed that significant hypometabolism in left middle frontal, hypermetabolism in bilateral basal ganglia and left temporal lobe (A). However, cerebral MRI in 38 days shows unremarkable signal alterations on T2-FLAIR image (B), DWI (C) and contrast enhanced MRI imaging (D). The FDG statistical deviation images analysis based on the z-score (E).

F I G U R E 2
A typical case of the PET scan time is 2-3 months from symptom onset (pat. 17 had intermittent fever for 2 months, cognitive impairment, and speech disorders accompanied by abnormal behavior). FDG-PET (A) at approximately 63 days revealed that hypermetabolism in left super temporal, hypometabolism in the left middle frontal and right frontal-parietal-temporal lobes (especially in the right temporal lobe), which correspond to mild hyperintense in MRI areas. T2-FLAIR (B) and DWI (C) of axial MRI in 49 days showed multiple asymmetric white matter hyperintensities, especially in the right frontal-temporal lobe, and lateral ventricle. There was a lack of enhancement in mentioned above areas of contrast-enhanced MRI (D). The FDG statistical deviation images analysis based on the z-score (E).
control in NMDAR encephalitis. Moreover, according to our previous research on the cerebral metabolic network, 26 the hub nodes are mainly located in the right frontal lobe in anti-NMDAR encephalitis instead of in the left frontal lobe, as it is in healthy controls.
We propose that hypometabolism of the left middle or medial frontal cortex may imply impaired neurons and that the right frontal lobe is a compensatory change. Third, we found that hypermetabolism of the superior temporal gyrus was more evident in the first 3 months (9/12 patients). Miao A. also found increased cerebral blood flow at the peak stage of the disease, 38  later. Furthermore, our study found that the later the therapeutic intervention, the more likely the patient was to relapse. Three of four patients who relapsed were treated after 3 months from symptom onset and relapsed 3-6 months later. Another patient was treated at 45 days and relapsed at 14 months. These data imply that earlier treatment results in a better prognosis. Moreover, our results were similar to those of other studies in general, 6,7,9,11,25,[27][28][29][30]  (Continues) treatment effects on PET images. However, we listed the time and will include therapy in future studies. This study aimed to evaluate the effects over time in patients with anti-NMADR antibodies by using PET/CT, MRI, and EEG, but to overcome the abovementioned limitations, increased sample sizes are required in future studies.
In conclusion, this study suggested that hypometabolism of the middle or media frontal cortex could be a common metabolic pattern over the disease course, whereas the metabolism of other regions and the performance of MRI and EEG reflected time-varying changes associated with progression. The sensitivity rate of FDG-PET was superior to that of MRI and EEG. These results have potential value for the diagnosis and clinical management of patients with anti-NMDAR-AE.

ACK N O WLE D G E M ENTS
This study was funded by the Shanghai Aging, Women, and

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data generated or analyzed during the study are available from the corresponding author by request.

PATI ENT CO N S ENT S TATEM ENT
The requirement for obtaining informed consent was waived. Right cerebellum 3 50.00

O RCI D
Note: Because the visual cortex and Broca's region are commonly affected by function, we do not list those regions and they should be interpreted with caution. Unilateral suggested no matter the left/right or both cortex involved in a patient, count only once (in italics).