Cerebral glucose hypometabolism and hypoperfusion of cingulate gyrus: an imaging biomarker of autoimmune encephalitis with psychiatric symptoms

We retrospectively reviewed patients diagnosed with AE in Beijing Tiantan Hospital between Match 2019 and Match 2022 (Fig. 1). The inclusion criteria were as follows: Age, 18–80 years old; positive antibodies in serum and/or cerebrospinal fluid (CSF) [1]; ¹⁸F-FDG-PET or ASL MRI scans; and follow-up duration > 12 months (unless the patient died from AE). The exclusion criteria were as follows: structural abnormalities in conventional MRI; seizures or subclinical seizures within 2 h before and after the PET examination; FDG-PET images with insufficient glucose uptake; FDG-PET images and ASL MRI with motion artifacts; and any missing data. We excluded patients with MRI lesions to avoid the impact of structural abnormalities on brain metabolism. Furthermore, in daily clinical practice, AE is often misdiagnosed as primary psychosis when MRI results are negative [17]. All patients were followed-up after at least 1 year from symptom onset and classified into two groups based on the absence or presence of psychiatric symptoms (patients with a good psychiatric prognosis and patients with a poor psychiatric prognosis).

A group of 39 healthy, age- and sex-matched participants were enrolled from the FDG-PET database of Beijing Tiantan Hospital. Ten of them were selected as healthy controls for ASL MRI analysis. The healthy control group had no history of neuropsychiatric disease, including, but not limited to, underlying brain injury, stroke, brain tumors, dementia, Parkinsonism, ataxia, epilepsy, and multiple sclerosis.

PET images were acquired using a PET/CT scanner (Elite Discovery, GE HealthCare, USA) with a matrix size of 192 × 192 and a slice thickness of 3.27 mm. Patients fasted for at least 6 h and their serum glucose levels were maintained below 8 mmol/L. Patients were administered an intravenous injection of 310 MBq/70 kg of body weight of 18F-FDG. PET data reconstruction was performed using the ordered subset expectation maximization (OSEM) algorithm.

ASL scans were acquired using the 3D pseudo-continuous arterial spin labeling technique on a 3 T Siemens Magnetom Prisma MRI scanner (Siemens Healthineers, Germany) with a 64-channel head coil. No contrast agents were used, and patients were required to be stable during scanning. The scanning parameters were as follows: repetition time (TR) = 4705 ms; echo time (TE) = 11.2 ms; field of view (FOV) = 128 × 128 mm; number of excitations (NEX) = 1; 30 × 4.0 mm axial sections with whole brain coverage; scan duration, 6 min and 14 s.

After the acquisition of ¹⁸F-FDG-PET and ASL images, all data were processed using MATLAB 2019b software with statistical parametric mapping 12 (SPM12) software. The cingulate gyrus, including the anterior cingulate cortex (ACC), the midcingulate cortex (MCC), and the posterior cingulate cortex (PCC), was selected as ROI based on the Montreal Neurological Institute (MNI) template, which was subsequently transformed into masks.

PET data preprocessing steps were as follows: first, the PET images were spatially normalized into MNI atlas anatomical space following a 12-parameter affine transformation and nonlinear transformations. This yielded images of 2 mm × 2 mm × 2 mm voxels. Next, to enhance the signal-to-noise ratio, the default SPM smoothing was applied using an 8-mm Gaussian kernel. Finally, the ratio of the standardized uptake value (SUVR) was calculated from the voxel size-weighted median FDG uptake within selected ROIs, normalized to the cerebellum.

In this study, we employed several methods for data analysis. Continuous variables are reported as the median (interquartile range; IQR). Categorical variables are presented as frequencies with corresponding percentages. We compared groups (patients with psychiatric symptoms and patients without psychiatric symptoms) using a t test for continuous variables that were normally distributed (calculated with the Shapiro–Wilk test), Mann–Whitney U tests for non-parametric data, and c² tests or Fisher’s exact tests for categorical variables. The differences in metabolism and perfusion among the three groups were analyzed using one-way ANOVA and the Tukey test for multiple comparisons. Clinical variables were comprehensively collected for possible inclusion in the risk model of psychiatric prognosis. We used binary logistic regression to assess independent risk factors associated with AE with a long-term poor psychiatric prognosis. Then all variables with p < 0.2 in the univariate analysis were included in the multivariable logistic regression model. Subsequently, the likelihood ratio test was used in a backward elimination process (p < 0.05 to retain, p > 0.1 to remove) to select the final set of independent risk factors for inclusion in the model. The odds ratio (OR) with a 95% confidence interval (CI) was presented for the logistic regression model. A two-sided p < 0.05 was considered to indicate statistical significance. SPSS 22.0 software package (IBM Corp., Armonk, New York, USA) and Prism 8 (GraphPad Software, CA, USA) were used for statistical analyses.

A total of 49 patients (32 males and 17 females) with AE were enrolled in this study. Among these, 27 patients presented psychiatric symptoms (referred to as the AE-PS group), while 22 patients (referred to as the AE-WPS) were without them. There were no significant differences in terms of age and gender between the two groups; the median age ranged from 50 to 60 years, with a male predominance (Table 1). Almost all patients developed seizures and memory deficits, regardless of whether they were suffering from psychiatric symptoms (AE-PS, p = 0.196; AE-WPS, p = 0.282). Importantly, AE-PS patients were more likely to have sleep disorders (AE-PS, 48.15%; AE-WPS, 18.18%, p = 0.038) and a smoking history (AE-PS, 33.33%; AE-WPS, 4.55%, p = 0.015). No significant differences were found in any laboratory test of CSF or serum between the two AE groups. Besides, the relapse rate and functional prognosis showed no association with the presence of psychiatric symptoms (AE-PS, p = 0.590; AE-WPS, p = 0.534).

The highest prevalence of psychiatric symptoms was in patients with anti-gamma-aminobutyric acid type B (GABAB) encephalitis (83%), and the lowest was in patients with anti-glutamic acid decarboxylase 65 (GAD65) encephalitis (14%). Furthermore, about half of the patients with anti-N-methyl-d-aspartate receptor (NMDAR) encephalitis (50%) and anti-leucine-rich glioma-inactivated 1 (LGI1) encephalitis (62%) had psychiatric manifestations. Agitation and thought blocking were typical psychiatric phenotypes in all encephalitis with different antibodies (50% in anti-NMDAR encephalitis, 70% in anti-LGI1 encephalitis, and 56% in anti-GABAB encephalitis). However, in anti-GAD65 encephalitis, the predominant symptoms were catatonia and a depressed mood. In addition, hallucinations and delusions were less common in anti-NMDAR encephalitis and anti-LGI1 encephalitis compared with anti-GABAB encephalitis and anti-GAD65 encephalitis (Fig. 2).

In addition to 49 AE patients, we also included 39 healthy controls (HCs). Among them all, 39 underwent PET/CT scans and 10 underwent MRI ASL scans. Metabolic and perfusion differences among the three groups, AE-PS, AE-WPS, and HCs, were compared. Regarding SUVR (Fig. 3), AE-PS patients showed hypometabolism within the whole cingulate cortex, MCC, and PCC compared to AE-WPS patients and HCs. However, in comparison with HCs, ACC hypometabolism was found only in AE-PS patients. Regarding rCBF, AE-PS patients showed hypoperfusion within all parts of the cingulate cortex compared to AE-WPS patients, while there was no perfusion difference compared to HCs. Consistently, no difference in metabolism or perfusion was found between AE-WPS patients and HCs in any cingulum region.

Among the 27 AE-PS patients, 6 (22%) had a poor psychiatric prognosis, indicating that they were still suffering from mental-emotional abnormalities at follow-up. Three of them received anti-psychotic drugs, but the treatment failed, although they all had a good psychiatric prognosis after receiving immunotherapy. Treatment failure is defined as a lack of improvement in psychosis after adequate anti-psychotic drug treatment, including psychiatric rehospitalization, suicide attempt, discontinuation, switch to another medication, or death. As for these three patients, they all switched to immunotherapy [18]. Based on the SUVR comparison between the two groups and the results of univariate analysis, only glucose metabolism in ACC could independently predict poor psychiatric prognosis (p = 0.043) (Fig. 4). No significant correlation was detected between poor psychiatric prognosis and any other characteristics (Table 2). Subsequently, three predictors with p < 0.2, including CSF protein, CSF antibody titer, and SUVR of ACC, were selected to construct a multivariable logistic regression model. A receiver operating characteristic (ROC) curve was obtained, with an area under the curve (AUC) of 0.865, specificity of 0.952, sensitivity of 0.500, and accuracy of 0.852 (Fig. 5).