Recognition of seizure semiology and semiquantitative FDG‐PET analysis of anti‐LGI1 encephalitis

Abstract Aims Anti‐leucine‐rich glioma‐inactivated 1 (LGI1) autoimmune encephalitis (AE) is characterized by complex manifestations of seizures. Here, we report a new seizure semiology, attempt to classify the disease by semiology type, and explore the metabolic pattern of each group. Methods Anti‐LGI1 AE patients were retrospectively screened between May 2014 and September 2019 in our tertiary epilepsy center. All enrolled patients had seizures during long‐range video electroencephalogram (EEG) recordings, and all patients (except one) underwent [18F] fluoro‐2‐deoxyglucose (FDG) positron emission tomography (PET) scans. Voxel‐based metabolic analysis and z‐distribution analysis were carried out to determine the metabolic pattern. Results Thirty‐three patients were enrolled. According to the patients’ seizure semiology, we divided the patients into four groups: focal impaired awareness seizures (FIAS, n = 17), faciobrachial dystonic seizures (FBDS)‐only (n = 6), FBDS‐plus (n = 8), and focal aware motor seizures (FAMS) (n = 2). No significant differences were found in the clinical manifestations or accessory tests except for the onset age (FIAS < FBDS‐plus) and seizure semiology. This was the first study to extensively describe the clinical manifestations and EEG of FAMS in anti‐LGI1 AE patients. In addition, we found that the patients with different semiologies all showed a wide range of abnormal metabolism, which is not limited to the temporal regions and basal ganglia, and extends far beyond our previous interpretation of FDG‐PET data. Conclusion Our results showed that FAMS can serve as a rare indicative seizure semiology of anti‐LGI1 AE and that individuals with this disease exhibited widespread functional network alterations.


| INTRODUC TI ON
At present, we have clearly realized that leucine-rich gliomainactivated 1 (LGI1) is an extracellular component of the voltagegated potassium channel complex protein, which is of great importance in bridging synapses. 1,2 Autoimmune encephalitis (AE) associated with anti-LGI1 antibodies often involves the limbic system and is characterized by symptoms of medial temporal lobe (MTL) damage (drug-resistant epilepsy, cognitive impairment, behavioral abnormalities, etc.), sleep dysfunctions and autonomic dysfunctions. 2 Notably, Irani et al. 3  Although controversy exists regarding whether FBDS are a type of seizure, recently some researchers detected generalized electroencephalogram (EEG) electrodecremental events, 4 and others found typical focal contralateral frontal waves, 5 both preceding the onset of muscle artifacts, indicating that FBDS were atypical seizures. In addition to this specific seizure semiology, the most common type of anti-LGI1 AE patient is focal seizures without awareness, resembling manifestations of MTL epilepsy. 6 Our previous study divided the seizure semiology of anti-LGI1 AE into three types: FBDS-only, epileptic seizures without FBDS, and FBDS plus epileptic seizures; this method of analysis revealed differences, such as ictal discharges, among the types. 7 Other groups tried to categorize seizure types of anti-LGI1 AE in terms of origin and consciousness 8,9 ; categories included the presence of FBDS, focal impaired awareness, focal aware motor/nonmotor seizures, autonomic, and generalized tonicclonic seizures (GTCS). However, the seizure semiology of anti-LGI1 AE has been expanding, and there is still room for improvement in the current classifications. Identifying types of unique semiology features will facilitate the early diagnosis and timely initiation of immunotherapy.
Focal aware seizures have been intensively investigated in previous studies; these seizures mostly manifest as abnormal feelings, including numbness, cold, pain, tingling, or others. [9][10][11] In 2011, Andrade et al. described abnormal tonic movements involving different body regions in three anti-LGI1 AE patients. 4 After that, several studies also reported a similar semiology and labeled them as epileptic spasms, dystonic posture, dystonic/clonic seizures, tonicdystonic seizures, or focal dystonia. 5,10,12,13,14 In fact, all of these symptoms represented the same entity as FBDS described by Irani et al. 3 However, focal aware motor seizures (FAMS), independent of FBDS, have only been mentioned in two studies to the best of our knowledge, and no detailed clinical features or EEG results have been described in detail. 8,15 [ 18 F]fluoro-2-deoxyglucose (FDG) positron emission tomography (PET) imaging plays increasingly important roles in the diagnosis, cancer screening, and follow-up of AE. Previous studies found that FDG-PET can significantly increase the sensitivity for detecting abnormalities in patients with AE compared with structural magnetic resonance imaging (MRI) or EEG. [16][17][18][19] Although considerable research efforts have been devoted to the study of AE patients' brain networks by using multimodal MRI, 20,21 the unique nature of FDG-PET in measuring synaptic activity is irreplaceable, as this data can reflect the impairment of the network from the perspective of metabolism. 22,23 Moreover, it is worthwhile to mention that computeraided brain FDG-PET analysis further greatly improved analysis sensitivity. 24 Here, we enrolled 33 anti-LGI1 AE patients in our tertiary epilepsy center, and they all had seizures during the long-range video EEG examination. According to the patients' seizure semiology, we divided the participants into four groups: focal impaired awareness seizures (FIAS), FBDS-only, FBDS-plus, and FAMS. We provided a thorough description of the clinical and EEG manifestations of the two FAMS patients. In addition, by comparing matched controls, we found distinct metabolic patterns of different groups with the help of computer-aided analysis. Since only one patient in the FAMS group underwent FDG-PET, we performed a z-distribution to evaluate the metabolic changes.

| Patient inclusion
Between May 2014 and September 2019, 57 anti-LGI1 AE patients were retrospectively screened in our tertiary epilepsy center (Beijing Tiantan Hospital). Among them, 21 patients were excluded due to incomplete PET examination or/and negative EEG results, one patient was excluded due to incomplete clinical data, and two patients were excluded due to indistinguishable forms of attacks (limb shaking, no EEG changes). Finally, the remaining 33 patients were enrolled in our study; it should be noted that although one patient had no FDG-PET imaging, his rare FIAS were captured by EEG recording. The diagnostic criteria of AE were based on a previous consensus, 25 and all patients were to have positive anti-LGI1 antibodies in their blood and/ or cerebrospinal fluid (CSF) for inclusion. Patients were evaluated by at least two experienced neurologists according to a standardized protocol; this evaluation included medical history interviews, neurologic examinations, and a battery of neuropsychologic tests. In addition, comprehensive blood and CSF tests, a 3.0 T MRI scan and a 24-h or longer EEG recording using the 10-20 system of scalp electrode placement were completed. Clinical information was obtained by reviewing the patients' charts and databases. For the subsequent analysis of FDG-PET, 31 healthy volunteers were recruited through advertisement. All healthy volunteers had no preexisting neurological or psychiatric illness.

| FDG-PET imaging acquisition and processing
Thirty-two patients underwent FDG-PET scans during initial clini- to match the number, sex ratio, and age of patients in the three different groups (FIAS, FBDS-only, and FBDS-plus). Images from the three groups of patients were tested for relative hypo/hypermetabolism by comparison with the corresponding reference database on a voxel-by-voxel basis using the general linear model by means of the SPM12 two-sample t-test design with age included as a covariate. An uncorrected threshold of p < 0.005 was used to compare paired groups. Only clusters of more than 50 voxels were considered.
For the only patient in the FAMS group, z-score mapping implemented in BrainVisa software (http://brain visa.info) was used to extract areas with differences between the patient and normal controls. Clusters of >100 voxels (8 ml) and voxels with absolute values of >1.96 z-score (p < 0.05) were considered to have significantly higher metabolism than healthy controls, and these areas were extracted before z-score maps were displayed onto anatomical images.
The details are provided in our previous literature. 24

| Statistical analysis
Analyses were performed by using SPSS 13.0 software (SPSS). For categorical variables, Fisher's exact test was used for group comparison. For quantitative variables, the Shapiro-Wilk test indicated that the data did not conform to a normal distribution; thus, we adopted a nonparametric test (Kruskal-Wallis H test) to make group comparisons. The onset age among groups was significantly different, and we made multiple post hoc comparisons by using the "all pairwise" method. Differences were considered significant at p < 0.05 (two-sided).

| Typical cases with FAMS
The two patients in the FAMS group were born at term to non- Seven seconds later, a low-amplitude fast rhythm appeared in the right central and parietal areas; then, the amplitude increased, and the frequency slowed down gradually and spread to the adjacent leads; simultaneously, a large number of motion and electromyography artifacts were detected.

| Patient 2
A 39-year-old woman presented with typical nocturnal GTCS 10 days prior to evaluation, which subsided within several minutes. sometimes combined with left-sided deviation of the head. Ten to 20 s later, the left upper limb began shaking, which abated after 10-40 s. Importantly, the patient was awake throughout the process.
Before immunotherapy, the patient underwent a brain FDG-PET scan. We calculated the z-distribution compared with age-matched healthy controls to understand the metabolic changes. As shown in Figure 2, the patient showed hypermetabolism in many scattered brain regions, indicating damage to multiple brain networks.

| Group analyses of brain glucose metabolism
In  Table S1 and Figure S1). In contrast, the brain areas with relatively low metabolism were mainly concentrated in the bi-  Figure S1).

| DISCUSS ION
This study retrospectively analyzed 33 anti-LGI1 AE patients in our tertiary epilepsy center. Seizures were observed in all patients during long-range video EEG recordings. Then, we grouped patients according to the seizure semiology and acquired patients' metabolic changes compared with the matched controls. This study was a larger sample study in which anti-LGI1 patients had clinical and subclinical seizures. Importantly, we first described the clinical and EEG manifestations of two FAMS patients in detail, and we explored the metabolic pattern of each group divided by seizure semiology. All FDG-PET scans were performed before immunotherapy due to its effect on metabolism. 18,26,27 Involuntary movements have been described in up to 40%-68.8% of patients with anti-LGI1 AE. 1,5 Despite naming differences, abnormal movements mostly belong to the FBDS entity, which is a common and characteristic manifestation of anti-LGI1 AE. 5,10,12,13,14 For example, authors reported tonic seizures when the tonic component was more prominent than the dystonic component. 4 7,19 or abnormal metabolism in entirely other brain regions. 27 These results suggest that our understanding of AE metabolism is still in its infancy, and the previous visualization method may be outdated.
Previously, we found that computer-aided semiquantitative analysis of FDG-PET scans could reduce the false negative rate of the MTL and basal ganglia up to 56% and 73%, respectively, compared with visual reading. 24 By using this method on voxelbased segmented brains, we found some surprising results that the brain areas with altered metabolism were much larger than we thought.  21 Thus, anti-LGI1 AE is not confined to the limbic system or basal ganglia but rather affects a wide range of brain regions and functional systems, and extralimbic symptoms and subclinical manifestations should be noted.
We acknowledge a number of limitations in this study and future directions that should be taken. First, the sample size was too small, and the patients were from a single center. Second, we found specific metabolic patterns of different groups, but it is not a simple superposition relationship; for example, the metabolic pattern of the FBDS-plus group was not the addition of the FIAS and FBDS-only groups; this finding is difficult to explain and may be due to individual differences, small sample sizes or different onset ages. Third, it is difficult to explain the causes of some brain regions' abnormal metabolism, and the results cannot currently be used in individual diagnosis. Fourth, although we have corrected for the number, sex ratio, and age of the patients, some patients had intracranial ischemic changes, which could cause metabolic changes. Considering the shortcomings of our research and the limitations in this field, multicenter collaboration to include more subjects is needed to verify the repeatability of the results. Furthermore, deep-level mechanisms, such as genetic mutations 37 and peripheral DNA methylation, 38 which is our future research direction, may contribute to a better understanding of our results.

| CON CLUS ION
Our results show that FAMS can serve as a rare indicative symptom of anti-LGI1 AE. Furthermore, we found that patients with different seizure semiologies all showed a wide range of abnormal metabolic patterns, which is far beyond our previous interpretation of FDG-PET data, indicating that this type of AE is associated with widespread functional network alterations. Future larger sample studies and studies from other perspectives 37,38 will contribute to the validation and interpretation of the current results.

CO N FLI C T O F I NTE R E S T
On behalf of all authors, the corresponding author confirms no conflict of interest. All authors agreed to the publication of the manuscript in its current form.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.