2.1 Patient Selection

The study was approved by the Ethics Committee of the Sanbo Brain Hospital, Capital Medical University. Nine children were diagnosed with SeLECTS with ESES at the Comprehensive Epilepsy Center, Sanbo Brain Hospital from August 2015 to July 2022. The median age was 7 (5.0, 8.5) years, including five males (56%) and four females (44%). The inclusion criteria were as follows: (i) age of onset between 3 and 15 years; (ii) clear diagnosis of self-limited epilepsy with centrotemporal spikes; (iii) SWI ≥ 50%; (iv) refractory epilepsy or epileptic encephalopathy. Patients with a history of prematurity (< 35 weeks), abnormal magnetic resonance imaging (MRI) of the brain, other epilepsy syndromes, neurosurgery, or severe brain injury were excluded. All the patients received follow-up for at least 6 months after TMS.

2.2 Repetitive Transcranial Magnetic Stimulation

In the early stages (2015–2019), the localization of the treatment was determined based on the clinical seizure presentation and EEG discharge distribution. The EEG international 10–10 system was used to localize C5/C6 (central facial movement representative area) (Figure 1A). After 2020, the localization was determined with TMS-specific navigation software for three-dimensional reconstruction of MRI images (Figure 2). This involved comparing image data from PET-CT and MRI fusion with navigation data to localize hypometabolic or hypermetabolic zones adjacent to the facial representation of the central region. Low-frequency stimulation (0.3, 0.5, and 1 Hz) by Rapid 2 (Magstim) was utilized based on the frequency of seizures and the patient's cooperation at the beginning of treatment. Eight patients received 10 workdays of stimulation with 1000–1500 pulses administered per day. The cTBS treatment modality by YD-MT600 (Neurosoft) was used in one patient. Stimulation intensity for rTMS treatment was set at 40%–60% according to our previous clinical experience and the patient's tolerance. The ASMs remained constant in all patients during the rTMS treatment without the other therapeutic interventions (Table 1).

2.3 EEG Recordings

EEG recordings were conducted using the standard international 10–20 system, employing a total of 19 scalp electrodes. The Nicolet recording system was used for EEG recordings during both sleep and wakefulness. SWI was reviewed by an independent electrophysiologist who guaranteed the confidentiality of clinical data. The sleep spindle was analyzed during the first 10 min of stage II sleep in the first sleep cycle. The sleep spindle, with a frequency of 10–15 Hz, was counted in the frontal, central, and parietal lobe using A1/A2 reference leads. Instances of synchronization between both hemispheres were counted once, while asynchronous occurrences with a time difference exceeding 1 s were counted twice, with the longest time course recorded. Manual counting was adopted due to significant discrepancies between initial trial software calculations and manual counts. Two senior electrophysiology experts independently conducted the manual counting, with discrepancies less than 3 being accepted without re-counting. In cases of inconsistent results with discrepancies exceeding 3, additional counts and discussion were undertaken, and results were confirmed after reaching consensus (Figures 1B and 3).

2.4 Slow Wave-Sleep Spindle Coupling

A machine learning algorithm utilizing YASA was employed. The slow wave screening frequency was set to 0.5–4 Hz, with a down duration range of 0.18–1.5 s and an up duration range of 0.06–1 s. The sleep spindle screening frequency was set at 10–15 Hz, lasting 0.5–4 s. The coupling identification time window was set as 1 s before and after the crest of the down portion of the slow wave. Slow waves and the sleep spindle were monitored during sleep phase II, with the coupled sleep spindles being sought by the YASA package among the identified slow waves. The coupled entries were conditioned, and the final number of results was obtained [16] (Figure 1C).

2.5 Neuropsychological Evaluation

All children underwent assessment using standard evaluation procedures. Intelligence was assessed before and after treatment for all patients, while memory was evaluated in children aged over 7 years. The Wechsler Intelligence Scale for Children, fourth edition (WISC-IV), was utilized to measure the intelligence level of the children, while the Wechsler Memory Scale was employed to evaluate the memory level.

2.6 Follow-Up Program

Seizure frequency, SWI, sleep spindles, and WISC-IV were assessed in nine patients, and Wechsler Memory Scale was assessed in five patients. Seizure frequency, SWI, and the sleep spindle were assessed at 3 months and 6 months after treatment. Wechsler Memory Scale, WISC-IV, and a slow wave and the sleep spindle coupling were assessed at 6 months after treatment.

2.7 Statistical Analysis

All data were tested for normality. Normally distributed data were expressed as X ± S, while non-normally distributed data were expressed as M (P25, P75), and counts were expressed as rate (%). Because of the small sample size, normality was determined by Shapiro–Wilk, and the values of sleep spindle and MQ in all groups conformed to normal distribution, and a Paired-Samples t test was used. Values for seizure frequency, SWI, and IQ that did not exhibit a normal distribution were analyzed via Wilcoxon signed-rank test, a nonparametric test. Correlation analysis was performed using the Spearman correlation coefficient. Comparative data for seizure frequency, SWI, and sleep spindle in pre-treatment, 3-month post-treatment, and 6 months post-treatment were subjected to Benjamini-Hochberg (FDR) validation. All test p-values less than 0.05 were considered statistically significant.

3.1 Clinical Information

Nine patients (four females and five males) with SeLECTS with ESES were enrolled. Eight of the patients had refractory epilepsy, and one had epileptic encephalopathy. Median age was 7 (5.0, 8.5) years. MRI results were normal for all patients. Seven patients presented salivation during non-seizure episodes, including one case with opercular syndrome. Before rTMS, the median number of seizures per month was 5 (1.13, 26.25), median SWI was 81% (72.5%, 89.5%) and mean number of the sleep spindle was 54.56 ± 43.97. Detailed information is provided in Table 2.

3.2 Seizure Frequency Before and After rTMS

All patients completed rTMS treatment. The overall number of seizures per month significantly decreased at 3 months and 6 months after rTMS (Table 3). Six out of the eight patients (75%) were seizure-free. Notably, these six patients were also seizure-free before treatment, and the data of patient 2 were not included in the seizure frequency analysis. One patient experienced a 60% reduction at 3 months and a 40% reduction at 6 months in seizure frequency. The other experienced a 47% reduction in seizure frequency at 3 months (p = 0.024) and returned to baseline at 6 months after rTMS (p = 0.018) (Figure 4A).

3.3 SWI Before and After rTMS

All the patients had SWI ≥ 50%. The median SWI after rTMS decreased from 81% (baseline) to 68% at 3 months and then to 57% at 6 months. The SWI at 3 months was significantly reduced by rTMS (p = 0.045). The SWI at 6 months significantly decreased in six patients, mildly increased in two patients, and remained unchanged in one patient after rTMS (p = 0.035) (Figure 4B).

3.4 Sleep Spindle Before and After rTMS

There was a significant increase in the sleep spindle of all patients at 3 months after rTMS (p = 0.002) and 6 months (p = 0.002); the mean number of the sleep spindle increased from 55 to 91 at 3 months and further increased to 147 at 6 months after rTMS (Figure 4C). The changes in the sleep spindle at 6 months after rTMS were observed in five aspects (Table 4): (i) All patients showed an increased number of sleep spindles; (ii) Longest single duration increased in seven patients (78%), decreased in one patient (11%) and was unchanged in one patient (11%), with the mean duration increasing from 1.87 s to 2.84 s; (iii) In the frequency range of 10–12.5 (Hz), the number of the sleep spindle increased in all patients; and the mean number of the sleep spindle increased from 33 to 79 before and after rTMS. In the frequency range of 12.6–15 (Hz), the number of the sleep spindle increased in eight patients (89%), and the mean number of the sleep spindle increased from 22 to 69; (iv) Seven patients (78%) showed an increase while two patients (22%) showed a decrease in maximum amplitude, and the mean maximum amplitude increased from 89 μV to 101 μV; (v) Seven patients (78%) showed improvement in amplitude modulation while two patients (22%) showed no alteration.

3.5 Neuropsychological Evaluation

All patients underwent IQ assessment, and five patients more than 7 years old also underwent memory assessment before rTMS and at 6 months after rTMS. All IQ and MQ scores were below the normal range before treatment. After rTMS, the median IQ increased from 72 to 83. The IQ of nine patients increased (p = 0.012), and the MQ of the five patients increased after stimulation (p = 0.016) (Figure 5). Improvement in IQ and increase in the sleep spindle were significantly positively correlated (p = 0.035).

3.6 Slow Wave-Sleep Spindle Coupling Before and After rTMS

The number of slow wave-sleep spindle coupling before and after rTMS was recorded for all the patients. The probability of spindle coupling in the slow wave “up” state was analyzed. The mean probability of the sleep spindle coupling in the slow wave “up” state increased from 28% to 55% at 6 months after rTMS (Figure 6).

3.7 Improvement in Other Clinical Symptoms

Seven patients presented salivation during non-seizure episodes before rTMS. Salivation symptoms during non-epileptic episodes improved in five patients (disappeared in 4 and improved in 1), while the other two patients showed no alterations after rTMS. The symptoms of opercular syndrome in one patient showed significant alleviation. The symptoms of significant memory loss, irritability, hyperactivity, and inattention in another patient improved. No adverse effects were reported during or after rTMS.

4.1 rTMS Inhibits Seizures, Decreases SWI and Improves Cognition

SeLECTS is a well-defined electroclinical syndrome characterized by focal seizures and typical EEG abnormalities [17]. 4.6%–5.8% of such patients will experience atypical evolution during the course of the disease, including ESES, Landau–Kleffner syndrome (LKS), motor dysfunction, refractory epilepsy, etc. [18]. Rolandic double/multiple spikes and slow-wave abnormalities in the interictal period may indicate ESES in SeLECTS [19]. Both atypical and typical ESES exhibit cognitive and executive function impairment [20]. Most of those with SeLECTS and ESES are refractory epilepsy patients with limited options to reduce seizures and improve cognitive-behavioral function. Currently, the optimal treatment regimen has not yet been established.

TMS utilizes electromagnetic coils to excite or inhibit neurons, and low-frequency repetitive pulses produce an inhibitory effect that reduces cortical excitability [21]. Extensive studies have demonstrated that rTMS yields significant clinical improvements in patients with various neurological and psychiatric disorders [22]. rTMS in both the ictal and interictal states, over the seizure focus in patients with focal epilepsy, could abort seizures and reduce seizure frequency or severity [23]. Low-frequency rTMS is effective and safe in the treatment of drug-resistant epilepsy [24]. Our previous study indicated that 62.5% of patients with SeLECTS with ESES got seizure-free and SWI significantly decreased within 3 months after stimulation [10]. In the present study, 75% of patients with SeLECTS with ESES achieved seizure-free. In addition, a significant reduction of SWI was found from 81% to 68% and 57% at 3 months and 6 months after rTMS respectively. It is important to note that rTMS alleviated symptoms of persistent salivation, opercular syndrome, and epileptic encephalopathy in patients with SeLECTS with ESES. It is well known that differences in the parameters of rTMS lead to differences in efficacy. High-frequency (> 5 Hz) rTMS is thought to induce excitatory effects on the cortex, while low-frequency (< 1 Hz) stimulation is considered to induce inhibitory effects. Whereas cTBS generally reduces cortical excitability [25], cTBS and low-frequency rTMS are thought to cause MEP amplitude reduction [26]. Low-frequency stimulation of rTMS was used in 8 of the 9 patients in this study. One patient with poor cognition could not cooperate with the comparatively longer-term treatment, so the patient accepted cTBS therapy. Both therapies have similar mechanisms of action and efficacy and could not impact the reliability of the results.

In the present study, all patients with SeLECTS with ESES had cognitive impairment indicated by a decrease in IQ. rTMS demonstrated various degrees of cognition improvement.

4.2 rTMS Increases Sleep Spindles and the Sleep Spindle Coupling in the Slow Wave Up-State

The sleep spindle is generated in the reticular nucleus of the thalamus, comprising a diffuse network of GABAergic neurons that encase other thalamic nuclei [27]. The sleep spindle appears as 10–15 Hz sinusoidal cycles and serves as a marker of the N2 stage of non-REM sleep on EEG. The sleep spindle reflects the strength and malleability of thalamocortical circuits, and its density correlates with markers of intelligence, as well as various cognitive deficits and dementia-related disorders [28]. Abnormal electrophysiology associated with sleep suggests disturbances in thalamocortical circuits, wherein dysfunction may hinder the generation of the sleep spindle, which is a brain rhythm crucial for memory processes [29]. Memory restoration occurs during the sleep spindle events and is subsequently reprocessed during spindle refractory periods [30]. Interventions that increase spindle activity have been shown to enhance sleep-dependent memory consolidation [31-33]. Meta-analyses revealed a positive relationship between the sleep spindle and overall cognition. A decrease in the sleep spindle is a predictive biomarker for cognitive impairment [34]. Our data showed a significant increase in the sleep spindle of all patients after rTMS. The mean number of sleep spindles increased from 55 to 91 and 147 at 3 months and 6 months, respectively, after rTMS. rTMS modulates the sleep spindle in duration, frequency, amplitude, and amplitude modulation in the majority of cases. Of most importance, improvement in IQ and increase in the sleep spindle were significantly positively correlated in the present study. By analyzing the alteration of the sleep spindle and associated neuropsychological performance after treatment, we postulated that an increase in the sleep spindle might be a potential mechanism for cognitive improvement in patients with SeLECTS with ESES.

Furthermore, we explored the role of rTMS on the coupling of the sleep spindle and slow waves. Slow wave-sleep spindle activation in the hippocampus is indicated to consolidate memories during sleep [35]. During brain maturation from childhood to adolescence, the precision of slow oscillation and the sleep spindle coupling increases, contributing notably to enhanced consolidation in declarative memory tasks, recall, and sleep dependence. This phenomenon may serve as an indicator of the development of the hippocampal–cortical memory system [36]. Additionally, functional coupling between hippocampal ripples and thalamocortical sleep spikes and slow waves during NREM sleep facilitates fine-tuned communication between the hippocampus and neocortex, thereby benefiting memory consolidation [37]. The timing of precise coordination between the NREM slow oscillation “up” state and the sleep spindle predicts the success of hippocampus-dependent memory consolidation [38]. The slow-oscillation “up” state is necessary for the replay and possible consolidation of previously learned tasks or experiences [39]. Our results revealed that the mean probability of the sleep spindle coupling in the slow wave rising “up” state increased from 28% to 55% at 6 months after rTMS. The sleep spindle coupling in the slow wave “up” state might be a potential mechanism for cognition improvement in patients with SeLECTS with ESES.