All the participants were diagnosed with drug-resistant epilepsy at the Beijing Institute of Functional Neurosurgery, Xuanwu Hospital, between January 2012 and December 2021. In total, this study included 65 patients. Their mean age was 24.37 ± 9.56 years and 36.9% (24 subjects) were women. The mean duration of epilepsy was 12.76 ± 7.28 years. The mean follow-up duration was 39.4 ± 20.9 months (ranging from 12 to 108 months). Experienced neurologists and neurosurgeons discussed and designed the selection of the stimulation target and the surgical plan of the DBS for each patient.

The inclusion criteria were as follows: (1) Patients diagnosed with drug-refractory epilepsy were based on the ILAE Classification of Epilepsies (19). (2) The preoperative evaluation indicated that the patient was inoperable or had contraindications for resective surgery, such as widely distributed epileptogenic zones (EZ), EZ located in the functional cortex, failed resective surgery, or the patient refused to undergo the resective surgery.

We performed long-term scalp video electroencephalography (VEEG) to record at least three habitual seizures for each patient using a video EEG monitoring system (Micromed, Treviso, Italy). In some patients, we identified the EZ by performing stereotactic electroencephalography (SEEG). All patients underwent a high-resolution magnetic resonance imaging (MRI) protocol performed using a 3.0-T MR Scanner (Siemens, Verio, Germany) and consisting of conventional axial, sagittal, and coronal T1-weighted spin-echo sequences. In some patients, we identified the EZ by performing magnetoencephalography and positron emission tomography-computed tomography. The patients who underwent the DBS procedure after the special committee consultation excluded resective surgery based on their clinical data. For each patient, we selected the target thalamic nucleus for DBS (ANT, STN, CMN, or PN) based on the patient's epilepsy or seizure type and the location of the epileptogenic focus, as well as the possible epileptic network involved (20, 21). We defined the baseline for each patient as their mean seizure frequency over the 3-month pre-implant period.

We implanted the DBS electrodes (Model 3387 or 3389; Medtronic, Inc., Minneapolis, MN, USA) with the assistance of a frame-based, microelectrode-guided, stereotactic technique under general anesthesia. All the patients receiving ANT-DBS, CMN-DBS, and PN-DBS underwent bilateral electrode implantation. In patients with STN-DBS, some patients with specific epilepsy types (such as those with the possible EZ located in the unilateral hemisphere) underwent unilateral electrode implantation. With the help of the high-resolution T1-weighted images, we delineated the thalamus nuclei based on the Morel Stereotactic Atlas. We performed the surgical procedure of the implantation of the DBS leads and the pulse generator (Model 3628 screener, Medtronic, Inc., Minneapolis, MN, USA) based on previous studies (22). Postoperative computed tomography was performed and registered with the T1-weighted images to confirm the locations of the electrodes.

One month after the implantation procedure, the pulse generator was initiated to be activated and programmed. The outpatient review of each patient was carried out 3 months after the operation to identify the occurrence of long-term complications. In addition, the stimulation parameters and contacts were adjusted based on seizure frequency and clinical response. We categorized patients with a ≥50% decrease in seizure frequency (mean seizure frequency for the last 3 months of follow-up, compared with the baseline as responders and patients with a < 50% decrease in seizure frequency as non-responders. All of the patients were followed up monthly or trimonthly, and the seizure frequency was reported by the patient or the family members. Noteworthy, the post-operative seizure needs to be verified with the habitual seizure. The data were recorded from outpatient reviews, medical record reviews, patients' daily diaries, and telephone interviews. The postoperative program control details for each patient were also documented.

The ANT is the most common stimulation target of DBS in epilepsy treatment (23). The unique anatomical relationship and the functional connection between ANT and the limbic system make this nucleus an ideal stimulation target for TLE treatment (24, 25). In our study, 80% of patients with ANT-DBS had TLE (including T-plus), and our results are in line with previous studies (13). Our previous SEEG study demonstrated that the ANT-DBS would desynchronize the epileptic network in patients with TLE. In addition, the position-specific correlation had also been reported between the DBS applied to the ANT and patients with TLE and EZ within the Papaz circuit or limbic system (26). Moreover, the ANT can receive the interictal period discharges that propagate from the epileptogenic zones in neocortical temporal and mesial temporal epilepsy (27). Therefore, combined with the previous clinical studies, our data suggest that ANT is an optimal stimulation target for patients with TLE. Fasano et al. suggested that patients with frontal seizures also benefit from ANT-DBS (28). The long-term follow-up of the SANTE trial showed that frontal onset seizures also respond well to ANT-DBS (14). In our study, patients with frontal epilepsy showed a good response to ANT-DBS. In addition, patients with multifocal epilepsy also benefited from the ANT-DBS. Previous studies suggested that ANT plays a role in a wider cortical network (29, 30), and other epilepsy types could be treated through ANT stimulation.

In patients with FBTCS, ANT-DBS showed good efficacy, which may be because ANT-DBS modulates the epileptic network excitability. Previous studies suggested that the ANT participates in the organization and maintenance of seizure activity (21). In addition, Tyvaert et al. observed a synchronous activity between the ANT and generalized epileptogenic network in patients with generalized epilepsy, indicating that the ANT is a potential propagation point (31). We speculate that, on the one hand, ANT-DBS might reduce the epileptic network excitability to some extent and raise the seizure threshold, making the seizure less likely to occur. On the other hand, the reduced network excitability might limit the propagation of the epileptic excitatory signal. This hypothesis is also supported by the reduced severity of postoperative seizures in patients with FBTCS. Therefore, based on the mechanism studies and the clinical results, ANT-DBS is also an alternative treatment for patients with FBTCS.

The reasons why some patients have poor responses to ANT-DBS are complex. In our study, ANT-DBS turned out to be poorly effective for patients with bilateral HS. We speculate that the EZ of these patients has excessive excitability, and ANT-DBS may have a relatively weak inhibitory effect on the sclerotic hippocampus. A previous epilepsy study reported that hippocampal DBS was less effective in patients with HS than in patients with normal MRI profiles (32). According to previous studies, the sclerotic hippocampus was related to neuronal reduction, which may prevent the hippocampus to provide enough available tissue for modulation (33). In addition, the sclerotic hippocampus might have an increased impedance and require a more intense stimulus (32). Regarding ANT-DBS in patients with bilateral HS, indirect stimulation based on the specific network may further weaken the regulation effect on the sclerotic hippocampus. Noteworthy, the aberrant circuits may be involved in patients with bilateral HS, which would induce inefficacy or even the paradoxical effect when the ANT-DBS was applied.

Currently, identifying patients who would benefit from ANT-DBS is difficult. Our results indicate that some epilepsy types would be refractory to this treatment. Therefore, different stimulation targets and corresponding surgical indications need to be explored further for patients with refractory epilepsy.

In some patients with drug-resistant epilepsy, resective surgery would be contraindicated due to the EZ being located in the primary motor cortex. Responsive brain stimulation (34) and ANT-DBS (13) offer an alternative treatment for these patients, but the response is not always satisfactory. In 2002, Benabid et al. first applied STN-DBS to treat epilepsy in a patient with focal centroparietal dysplasia and reported an 80.7% reduction in seizure frequency (16). Subsequent studies confirmed the safety and effectiveness of STN-DBS in patients with motor seizures (35, 36). Regarding the mechanism, our team's prior study demonstrated the interaction between STN and the motor cortex. In addition, STN-DBS with high-frequency stimulation suppressed the interictal spikes and high-frequency oscillations in patients with motor seizures (37).

Based on existing clinical evidence and our knowledge of the mechanism, we choose STN as the target for patients with an EZ overlapping the sensorimotor cortex. STN-DBS significantly reduced motor seizures in these patients in concordance with previous studies (38). Therefore, STN-DBS can be a potent treatment option for patients with motor seizures. Nevertheless, STN-DBS needs to be further investigated in large-scale randomized controlled trials and specific regulative mechanism studies. Notably, patient 57 reported a seizure frequency increase (four times per month) after the STN-DBS, which might be related to the lower baseline (2–3 times per month) and require further stimulation parameter adjustment and follow-up. The EZ of patient 50 was located in the frontoparietal region with focal motor seizures and FBTCS. After STN-DBS, his focal seizures disappeared, and the FBTCS frequency decreased non-significantly; the pulse generator was removed after 14 months due to an increase in FBTCS frequency. We speculated that the poor response may be due to improper stimulation parameters and contacts. Therefore, even when selecting the optimal stimulation target, elaborate postoperative program control is particularly important.

Patients with LGS present specific EEG abnormalities and multiple seizure types, such as generalized tonic seizures (17). Previous studies recorded epileptiform EEG activity in the CMN of patients with generalized tonic seizures from LGS (39). In addition, CMN has diffuse connections with the diffuse frontal areas, brainstem, and striatum, which prompted us to choose the CMN rather than the ANT, as the stimulation target in LGS or LGS-like epilepsy cases (40). Velasco et al. performed CMN-DBS on five patients with drug-resistant epilepsy and reported a significant reduction in secondary generalized tonic-clonic seizures (GTCS) frequency (41). Subsequent randomized controlled and small open-label studies reported significant efficacy for CMN-DBS in generalized seizures, especially in patients with primary or secondary LGS (17, 42). Based on this encouraging clinical data, we performed CMN-DBS in three patients with LGS-like epilepsy and also observed a good response. Therefore, we consider CMN-DBS to be an alternative treatment for patients with generalized-onset epilepsy.

As the largest thalamus nucleus, the PN has extensive connections with areas of the cortex, such as the mesial temporal lobe, the parietal cortex, and the occipital lobe (43–45). The anatomical features of the PN indicate that it is a potential neuromodulation target to treat epilepsy. Compared with the other targets, clinical reports on the stimulation of PN for the treatment of epilepsy are relatively rare. Filipescu et al. investigated PN stimulation on temporal lobe seizures and first suggested that PN-DBS could be a well-tolerated and effective approach for drug-resistant epilepsy (44). In a study of responsive neurostimulation targeting the PN to treat epilepsy, it was effective for drug-resistant epilepsy with posterior quadrant origin (18). In our study, we performed PN-DBS on one patient with bilateral occipital lobe epilepsy, significantly reducing seizure frequency. Although this is only one case, PN does seem to be an alternative target for neuromodulation to treat occipital lobe epilepsy.

Based on our single central clinical results, we summarized empirical guidance for the selection of stimulation targets for patients with refractory epilepsy. Our results show that ANT-DBS is effective for patients with either TLE (including T-plus) or ETLE (including FLE and multifocal epilepsy). However, in patients with bilateral HS, ANT-DBS should be applied with more caution. In addition, ANT-DBS is effective for patients with FBTCS. For patients with motor seizures, especially with the EZ overlapping the sensorimotor cortex, STN-DBS might be a powerful treatment. CMN-DBS and PN-DBS might be alternative options for patients with LGS-like epilepsy and occipital lobe epilepsy, respectively.

The small cohort of our study prevented the investigation of the efficacy of DBS in a wider variety of epilepsy and seizure types. In addition, we only reviewed the efficacy of DBS in different thalamus nuclei. In future studies, we would investigate the details of postoperative program control, such as the parameter settings and side effects and the influence factors of surgical outcome.