Pathogenesis, diagnosis, and treatment of electromagnetic stimulation–mediated neuromodulation therapy and new technologies

Authors: Dian Jiao, Lai Xu, Zhen Gu, Hua Yan, Dingding Shen, Xiaosong Gu

Abstract

Epilepsy is a severe, relapsing, and multifactorial neurological disorder. Studies regarding the accurate diagnosis, prognosis, and in-depth pathogenesis are crucial for the precise and effective treatment of epilepsy. The pathogenesis of epilepsy is complex and involves alterations in variables such as gene expression, protein expression, ion channel activity, energy metabolites, and gut microbiota composition. Satisfactory results are lacking for conventional treatments for epilepsy. Surgical resection of lesions, drug therapy, and non-drug interventions are mainly used in clinical practice to treat pain associated with epilepsy. Non-pharmacological treatments, such as a ketogenic diet, gene therapy for nerve regeneration, and neural regulation, are currently areas of research focus. This review provides a comprehensive overview of the pathogenesis, diagnostic methods, and treatments of epilepsy. It also elaborates on the theoretical basis, treatment modes, and effects of invasive nerve stimulation in neurotherapy, including percutaneous vagus nerve stimulation, deep brain electrical stimulation, repetitive nerve electrical stimulation, in addition to non-invasive transcranial magnetic stimulation and transcranial direct current stimulation. Numerous studies have shown that electromagnetic stimulation-mediated neuromodulation therapy can markedly improve neurological function and reduce the frequency of epileptic seizures. Additionally, many new technologies for the diagnosis and treatment of epilepsy are being explored. However, current research is mainly focused on analyzing patients’ clinical manifestations and exploring relevant diagnostic and treatment methods to study the pathogenesis at a molecular level, which has led to a lack of consensus regarding the mechanisms related to the disease.

Introduction

Epilepsy is a severe neurological condition characterized by repeated seizures (Yu et al., 2024). Recurrent epilepsy attacks often lead to symptoms such as depression, anxiety, and insomnia, which greatly impede the patient’s ability to perform daily activities (Keezer et al., 2016; Devinsky et al., 2018). The mechanism behind epilepsy involves abnormal protein expression and an imbalance of neurotransmitter levels, although the precise details are still not fully understood. The diagnosis of epilepsy typically relies on reviewing the patient’s medical history and conducting electrophysiological tests, such as electroencephalography (EEG), which monitors and records the brain’s electrical activity to detect asynchronous and abnormal firing patterns of neurons. Additional imaging tests like magnetic resonance imaging or computed tomography scans may be performed to identify any structural issues that could be causing epilepsy. Currently, the conventional treatments for epilepsy include the use of antiepileptic drugs and surgery (Spencer, 2002; Duncan et al., 2016). Currently, neuromodulation therapies, including vagal nerve stimulation (VNS) and deep brain electrical stimulation (DBS), have garnered approval for treating intractable epilepsy. These advanced techniques effectively regulate nerve activity through precision delivery of electrical stimulation to targeted regions of the brain (Cagnan et al., 2019; Ryvlin et al., 2021). Recent research has further elucidated the profound potential of electromagnetic stimulation-mediated neuromodulation as an emerging therapeutic avenue. By adroitly administering high-frequency electrical stimulation to the anterior thalamic nucleus, this therapy achieves desynchronization of epileptic networks, thereby markedly mitigating seizure frequency and limiting the extent of epileptic diffusion. This groundbreaking approach offers renewed hope for patients battling this challenging neurological condition (Yu et al., 2018). This review aims to provide readers with an up-to-date, scientifically accurate, and comprehensive understanding of the progress and challenges associated with electromagnetic stimulation treatment in the field of epilepsy. It discusses the previously reported underlying mechanisms of epilepsy, analyzes existing EEG monitoring and diagnostic data, and provides an overview of clinical interventions and treatments. By incorporating the latest research data and considering the realities of clinical practice, this review ensures that readers are equipped with valuable information.

Search Strategy

In this narrative review, we have included research on the pathogenesis, monitoring, diagnosis, and treatment of epilepsy. We conducted a literature search using the PubMed database, using search terms such as epilepsy, pathogenesis, diagnosis, treatment, and neural regulation. Articles were selected based on their titles and contents, with a focus on those published within the last 10 years.

The Pathogenesis of Epilepsy

The underlying mechanisms involved in epilepsy have not been defined or unified due to their complexity and controversy. Recent research has established a link between abnormal neuronal discharge and this disease, including aberrant microRNA (miRNA) expression, dysregulation of genes and proteins, and abnormal ion homeostasis and ion channel function (Song et al., 2011; Henshall, 2014; Piletič and Kunej, 2016; Ren et al., 2016; Zhan et al., 2016; Organista-Juárez et al., 2019; Almeida Silva et al., 2020; Guerra Leal et al., 2022; Rastegar-Moghaddam et al., 2022; Li et al., 2023; Table 1).

Abnormal miRNA expression

Several miRNAs associated with epilepsy have been identified. Wang et al. (2017) discovered that a reduction in miR-219 expression in mouse oligodendrocytes led to a myelin deficiency accompanied by tremors, which progressed to severe epilepsy, ataxia, and death 4 months after the onset of seizures. In a mouse model of epilepsy, Venø et al. (2020) observed anti-epileptic phenotypes following miR-10a-5p, miR-21a-5p, or miR-142a-5p knockout, and simultaneous inhibition of these three miRNAs reduced the occurrence of spontaneous epileptic seizures. Several studies have found that decreased miR-128 expression in dopaminergic neurons of postnatal mice results in severe motor functional impairment and even fatal epilepsy. Using microelectrode arrays, researchers found that inhibiting miR-128 significantly increased neuronal activity in the primary mouse motor cortex (Tan et al., 2013; Henshall et al., 2016; McSweeney et al., 2016). miR-134 is implicated in epilepsy; it promotes actin motility by targeting LIM kinase to double-decompose cofilin phosphorylation, thereby regulating the morphology of dendritic spines. The upregulation of miR-134 induces long-term nociceptive seizures. Based on these findings, Ant-134 was found to inhibit miR-134 and regulate synaptic microstructures, thereby impairing the capacity of neural networks to transmit super-synchronous discharges and producing practical and durable therapeutic effects on epilepsy (Jimenez-Mateos et al., 2012; Brennan and Henshall, 2020; Reschke et al., 2021). Another study developed a dTg-211 mouse model by inhibiting the ectopic expression of miR-211 in the forebrain with doxycycline, which was accompanied by convulsive seizures, increased synaptic activity, Ca^2+^ transmembrane transport, transforming growth factor beta receptor II signaling, and synaptic pathway signaling (Bekenstein et al., 2017). In contrast, a reduction in miR-211 induced hyper-synchronization and non-convulsive or convulsive seizures, indicating that the dynamic expression of miR-211 could cause epilepsy. Furthermore, Kim et al. (2021) suggested that miR-106b family members induce the expression of potassium voltage-gated channel subfamily Q member 2 (KCNQ2) by binding to the 3′-untranslated region of KCNQ2 messenger RNA, and further proposed that miR-106b-5p can increase the excitability of hippocampal neurons in epilepsy through KCNQ2, thus leading to epileptic seizures.

For ease of viewing, the different possible epileptic miRNAs are summarized in Table 1 and Figure 1, which shows how the related pathways can cause epilepsy.

Dysregulation of epilepsy-related genes and proteins

With the development of genomics technologies, genes related to epilepsy have been increasingly explored, and more and more genes associated with the pathogenesis of epilepsy have been discovered. In recent years, many studies have reported on proteins and genes that may cause epilepsy, showing the possible effects of different protein or genetic changes followed by secondary epilepsy (Kovacevic et al., 2018; Uddin et al., 2018; Miyamoto et al., 2019; De Franco et al., 2020; Lammertse et al., 2020; Parras et al., 2020; Perenthaler et al., 2020; Table 2).

Ion homeostasis/ion channel dysfunction

Ion channels are the foundation of neuronal electrical activity; their dysfunction can trigger epilepsy by initiating abnormal activity in the central nervous system. Previous research has demonstrated that cyclic nucleotide and voltage-gated ion channels are involved in epileptic seizures (Ramentol et al., 2020; Shen et al., 2023).

Cyclic nucleotide-gated channels

The hyperpolarization-activated cyclic nucleotide-gated channels (hydrocyanic acid channels; HCN) are strongly associated with epilepsy (Ramentol et al., 2020). The HCN1 isoform is closely associated with epilepsy, and the HCN1 M305L variant has been detected in developmental and epileptic encephalopathy patients (Bleakley et al., 2021). Hsieh et al. (2020) noted that the neurons of focal cortical malformation mice abnormally express HCN4, increase intracellular cAMP concentration, influence HCN4-gated channels, cause focal cortical malformation neurons to discharge repeatedly, and ultimately induce epilepsy. Epileptogenesis could be prevented by inhibiting HCN4 activity in focal cortical malformation neurons, indicating that the ion channel protein HCN4 plays a crucial role in epileptic seizures (Hsieh et al., 2020).

Voltage-gated sodium ion channels

Nav1.1, a voltage-gated sodium channel subunit, is required to regulate interneuron excitability. Deficiencies and mutations in Nav1.1 can inhibit interneuron excitability. Mice lacking Nav1.1, an essential component of the E1 ligase for neddylation, exhibited spontaneous convulsions in their parvalbumin interneurons and died prematurely. It has been reported that Nav1.1 decreases parvalbumin interneuron excitability and gamma-aminobutyric acid (GABA) release while increasing the excitability of the pyramidal neuron network (Mareš and Kubová, 2020).

Mutations in the sodium voltage-gated channel alpha subunit 2 (Scn2a) gene increase neuronal excitability, which is a risk factor for epilepsy in infants and young children (Ben-Shalom et al., 2017). Another study has shown that the SCN2A variants R1882Q and R853Q can cause developmental and epileptic encephalopathy; the R1882Q variant boosts neuronal activity and the R853Q variant reduces action potential firing (Berecki et al., 2018). Severe epilepsy and malformations of the cortex are caused by the abnormal expression of the sodium voltage-gated channel alpha subunit 3 (SCN3A) subunit encoding the Nav1.3 channel in children. When combining information from 22 patients with pathogenic SCN3A variants and whole-cell voltage-clamp electrophysiological recordings, we found that SCN3A mutations caused intractable epilepsy, which was sometimes accompanied by cortical dysplasia. Domain II to IV transmembrane segments were clustered with pathogenic variants. As evidenced by a voltage-dependent persistent current increase or activation, pathogenic variants exhibited an increase in channel function (Zaman et al., 2020). Using dCas9-mediated sodium voltage-gated channel alpha subunit 1 (Scn1a) gene activation, Colasante et al. (2020) restored Scn1a haploinsufficiency in a mouse model of Dravet syndrome and rescued the physiological levels of Scn1a. This single guide RNA increased Scn1a gene expression in cell lines and primary neurons with high specificity, thereby elevating the level of Nav1.1 protein and enhancing the ability of wild-type (WT) immature GABAergic interneurons to stimulate action potentials. The enhancement of Scn1a transcription was achieved in mature Dravet syndrome model interneurons, thereby restoring their divergent capacity (Colasante et al., 2020).

Voltage-gated potassium ion channels

The potassium two-pore domain channel subfamily K member 4 (KCNK4) is a member of the two-pore domain (K2P) KCNK4/two-p-domain in a weakly inwardly ratifying K^+^ channel (TWIK)-related K^+^ channel (TREK) subfamily of mechanosensitive ion channels. De novo missense mutations in KCNK4 can induce pathological responses, including epilepsy (Bauer et al., 2018, 2019). KCNQ2–KCNQ5 of the KCNQ family of K7 voltage-gated potassium channels regulate neuronal excitability; their dysfunction correlates with neurological disorders, including epilepsy. In contrast, closed-state inactivation, or channel inactivation during subthreshold depolarization, is a dynamic state that controls neuronal excitability and potential action backpropagation in response to excitatory synaptic input. Ca^2+^ enters neuronal dendrites and mediates spike-timing-dependent plasticity, resulting in neurological pathology. Lin et al. (2018) suggested a connection between the Kv4.2 channel CSI and infantile epilepsy and autism.

Transmission of R-type calcium currents is initiated by conduction through the voltage-gated Ca2.3 channel. CACNA1E, which encodes a subunit of the Ca2.3 channel, can be mutated to cause developmental and epileptic encephalopathy by increasing R-type calcium currents (Helbig et al., 2018). Cacna1h/Cav3.2, a T-type calcium channel associated with autism and epilepsy, is inhibited in the parvalbumin interneurons of the medial prefrontal cortex (Shen et al., 2021).

Neurotransmitter GABA receptors

GABA receptors are G protein-coupled receptors of the class-C family. These receptors are important inhibitory neurotransmitters in the central nervous system, consisting of both phasic and tonic inhibition forms. The former is controlled by GABA release-mediated activation of the ionotropic GABAA receptor in post and perisynaptic membranes and the metabolic receptor GABAB. Simultaneously, the latter is mediated by ambient GABA that diffuses in the extracellular space (Stell and Mody, 2002). GABA receptors have been reported to mediate neuronal signaling in the developing brain and most of the rapid synaptic inhibition in the brains of mature animals (Shen et al., 2017; Shaye et al., 2021; Bhat et al., 2023; Bonalume and Magnaghi, 2023). Kourdougli et al. (2017) showed that GABA depolarization might play an essential role in epileptic seizures and may have long-term effects on the morphology and function of glutamatergic networks, thereby triggering seizures.

GABRB3, the GABA receptor subunit β3, is highly expressed during brain development and plays an essential role in the assembly, transport, and differentiation of embryonic brain-derived stem cells (Møller et al., 2017). Multiple techniques, including high-throughput flow cytometry and exome sequencing, have confirmed that gene mutations in non-neuronal cells and neurons cause neuronal dysfunction. Variable surface expression is observed in GABRB3 mutants, which causes the abnormal expression of the ligand γ2 subunit. γ2 is correlated with the aggregation of GABA receptors, and mutation of the β3 subunit impairs the postsynaptic aggregation of the WT GABAA receptor γ2 subunit and its entry into GABAA receptors at the synapse. In addition, the WT γ2 subunit was downregulated, and its aggregation at inhibitory synapses was decreased in Gabrb3-knockout mice. These studies indicated that mutant GABRB3 is a significant factor in the genesis of epilepsy (Epi4K Consortium et al., 2013; Shi et al., 2019). De novo missense mutations in GABRB3 were identified in patients with early-onset epileptic encephalopathy (Shi et al., 2019). Hernandez et al. (2017) noted that the mutated residue was part of a domain (formed a GABA binding/channel-gated coupling junction) and channel holes (functionally coupled during receptor activation). The mutant even linker residues result in the rearrangement and formation of new hydrogen bonds in the open state, while the mutant pore residues reform the pores. Uncoupling of the mutant coupling residues during activation led to a functional enhancement. The mutant pore residues promoted the difference in low-conductivity receptors and sensitivity to diazepam, which may be the cause of early-onset epileptic encephalopathy (Hernandez et al., 2017).

GABRD, the GABAA receptor subunit δ, encodes extrasynaptic GABAA receptors. GABRD variation can be detected in patients with generalized epilepsy. SLC6A1-induced deficiency of extrasynaptic receptors, including GABAA receptors, was observed in patients with neurodevelopmental disorders (Ahring et al., 2022). Experiments on patients with epilepsy with GABRB2 variants revealed it significance in various epilepsy pathologies (El Achkar et al., 2021). Additional research revealed that adult epilepsy-like seizures originated in circuits with defective GABAA-mediated inhibition (Pathak et al., 2007; Figure 2).

Oxidative stress

Energy metabolism disorders in brain tissue can lead to mitochondrial dysfunction, enzyme dysfunction, and disruption of the interstitial glutamic acid-glutamine cycle. It has been demonstrated that oxidative stress occurs in neurons and astrocytes at the onset of epilepsy, followed by damage to the hippocampus in status epilepticus. The oxidative stress response in epilepsy is linked to the inflammatory factor high mobility group box 1 (Pathak et al., 2007). Furthermore, in epilepsy, excessive production of reactive oxygen species causes hippocampal sclerosis, followed by mitochondrial failure. Through its primary sensor C151, the nuclear factor erythroid 2-related factor 2 (Nrf2) activator RTA-408 inhibits recombinant Kelch like ECH associated protein 1 (KEAP1). In vitro models of epileptic seizures revealed the pathological phenomenon of RTA-408 inhibiting the production of reactive oxygen species, mitochondrial depolarization, and cell death. After in vivo status epilepticus, RTA-408 increased glutathione and ATP expression and prevented neuronal death (Mareš and Kubová, 2020).

Intestinal microorganisms

Human microflora is a large symbiotic organism accompanying human hosts throughout their lives. A healthy intestinal microbiome is made up of microbial communities and has both anti-inflammatory and pro-inflammatory effects. The gut–brain–gut axis participates in the physiological activities of the nervous system by synthesizing and secreting neurotransmitters, synthesizing metabolites, and stimulating the production of various cytokines. Consequently, a balance exists between the intestinal microenvironment and healthy nervous and immune systems (Round and Mazmanian, 2009; Tillisch, 2014). Interleukin-1, high mobility group box 1, and S100 calcium-binding protein B, which promote convulsions, are overexpressed in the glia of experimental and human epileptic tissues, neurons, and blood-brain barrier endothelial cells. The activation of the intracellular signaling cascades of inflammatory molecules by autocrine and paracrine functions suggests that chronic inflammation plays a significant role in epileptic seizures (Vezzani, 2014). In addition, the intestinal microenvironment can modulate immune and inflammatory responses, thus contributing to the development of epilepsy. Recent studies have found that, compared to a healthy control group, patients with epilepsy have significantly different intestinal flora (manifested by a lower distribution of beneficial microflora, such as Bacteroides and Actinomyces) and a higher distribution of harmful microflora (such as Firmicutes, Proteus, Verrucomicrobia, and Clostridium). Patients with four or fewer seizures per year had a greater abundance of Bifidobacterium and lactic acid bacteria than those with frequent seizures, suggesting that an imbalance in the intestinal flora may be linked to epilepsy (Arulsamy et al., 2020). Medel-Matus et al. (2018) discovered that transplanted bacteria could alter the susceptibility of rats to epileptic seizures, indicating that a micro-ecological disorder of the intestinal flora was responsible for the epileptic seizure hepatitis B state (Figure 3).

Nerve regeneration

At present, many unsolved mysteries regarding the neural regeneration mechanism of epilepsy remain; however, researchers have made some progress. Because the hippocampus is the main focus of epilepsy, and the generation of new neurons mainly occurs in the dentate gyrus of the hippocampus, nerve regeneration in the dentate gyrus was once the focus of the experimental epilepsy nerve regeneration research field (Schmidt-Hieber et al., 2004; Espósito et al., 2005; Shapiro et al., 2005; Pun et al., 2012; Myers et al., 2013; Zhou et al., 2022). Since the 1990s, some scholars have suggested that adult nerve regeneration in hippocampus occurs as a rapid increase over a short period and then a decrease in a chronic period after epilepsy. Thus, it is hypothesized that abnormalities in the morphology, location, and electrophysiological characteristics of regenerative neurons may be related to the formation of an epileptic abnormal excitation loop, that is, adult-born dentate granular cells may directly participate in the occurrence of epilepsy (Pun et al., 2012; Cho et al., 2015; Hosford et al., 2016). To further study the role of stem cells in the treatment of epilepsy, Cunningham et al. (2014) suggested that mature GABAergic intermediate neurons derived from human pluripotent stem cells widely migrated in the epileptic brain of mice, resulting in an inhibitory postsynaptic response, and that transplanted neurons could inhibit epileptic seizures and improve behavioral abnormalities, thus treating post-epileptic cells. Zhou et al. (2019) discovered that the presence of granular granulocytes in the dentate gyrus plays a crucial role in the development of epilepsy and the occurrence of spontaneous recurrent seizures. Their research demonstrated that the dysregulated integration of hippocampal newborn dentate granule cell (DGC) circuits leads to the formation of transitional excitatory connections and repetitive excitatory circuits. In a murine experimental model of epilepsy, the activation of DREADD receptors in the hippocampal newborn DGCs effectively decreased the frequency of epileptic spikes and the occurrence of recurrent seizures (Zhou et al., 2019).

Clinical observations have shown that anterior temporal lobe resection can control the seizures in patients with temporal lobe epilepsy as high as 80%. Following the operation, through neuropsychological tests and related evaluations, and coding the memory paradigm of functional magnetic resonance imaging, it was found that the working memory after temporal lobe resection depends on participation of the posterior medial temporal lobe and eloquent cortex. In addition, reorganization of the contralateral hippocampus is short-lived and inefficient 3 months after the operation. Functional remodeling of the contralateral hippocampus was performed 12 months after the operation to further repair the nerve defects caused by epilepsy (Yogarajah et al., 2010; Stretton et al., 2014; Sidhu et al., 2016).

Electroencephalography Monitoring and Diagnosis

Epilepsy monitoring technology

Currently, carbon nanotube (CNT) electrode arrays, genetically encoded calcium indicator GCaMP7 calcium imaging, jitter analysis, video monitoring, voice analysis, heart rate and heart rate variability analysis, and EEG pattern recognition are the primary methods for monitoring epilepsy, among others.

CNT electrode arrays

Given the fundamental roles of optical and electrical hybrid modalities in studying neural interface connectivity and neuronal circuits, optogenetic stimulation and multichannel electrophysiology have been used to map neural networks. The high spatial and temporal resolution of electrophysiological and optical imaging techniques allows for co-localization of neural structures and function (Nikolenko et al., 2007; Carlson and Coulter, 2008; Gonçalves et al., 2013). Researchers have developed a transparent, stretchable electrode array from woven CNT films for simultaneous electrical and optical neural interfacing in mechanically active environments. The stretchable transparent CNT electrode, which maintains high optical transparency over a broad range of wavelengths, demonstrates a favorable performance when stretched and can withstand multiple mechanical stretching cycles. In addition to high temporal resolution electrical recordings, single-activated neurons can be recorded in status epilepticus (SE) brains, allowing for real-time, sustained electrophysiological monitoring of cortical activity following traumatic brain injury (Zhang et al., 2018).

GCaMP7-based calcium imaging

Prior to the development of glioma, epileptic seizures may occur; therefore, understanding its role in the pathogenesis of brain cancers can serve as the basis for the phasic treatment of extremely drug-resistant epilepsy. By deleting tumor suppressor genes in the uterus of mice, researchers have recently developed models of gliomas. To explain the reproducible transition from hyperexcitability to convulsive seizures, the occurrence of cortical epilepsy during tumor infiltration was continuously monitored using EEG and imaging of the genetically encoded calcium indicator GCaMP7. Epileptic seizures are preceded by a loss of suppressor cells and their protective scaffolds, an increase in glial glutamate antiporter X-ray transmission computed tomography expression, an increase in astrocytes, and microglial inflammation. As the disease progresses, the inflammation becomes increasingly severe and spreads beyond the perimeter of the tumor. Additionally, it has been shown that inhibiting glial X-ray transmission computed tomography activity can prevent future seizures (Hatcher et al., 2020). Lau et al. (2022) used GCaMP7 as an important recording method for detecting wires, and combined with electrophysiological results, it showed that the probability of epileptic seizures at the level of a single neuron was random, that is, the onset of epileptic seizures can activate any neuron and trigger a large-scale epileptic seizure through a neural network, which made it possible for a large number of neurons to be juxtaposed.

EEG

Epilepsy is typically diagnosed using ambulatory EEG (aEEG), video EEG (vEEG), and stereotaxic EEG (sEEG), among which aEEG is the most useful and non-invasive; however, it is also vulnerable to a variety of artifacts. vEEG, or EEG combined with video equipment, can simultaneously monitor seizures and electrical activity in the brain (Karoly et al., 2021). vEEG has been applied to monitor sudden unexpected death in epilepsy (Ryvlin et al., 2013). Furthermore, vEEG can be used in rodent models of epilepsy in addition to observing patients’ clinical conditions. C57BL/6J mice implanted with bipolar electrodes were subjected to prolonged electrical stimulation, which produced SE. vEEG images revealed spontaneous seizures 1 month after electrode implantation, in addition to significant spatial memory impairment, depressive-like behaviors, and increased expression of inflammatory factors 7 days after SE (Thergarajan et al., 2022). Using 24-hour vEEG monitoring, Sullivan et al. (2020) recently identified epileptic seizure phenotypes, interictal spike frequency, sleep dysfunction, and hyperactivity in Syngap1^+/–^ mice, with interictal spikes primarily occurring during non-rapid eye movement sleep. In behavioral state transitions, an EEG spectral power analysis revealed an appreciable loss of gamma power modulation. Stereoscopic three-dimensional EEG was able to record the origin and propagation of epileptic discharges and evaluate the location of epileptic lesions in time and space based on clinical symptoms, cortical discharges, and neuroanatomy. A study using sEEG tracing and follow-up revealed that sEEG was most useful for patients with a history of bilateral epilepsy, deep brain stimulation, or surgery (Tandon et al., 2019). Using surgical robots for intracranial recording and localization of epilepsy has become the standard method (Tandon et al., 2019).

While scalp EEG cannot adequately localize the seizure lesion, intracranial microelectrodes are required for epilepsy EEG monitoring during surgical evaluation (Shih et al., 2018). Intracranial microelectrodes were utilized to monitor and record the encoding and retrieval of episodic memory by organizing temporal information. The accumulation of evidence over the past decade suggests that populations of “time cells” within the hippocampus of rodents encode temporal information. In 27 epileptic patients who completed a scene memory task, researchers recorded time cells using sEEG, demonstrating that time cell activity could predict the temporal organization of the retrieved memory items. In addition, it provided evidence for an increase in human neuronal activity and multi-aspect real-time monitoring of epilepsy complicated by other neurodegenerative diseases (Umbach et al., 2020).

Other detection methods

Rhythmic movements can be identified using video detection. Repeated body movements are required for vibration analysis to detect epileptic seizures. Using leg-mounted accelerometers, Cuppens et al. (2009) detected convulsions in patients with frontal lobe epilepsy with a sensitivity of 92% and a specificity of 84%. In addition, three portable devices have been used to directly detect and monitor limb or body tremors in tonic–clonic seizure patients. EpDetect employs smartphone accelerometers to detect jitter at a frequency of 5 Hz for at least 10 seconds (EpDetect, 2023). Using wrist accelerometers, the SmartWatch was shown to detect rhythmic jitter and monitor epilepsy in six patients with tonic–clonic seizures. One seizure was missed because of a dead battery (Lockman et al., 2011). The accelerometers and algorithms in the ERT watch was used to monitor 15 epileptic patients for 1692 hours, which correctly identifed 20 of 22 seizures and generated eight false positives (Kramer et al., 2011).

Electrophysiological epilepsy markers

Interictal epileptic discharge

Interictal epileptic discharge is the temporary and synchronized activity of pathologically interictal epileptiform discharges (IEDs) during epileptic seizures. The principal research interest in epilepsy is the role of transient IEDs in epileptic seizures. One study revealed that the onset of epileptic seizures is a gradual process characterized by a gradual loss of neuronal network plasticity. The principle of critical deceleration governs this slow transition from a kinetic standpoint. In epilepsy, this process is regulated by synchronized synaptic inputs from the IEDs, which are external perturbations that induce phasic changes during the slow progression and have opposing effects on the dynamics of seizure networks (Chang et al., 2018). In 2019, scientists demonstrated that IEDs are strikingly coupled to spindles in discrete, personalized brain regions outside epileptic networks (Dahal et al., 2019).

Paroxysmal slow wave events

Analyzing the EEG data of epileptic patients revealed cortical slowing, also known as paroxysmal slow wave events, which consist of a transient burst slowing of cortical networks and occur in dysfunctional blood–brain barrier cortical regions, as confirmed in vitro. Researchers have discovered epileptic activity in patients with Alzheimer’s disease; they injected serum albumin into the ventricles of young mice to increase the frequency of paroxysmal slow wave events, and identified paroxysmal slow wave events as an EEG feature of non-convulsive seizures in epilepsy patients (Milikovsky et al., 2019).

High-frequency oscillations

Recently, high-frequency oscillations (HFOs), or electrical activity with frequencies > 80 Hz, have been used to pinpoint seizure onset zones (SOZs). SOZs quantitatively reflect the severity of epilepsy and can be used to determine the efficacy of anti-epileptic drugs. Liu and Parvizi (2019) monitored spontaneous epileptic activity as HFOs and documented stimulus-locked physiological responses as physiological high-frequency broadband activity to demonstrate interactions and behavioral correlations between physiological responses of epileptic tissues and cognitive stimuli. According to the results, all epileptic sites in six patients with non-pathological epilepsy exhibited abundant normal physiological responses to cognition-related stimuli. If spontaneous HFOs occurred 850–1050 ms before or 150–250 ms after relevant cognitive stimuli, these physiological responses were more likely to be “captured” (delayed or missed). Patients’ memory performance was reported to be impaired by spontaneous HFOs in the medial temporal lobe. Thus, HFOs enable clinicians to quantitatively differentiate between pathological and physiological high-frequency activity in epileptic lesions and to indirectly assess the potential risks associated with surgical resection (Liu and Parvizi, 2019).

Sharp wave ripples

Sharp wave ripples are field potentials generated by the hippocampus during slow-wave sleep. During sharp wave ripples, memory traces are selectively reactivated. Valero et al. (2017) recorded and analyzed the activity of single hippocampal cells during physiological and pathological spike ripples in rats with normal and epileptic memories. CA1 pyramidal cells fired selectively during spike ripples, a phenomenon that may govern cell and synaptic drive specificity. Changes in the intracellularly determined excitatory/inhibitory ratio in single cells may be responsible for this firing.

Automated diagnosis of epilepsy

EEG is a standard non-invasive method for monitoring epilepsy. It takes hours for a neurologist to analyze and diagnose the 24-hour EEG data of epileptic patients, placing a significant burden on clinicians. Consequently, an automated diagnosis of epilepsy is crucial. EEG spike detection techniques include simulation technology, morphological analysis, template matching algorithms, the parameter method, independent component analysis, artificial neural networks, clustering technology, knowledge-based rules, and data mining and classification technology. Diverse approaches, including nonlinear models, artificial neural networks, Bayesian methods, independent component analysis, variance-based methods, and support vector machines, have been developed to detect epilepsy from EEG data.

Ullah et al. (2018) developed a 13-layer deep convolutional neural network algorithm for seizure interval and triage detection with an average accuracy of 88.7%, specificity of 90%, and sensitivity of 95%. This may be lower than classification models based on EEG feature screening and machine learning; however, this method does not require feature extraction and feature separation, and algorithm performance could be improved by increasing sample sizes. Li et al. (2017) proposed a new approach based on wavelet envelope analysis neural network ensembles to detect seizure intervals and epileptic seizure signals, with a 98.78% recognition accuracy rate.

The automated diagnosis of epilepsy has become increasingly sophisticated with the development of machine learning and deep learning technologies. Over 90% of normal, interictal, and epileptic seizures are correctly classified. Portable devices, such as smart watches, can detect and record epileptic seizures using heart rate and respiration measurements (Garbarino et al., 2014). Compared to EEG, portable devices are convenient and cause minimal patient disruption. Yoo et al. (2013) created a device for epilepsy detection that utilizes an eight-channel scalp EEG to detect and continuously record patient-specific epileptic seizure activity. This apparatus featured eight analog front-end channels with a wide dynamic range, a machine-learning epilepsy classification processor, and 64 kB of static random-access memory. With the development of dry electrodes, gel electrodes, wireless EEG, and other technologies, epilepsy detection devices based on EEG will soon be portable.

Prediction of seizures

Typically, epileptic seizures begin and end with no outside interference. Seizures that cannot be predicted are usually accompanied by an increased risk of accidents, such as falls and head injuries. It is necessary to predict epileptic seizures within a reasonable amount of time prior to their occurrence to reduce the occurrence of other accidents. This could also provide trigger signals for closed-loop stimulation treatments, such as reactive nerve stimulation (RNS) or serve as a reminder to take anti-epileptic drugs. Currently, epilepsy can be detected using EEG pattern recognition; however, this is impractical due to a high rate of false positives.

Several studies have examined the spectral characteristics of EEG and have determined that spectral power changes in specific frequency bands may serve as early warning signs for epileptic seizures. Park et al. (2011) explained that, in the initial stage, δ wave power (< 4 Hz) was relatively lower than other frequency bands; thus, they proposed an algorithm for predicting epilepsy based on the energy characteristics of EEG spectra and support vector machine classification.

Based on epileptic seizure precursors, Chu et al. (2017) developed a novel attractor state analysis for seizure prediction. They examined the transition from a normal to an epileptic attractor state, identified precursor phenomena prior to reaching an epileptic attractor state, and established quantitative spectral scalp EEG measurements for epileptic seizure prediction. Each of the six EEG bands was assigned a specific spectral size based on the extracted Fourier coefficients of each semi-overlapping 20-second window. Prior to epileptic seizures, the early warning index was closer to the bifurcation point that triggered the transition from the normal to epileptic state, and the spectral density of the low-frequency band of the attractor disturbance in the EEG was elevated. A low-complexity algorithm for seizure prediction employing this property was evaluated and found to achieve a high sensitivity (86.67%), a low false prediction rate (0.367), and an average prediction time of 45.30 minutes (Chu et al., 2017). Within seizure prediction models, nonlinear dynamics and the chaos theory have been applied to the study of seizure interval conversion. Iasemidis et al. (2005) utilized nonlinear dynamics to predict seizures 91 minutes beforehand with a 91.3% accuracy and a false-positive alert every 8.27 hours; this report has demonstrated the most accurate predictions to date. A study utilizing electrodes implanted in the medial temporal lobe to record seizures in 21 patients for 582 hours predicted 88 seizures, which was only marginally better than chance (Winterhalder et al., 2003). In addition to nonlinear dynamic parameters, cumulative energy, wavelet transforms, and baseline voltage crossing have all been applied to seizure prediction (Osorio et al., 2002; Esteller et al., 2005; Zandi et al., 2009).

EPILEPSIAE (http://www.epilepsiae.eu; the evolutionary platform to improve the life expectations of epileptic patient campaign) collaborates with the Micromed Group (http://micromed-it.com/) to test predictive algorithms by creating a database of epileptic seizures. In addition, Bluetooth-based portable EEG recorders are being developed.

Once a reasonably sensitive and specific seizure prediction method has been demonstrated, it will be possible to incorporate it into a “closed-loop” therapeutic device. These devices may stimulate the brain or vagus nerve to reduce the likelihood of seizures, administer anti-epileptic drugs to brain regions or elsewhere, use brain cooling devices, or instruct patients to take fast-acting (e.g., nasal spray) drugs.

Clinical Treatment of Epilepsy

Considering the severe effects of epilepsy, the current clinical treatment methods can be categorized primarily into two drug treatments and non-drug treatments.

Drug treatments

Drug therapy is currently regarded as the most effective anti-epileptic treatment; however, although medications can control most epileptic seizures, they cannot eradicate them. Reasonable half-lives and optimal drug concentrations in the blood are conducive to promoting tolerance in the central nervous system and minimizing drug side effects, resulting in maximum efficacy and anti-epileptic action.

Intervention using related dysregulated miRNAs

Currently, targeted drug interventions based on the aberrant expression of miRNAs are accessible for clinical use. Normalization of the complex molecular milieu impaired in epilepsy (seizures, traumatic brain injury, and stroke) continues to be one of the most significant obstacles in the development of anti-epileptic drugs (Morris et al., 2021). Targeting a single pathway or molecule is ineffective because disease development is caused by compensatory mechanisms or additional pathways. However, a single miRNA can simultaneously regulate multiple epilepsy-preventing or -delaying pathways. Therefore, targeting specific miRNAs can influence numerous cellular processes and may serve as an effective intervention strategy following epileptogenic damage. Oligonucleotides are the predominant therapeutic method for manipulating miRNA levels, resulting in potent and long-lasting reductions. miRNA-targeted therapies have been developed and evaluated in animal models.

As stated previously, miR-134 is associated with epileptogenesis. An intraventricular injection of antagomir, an oligonucleotide that targets miR-134, decreased epileptic seizures in animal models. Intraperitoneally or intravenously administered Ant-134 inhibits the epilepsy-induced elevation of miR-134 in the hippocampus. Furthermore, antagomir-mediated silencing of miR-134 reduced SE-induced seizures and hippocampal damage in mice, possibly through the inhibition of LIM kinase 1. The inhibition of miR-134 has a long-lasting suppressive effect on epilepsy and neuroprotective effects. Therefore, miR-134 antagonists may emerge as potential new treatments for epilepsy (Jimenez-Mateos et al., 2012, 2015; Morris et al., 2018; Reschke et al., 2021).

Targeted intervention for the abnormal expression of related genes and proteins

For excitatory synaptic transmission, the alpha-amino-3-hydroxy-5-methyl-4-isoxazole-propionic acid receptor (AMPAR) is indispensable. It mediates the majority of excitatory neurotransmission, which can lead to seizures and spread. AMPAR antagonists can be used as anti-epileptic medications, making them an appealing therapeutic target for epilepsy. The only Food and Drug Administration (FDA)-approved anti-epileptic medication, Perampanel (Fycompa), can cause vertigo and dystonia. Researchers have discovered that inhibiting AMPARs in the forebrain, but not the cerebellum, has an anti-epileptic effect without causing dyskinesia (Kato et al., 2016). The forebrain-specifically expressed molecule LY3130481 can modulate the mutant AMPAR-2. It is replaced with trace amine-associated receptor 8 (TARPγ-8) amino acid at the corresponding position, which confers selectivity and exerts an antagonistic effect on AMPARs expressed in the hippocampus of epileptic patients but not in the cerebellum. This demonstrates that LY3130481 inhibits epileptic seizures without causing motor side effects. Thus, research has identified the first molecule to target specific neural circuits, demonstrating its therapeutic value.

Medications for abnormal ion homeostasis/ion channel function

Voltage-gated potassium channels, which include the widest variety of ion channels, are essential for regulating neuronal excitability (Carmeliet, 1991; Oyrer et al., 2018). Only two potassium channel-targeting drugs, haloperidol, which acts on the Kv7.2 potassium channel, and retigabine, which acts on the Kv7.3 potassium channel, have demonstrated anti-epileptic effects (Grunnet et al., 2014). Retigabine has been extensively utilized in clinical practice, whereas haloperidol can cause liver damage, despite being effective. Therefore, researchers continue to focus on developing anti-epileptic drugs with fewer side effects. Neuronal M currents are generated by the voltage-gated potassium channel KCNQ2. It is a key target for the treatment of multiple neuronal hyperexcitation-associated disorders, such as epilepsy and pain. Cryo-electron microscopy was utilized by Li et al. (2021) to determine the structure of human KCNQ2 in its apolipoprotein form; they found that retigabine binds to the pore domain and activates the KCNQ2 channel via allosteric modulation, resulting in adequate control of epileptic symptoms.

The voltage-gated sodium channel initiates the transduction of electrical signals to generate and transmit action potentials (Payandeh et al., 2011). The crystal structure of voltage-gated sodium channels determines its channel selection properties, allowing several small-molecule hydrophobic pore-blocking drugs (such as phenytoin and carbamazepine) to enter and exert an anti-epileptic effect (Mantegazza et al., 2010; Dudev and Lim, 2014). Some scholars have found that miR-335-5p may affect neuronal excitability and epileptic seizures by regulating voltage-gated sodium channels (Heiland et al., 2023).

Ca3.1–Ca3.3 represent the T-type or low-voltage-activated subfamily of the ten subtypes of mammalian voltage-gated calcium channels; their abnormal activity is associated with epilepsy and psychiatric disorders (Zhao et al., 2019). Various blockers can selectively obstruct calcium channels in various pathologies. Selective or nonselective T-type channel blockers are regarded as potential therapeutic targets for neurological diseases including epilepsy and pain (Dibué-Adjei et al., 2017).

Valproate, a T-type calcium channel blocker, is a low-affinity Cav3 channel subtype blocker that selectively inhibits inactive channels and produces use-dependent inhibition (Zamponi, 2016). Additionally, valproate inhibits sodium channels, histone deacetylases, and additional targets. Lamotrigine, a second calcium track blocker widely used for focal and absence seizures, blocks the transiently expressed R-type Cav2.3 channels in a concentration-dependent manner within a low concentration range. The auxiliary α2δ-1 subunit is a therapeutic target of gabapentinoids, demonstrating efficacy in treating certain forms of epilepsy and chronic neuropathic pain (Dibué-Adjei et al., 2017). Nicotinic acetylcholine is a ligand-gated ion channel that mediates rapid chemical neurotransmission at the neuromuscular junction and has multiple signal transduction functions in the central nervous system (Morales-Perez et al., 2016).

GABA-targeted drugs

The onset of epilepsy is caused by the binding of the neurotransmitter GABA to GABAA receptors, which produces fast inhibitory neurotransmission. Benzodiazepines and barbiturates are current anti-epileptics that act on GABAA receptors (Zhu et al., 2018). Persistent epileptiform discharges are linked to excessive chloride ions in neurons because of decreased GABAA receptor conductance and chloride excretion. Epilepsy can also be treated with benzodiazepines, a positive modulator of chloride-permeable GABAA receptors; however, these drugs have no effect on children with SE. This is the result of the evolution of epileptiform activity patterns induced by the removal of Mg^2+^. Optogenetic stimulation of GABAergic interneurons in a state resembling SE increased epileptiform activity in a GABAA-dependent manner. Low concentrations of the barbiturate phenobarbital exacerbated SE-like movement, while high concentrations diminished or prevented epileptiform discharges. Phenobarbital at high concentrations activated GABAA receptors and hyperpolarized neuronal membrane potentials, preventing the generation of action potentials (Nakajima et al., 2012; Zhu et al., 2018; Burman et al., 2019).

When the concentration of GABA in the brain falls below a certain threshold, neuronal hyperexcitation causes twitching (Gale, 1989). This neurotransmission imbalance is correctable by inhibiting GABA aminotransferase, the enzyme that catalyzes the conversion of GABA to the excitatory neurotransmitter L-glutamate. Thus, GABA aminotransferase inhibitors can increase brain GABA levels, resulting in anti-alertness in epileptic patients, which is a novel epilepsy treatment strategy (Silverman, 2018; Auer et al., 2020). Based on this mechanism, the FDA has approved (1S,3S)-3-amino-4-(difluoromethylenyl) cyclopentane-1-carboxylic acid, also known as CPP-115. It has completed phase 1 clinical trials for safety and possesses a high therapeutic potential for a variety of epilepsies (Juncosa et al., 2018).

Several researchers have proposed GABA-mimicking drugs as anti-epileptic and anti-convulsant medications; however, their therapeutic effect depends on the chloride gradient along the plasma membrane. An altered functional balance of chloride transporters contributes to the pathophysiology of epilepsy, and the dissolution of chloride gradients may render GABA mimics ineffective. Bumetanide, an inhibitor of the Na–K–Cl co-transporter, has emerged as a new-generation chloride gradient-based anti-epileptic GABA mimic drug that re-establishes the chloride gradient and the hyperpolarizing effect of GABA mimic drugs, exerting anti-epileptic and anti-convulsant effects (Auer et al., 2020).

Non-drug interventions

Only 70% of patients can completely control their epileptic symptoms with anti-epileptic medication. Consequently, one-third of patients with refractory epilepsy continue to suffer from epilepsy, which can disrupt their daily lives. Therefore, the treatment of refractory epilepsy has become of paramount importance.

Ketogenic diets

Ketogenic diets produce ketones by controlling the ratio of fat, protein, and carbohydrates consumed to achieve a state of fat-powered metabolism. This diet provides a more efficient energy source for neurons and astrocytes in the brain than glucose, resulting in favorable metabolic changes, such as increased levels of adenosine. Some experts have also observed that a medium-chain triglyceride ketogenic diet is an effective treatment for drug-resistant epilepsy because it can increase the plasma levels of capric acid and ketones (Chang et al., 2016; Pain et al., 2021). Consequently, it has gradually become an integral part of the treatment for refractory epilepsy. This diet induces the binding of AMPARs to sites on the M3 helix of the AMPA-GluA2 transmembrane domain and inhibits AMPARs directly, playing an essential role in the inhibition of excitatory neurotransmission. At relevant concentrations, decanoic acid acts as a noncompetitive antagonist against epilepsy in a voltage- and subunit-dependent manner and alters cellular energetics via mitochondrial biogenesis. Rather than ketones, medium-chain fatty acids may prevent epileptic seizures and increase the threshold for seizures (Chang et al., 2016; Augustin et al., 2018). In a study comparing epileptic and healthy children, cronobile levels in the intestinal microbial zone of epileptic children were significantly higher. The frequency of epileptic seizures in children was reduced by 50% after 1 week of treatment with a ketogenic diet. Using 16rSDNA gene sequencing, it was discovered that the prevalence of harmful bacteria in epileptic children before treatment was significantly higher than in children from healthy families. Significant increases were observed in the genera Prevost and Bifidobacterium following a ketogenic diet. Therefore, children with epilepsy were reported to have an unbalanced gut microbiome. A ketogenic diet can relieve epilepsy symptoms by reshaping the gut microbiome (Xie et al., 2017). After 13 months of treatment with a ketogenic diet, the effectiveness rates for 389 patients with drug-resistant epilepsy were 66.8% and 83.1%, confirming once again the efficacy of treatment with a ketogenic diet (Guzel et al., 2019). Clinical research has revealed that patients with epilepsy have a higher percentage of Firmicutes/Bacteroides (Sorboni et al., 2022).

Further investigation by some scientists revealed that the ketogenic diet could reduce the diversity of the intestinal microenvironment in vivo and that the depletion of the microbial flora with high-dose antibiotics increased epileptic seizure susceptibility in WT and Kcna1^–/–^ mice (Lindefeldt et al., 2019). The process was linked to a decrease in whole-body glutamyl amino acids and an increase in hippocampal GABA/glutamic acid levels. It was possible to protect against seizures by introducing intestinal microflora from the ketogenic diet, treating with Akkermansia muciniphila, and using p-hydroxybenzoate (Olson et al., 2018).

Ketogenic diets may inhibit glutamyltranspeptidase activity in mouse feces due to an increase in bacterial plugging caused by muciniphila. This decreases the levels of glutamyl amino acids in the serum and hippocampus, which in turn reduces the frequency of seizures. In the hippocampus, increasing the ratio of GABA to glutamic acid may have anti-epileptic effects. Experts classified ketogenic diet therapy as an indication treatment in 2018 based on the clinical and experimental studies listed above (Kossoff et al., 2018).

Neuroregenerative gene therapy

Among the many epilepsy treatment methods, 20%–30% of patients are insensitive to anti-epileptic drugs, surgical resection has certain surgical risks and complications, and the effect of the ketogenic diet is limited. Based on the above problems, Zheng et al. (2022) transdifferentiated astrocytes in the hippocampus into inhibitory intermediate neurons in situ, which effectively reduced seizures and enabled GABA to regenerate the neurons. Nerve regeneration gene therapy can also directly regenerate new neurons in the epileptic region to restore the balance between excitement and inhibition, and improve cognitive and emotional disorders in a rat model of temporal lobe epilepsy, thus opening up a new way for the effective treatment of drug-resistant temporal lobe epilepsy (Zheng et al., 2022).

Neuromodulation therapy

Neuroregulation therapy is an alternative treatment for epilepsy patients who are resistant to conventional medications. By altering the function of the nervous system, neuroregulation can stop a seizure in its tracks. The precise treatment of neuromodulation is reversible and subject to regulation. Neuromodulation therapy for epilepsy primarily consists of invasive and non-invasive nerve stimulation, with invasive nerve stimulation comprising percutaneous VNS, DBS, and RNS. Currently, transcranial direct current stimulation (tDCS) and transcranial magnetic stimulation (TMS) are the most efficient treatments for epilepsy (Sui et al., 2022). Thus, patients with drug-resistant epilepsy now have access to effective treatments for seizure control, including VNS, DBS, and RNS (Figure 4).

At a predetermined time, open-loop stimulation can be applied. In contrast, closed-loop stimulation is a neural regulation technique that uses epilepsy biomarkers (EEG, electrocorticogram (ECoG), local field potential, etc.) to determine the stimulation’s magnitude and timing. Included in the neuromodulation therapy mode are these two types of stimulation. Closed-loop stimulation aims to minimize treatment-related side effects by delivering stimulation only during a predetermined time window. The viability of closed-loop stimulation is primarily dependent on the accurate detection of biomarkers that accurately reflect the severity of patients’ symptoms and changes in response to treatment. Using numerous signals, including pathological neural activity and peripheral measurements, it has been possible to calculate when and how much to stimulate. Instead of being directly associated with the disease mechanism, biomarkers should be linked to the severity of disease symptoms and used to monitor the efficacy of intervention therapy (Benabid et al., 2002; Kobayashi and Pascual-Leone, 2003; Cagnan et al., 2019). Figure 5 depicts the closed-loop stimulation mode. Epilepsy is predicted by decoding continuously collected neurophysiological signals. The instructions are then transmitted to the stimulator, which adjusts the stimulation parameters to stimulate a specific brain region.

By controlling brain networks through the afferent nerves of the vagus nerve, VNS lowers the frequency of epileptic seizures. Currently, the FDA has approved VNS for treating numerous epilepsy conditions, including Dravet syndrome and severe myoclonic epilepsy in newborns. Rodent studies have demonstrated the effects of VNS on a variety of neurophysiological traits and neurotransmitters, including the increase in the norepinephrine concentration related to an anticonvulsant effect and the decrease in the synchronization of epileptic EEG activity (Ben-Menachem, 2002; Liu et al., 2018). However, the long-term improvement of VNS seizure control may be through other mechanisms, such as the regulation of GABA function and neuroinflammation (Okun, 2012). According to the meta-analysis based on 16 clinical studies in 2019, VNS technology was applied to 1061 patients, showing that 53% of patients had a reduction of more than 50% of seizures. In addition, with the extension of implantation time, the effective rate of treatment also increases, which is also supported by other studies (Chrastina et al., 2018; Kuba et al., 2009; Wang et al., 2019).

In the year 2000, transcutaneous VNS (tVNS) was introduced as the first treatment for epilepsy (Révész et al., 2018). Traditional VNS is more expensive than tVNS. Furthermore, tVNS does not require invasive surgery, has fewer adverse effects and secondary tissue damage, and is easy to use. A small stimulator is implanted in the outer ear for tVNS, which employs electrical impulses to stimulate a branch adjacent to the vagus nerve, thus stimulating the vagus nerve (Sharon et al., 2021; Steidel et al., 2021). The efficacy and safety of tVNS-stimulated and control groups were evaluated in a randomized, double-blind controlled trial (cMPsE02). Through efficacy assessment, subgroup analysis, and safety evaluation of 25 and 1 Hz tVNS stimulation, it was determined that 25 Hz stimulation reduced the frequency of epileptic seizures more effectively while maintaining a comparable response rate and safety profile (Bauer et al., 2016). Tecoma and Iragui (2006) compared the parameters of VNS and found that a pulse width of 0.25 ms was better tolerated than 0.5 ms, with comparable therapeutic effects. Stimulation at a frequency lower than 20 Hz stimulated demyelinating C fibers with greater adverse effects. There is evidence that the curative efficacy and adverse effects of heart-based closed-loop VNS are superior to those of conventional VNS (Tzadok et al., 2019; Haberbusch et al., 2022). Several neurophysiological and imaging biomarkers are associated with the VNS response, including P300 amplitude, left thalamus and limbic network fractional anisotropy and functional connectivity, median nerve somatosensory evoked field, and heart rate variability (Neuhaus et al., 2007; Wostyn et al., 2017). A systematic review demonstrated that tachycardia is associated with the majority of seizures (Theodore and Fisher, 2004). Consequently, a closed-loop VNS can automatically trigger stimulation by detecting heart rate change during seizures.

During brain stereotactic surgery, the DBS procedure employs implanted electrodes and neurostimulators to stimulate specific deep brain nuclei and correct abnormal cerebral electrical circuits (Zhang et al., 2023). DBS of the anterior thalamic nucleus has become a promising treatment for drug-resistant epilepsy in specific areas. DBS aims to prevent the transmission of epileptic seizure activity by targeting the most influential “transmission point” downstream of the epileptic network (Gooneratne et al., 2016; Yu et al., 2018). The current consensus regarding the mechanism of DBS is that high-frequency stimulation restores function by regulating the neural activities of afferent and efferent brain regions (Cagnan et al., 2019). The severity of a patient’s symptoms is proportional to the amount of excessive rhythmic nerve activity in the DBS target and projection target region. It has been shown that high-frequency electrical stimulation inhibits rhythmic nerve activity while alleviating patient symptoms. In a clinical study involving nine patients with drug-resistant focal epilepsy, the anterior thalamic nucleus was stimulated at 130 Hz to record local field potentials in the cortex. DBS on the anterior thalamic nucleus could significantly reduce phase spikes and HFOs during epileptic seizures and decouple large-scale neural electrical activity in the hippocampus and cortical regions, which may be associated with the effect of high-frequency stimulation-induced ipsilateral hippocampal activity desynchronization (Yu et al., 2018; Ren et al., 2020). In a study of eight epilepsy patients treated with deep brain stimulation, the frequency of seizures decreased by approximately two-thirds (Lee et al., 2012). In contrast, in a study of nine patients, seizures decreased by 66%–100% after a mean of 30 months of follow-up (Boëx et al., 2011). A single-blind cross-over study revealed that centromedian thalamic nuclei DBS was effective for generalized seizures (all six patients with epilepsy responded to treatment), but ineffective for frontal lobe seizures (only one in five patients responded) (Valentín et al., 2012). The use of thalamic DBS for the treatment of drug-resistant partial and secondary generalized seizures has been approved. Yu et al. (2018) examined nine patients with temporal lobe epilepsy (TLE) and recorded intracranial EEG using electrodes in the anterior thalamic nucleus. High-frequency stimulation of the anterior thalamic nucleus caused broadband local field potentials in the anterior thalamic nucleus to be out of phase with those in the ipsilateral hippocampus and neocortex. Scher et al. (2020) investigated intermittent anterior thalamic nucleus DBS in 14 patients with TLE using scalp EEG; the results supported the conclusion of desynchronization. The study found that stimulation caused θ and α activities to be out of sync, which could be used to treat epilepsy. High-frequency DBS can prevent and synchronize the transmission of epileptic seizure activity from the epileptic seizure region to the brain-led network. Through inhibition of the production of local epileptic seizure activity, DBS can also prevent the spread of epileptic seizures (Nair et al., 2020). Segmented electrodes allow for the independent control of VTA electrode contacts through field manipulation. In addition, thin film planar arrays can further increase the spatial specificity of stimulation, thereby maximizing the curative effect and minimizing side effects. For stimulation opportunity control, closed-loop DBS relies primarily on signal characteristics, such as neural activity or frequency bands that reflect the severity of the patient’s symptoms. In patients with primary epilepsy, the characteristics extracted from thalamic field potentials have been used to decode epileptic seizures (Tan et al., 2019; He et al., 2020). A recent study has shown that the closed-loop direct stimulation of patients’ striatum can enhance their cognitive control ability during the task and can be directly decoded through neural features (Basu et al., 2023).

The RNS is a closed-loop system intended to continuously monitor the intracranial EEG of the seizure area. Based on the EEG pattern of an impending seizure, RNS delivers high-frequency, short-pulse electrical stimulation to halt the seizure. RNS was designed to inhibit seizures by inducing the SOZ’s “response” to recorded SOZ epileptic activity. RNS is a closed-loop system in which both the receiver and transmitter electrodes are located in the SOZ; it suppresses local or regional synchronization during and between epileptic seizures. This technology was approved by the FDA in 2013 for drug-resistant focal epilepsy patients > 18 years of age with two or more epileptic foci (Kossoff et al., 2004; Stacey and Litt, 2008). A 2-year study of 191 patients with drug-resistant partial seizure disorder revealed a 44% and 53% decrease in median seizures 1 and 2 years following RNS device implantation (Sun and Morrell, 2014). A 100-hour RNS study on four rats utilizing a TLE model predicted an 88%–96% decrease in seizure frequency. According to the report, 11 consecutive patients with focal epilepsy who had RNS systems implanted within the previous 2 years were documented and followed up. Electrophysiology determines the mode of action of RNS based on the onset latency associated with the initiation of stimulating events. The average frequency, severity, and duration of seizures were evaluated and studied using questionnaires. Direct and indirect effects dominated the electrophysiological characteristics of stimulation-induced modulation of the seizure network. In these 11 patients, epilepsy symptoms and survey scores were ameliorated by indirectly influencing the location of the focus to trigger stimulation via spontaneous epilepsy arrangement, frequency adjustment, fragmentation adjustment, and duration adjustment. The therapeutic effectiveness of RNS is a result of the stimulation of epilepsy network activity over time. To comprehend how RNS regulates epileptic network activity, researchers found that functional connectivity and frequency bands decrease in epileptic seizure regions. The frequency band’s functional connectivity was greater in “super” responders (reduced by 90% during seizures) than in poor responders (reduced by 50% during seizures). This result demonstrated that RNS can desynchronize epileptic networks and reduce the risk of epileptic seizures, which are caused by epileptiform discharges between seizures. Long-term neuromodulation of epileptogenic networks may be responsible for the efficacy of RNS, as opposed to simply preventing seizures (Kokkinos et al., 2019). Closed-loop RNS induces plastic alterations in the epileptogenic region and decreases the frequency of focal seizures. DBS stimulation of the anterior thalamic nucleus also decreases epileptic discharge and HFOs during the epileptic seizure phase, providing additional evidence that epileptic seizure areas are plastic. However, the effects of stimulation on brain networks and their dynamics require further study (Yu et al., 2018).

TMS refers to the use of electromagnetic pulses or alternating magnetic fields to excite or inhibit tissue currents. TMS is non-invasive, painless, and secure. Nine patients with partial or secondary systemic epileptic seizures were treated with TMS (Tergau et al., 1999). The dorsal region of the head was stimulated using a circular magnetic coil in two sequences per day, with 500 pulses per sequence and one pulse every 3 seconds. The patients’ weekly seizure frequency decreased significantly from 10.3 ± 6.6 to 5.8 ± 6.4 after treatment. A series of subsequent studies demonstrated the efficacy of TMS in preventing epileptic seizures (Menkes and Gruenthal, 2000; Graff-Guerrero et al., 2004; Misawa et al., 2005; Rotenberg et al., 2008).

Fregni et al. (2006b) used TMS to target epileptogenic foci in 21 patients with drug-resistant seizures. Patients received five consecutive 20-minute stimulations per day. After 2, 4, and 8 weeks, EEG epileptiform discharges and seizure frequency decreased by 72%, 53%, and 58%, respectively. Two studies administered TMS to 43 patients with drug-resistant epilepsy using two daily sequences of 500 stimuli per sequence at a frequency of 0.3 Hz and an intensity equal to 100% of the motor-evoked threshold (Cantello et al., 2007; Walton et al., 2021). No significant changes in seizure frequency or EEG epileptiform activity were observed. A second study examined the effectiveness of TMS in 21 patients with focal epilepsy, who were treated twice per day for 15 minutes per session for 1 week at a frequency of 1 Hz and an intensity of 120% of the motor-evoked threshold. With the coil placed in the estimated focal region, neither focal nor systemic seizures significantly improved (Theodore et al., 2002). Depending on the stimulation area and parameters such as frequency, intensity, and duration, TMS can have various therapeutic effects. Even a single pulse of TMS can result in seizures.

tDCS is a novel non-invasive intervention that induces excitability changes in the cerebral cortex by passing a weak current through the scalp. Anodal tDCS increases cortical excitability by depolarizing cell membranes, while cathodal tDCS decreases cortical excitability by hyperpolarizing cell membranes. Cathode tDCS is a safe and effective method for treating drug-resistant focal epilepsy (Lafon et al., 2017; Wang et al., 2020, 2022; Sudbrack-Oliveira et al., 2021). tDCS is currently utilized primarily to treat depression and insomnia. A study found that patients with drug-resistant focal epilepsy treated with cathodal tDCS had decreased EEG epileptiform discharge rates and no adverse events (Sudbrack-Oliveira et al., 2021). In another study, the tDCS anode was placed in the F3 region (left hemisphere) for 20 minutes, while the cathode was placed in the F4 region (right hemisphere) (1.5 mA); according to their analysis, tDCS showed potential to alleviate depression, anxiety, and stress in epileptic patients (Azmoodeh et al., 2021). Ten patients with drug-resistant TLE were treated with transcranial direct current stimulation, with the cathode placed in the epileptogenic zone and the anode above the contralateral homologous region. The stimulation current was 1 mA and lasted for 20 minutes. After 1 week of stimulation, seizure frequency decreased by 71.33% (Assenza et al., 2017). Patients with drug-resistant epilepsy and cortical development malformations participated in a controlled trial of tDCS. After a single stimulation, Fegni et al. (2006a) discovered that EEG epileptic discharges decreased by 64.3%, but seizure frequency did not change significantly. Neither seizure induction nor an increase in EEG discharge rate were observed during or after stimulation. Zoghi et al. (2016) investigated the effects and duration of tDCS in a sample of drug-resistant TLE patients. A 1 mA stimulation current was administered, followed by two 9-minute treatments separated by a 20-minute interval. This treatment reduced seizure frequency by 42% and prolonged the effects of cathodal stimulation. In a double-blind, controlled trial involving children with Lennox–Gastaut syndrome, Auvichayapat et al. (2016) increased the stimulation duration by delivering 2 mA of stimulation current for 5 consecutive days over 20 minutes per session. After 4 or 3 weeks, the number of daily epileptic seizures and epileptiform discharges significantly decreased. These effects diminished in magnitude over time (Auvichayapat et al., 2016). Yang et al. (2020) performed tDCS on 70 patients with refractory focal epilepsy. Stimulation was administered twice daily for 20 or 40 minutes, with a 20-minute interval and a 2-mA stimulation current. The anode was placed in the contralateral “silent” area, and the stimulation lasted 4 or 5 weeks (long period). The frequency of seizures was reduced by 64.98%–66.32% at the 8-week follow-up in the group that received the longer treatment course (Yang et al., 2020). In a cross-over study, Tekturk et al. (2016) administered modulated cathodal stimulation with a sinusoidal waveform to participants with TLE due to hippocampal sclerosis. The peak current intensity was 2 mA at a frequency of 12 Hz for 30 minutes, three times per day. The purpose of this intervention was not only to inhibit the epileptic region, but also to “restore the natural rhythm of cortical activity”. Following stimulation, the average monthly frequency of seizures significantly decreased from 10.58 at baseline to 1.67 1 month later. In conclusion, tDCS is a safe and effective intervention for drug-resistant focal seizures. It also decreases the time between treatments and lengthens the duration of the effects. The cathode is either placed in a specific parietal/temporal lobe-related region to enhance the activity of the so-called antiepileptic system, or in a relevant central processing region of the epileptic network (Yang et al., 2019). The stimulation current is typically 1–2 mA, with single stimulations lasting 19–40 minutres and daily stimulations ranging from 1 to 14. Generally, seizures and increased EEG discharges are not induced by stimulation.

VNS is a more secure invasive neurostimulation technique than DBS. In terms of efficacy data, thalamic DBS and RNS demonstrated greater reductions in seizure frequency compared to baseline than VNS. Regarding acceptability and cost, noninvasive neurostimulation is more desirable than invasive neurostimulation. In addition to enhancing the temporal and spatial specificity of stimulation, segmented electrodes, thin-film planar arrays, and closed-loop stimulation can reduce side effects and improve efficacy.

Summary and Perspectives

Epilepsy is characterized by recurrent seizures, which have serious physical and mental health consequences. Accurate diagnosis and treatment require a comprehensive understanding of the pathogenesis. Studies have shown that the pathogenesis of epilepsy is complex, involving changes in gene expression, protein expression, ion channel expression, energy metabolites, intestinal microenvironment, and other variables. With traditional treatment methods, a radical cure is still difficult to achieve. To address the pain caused by epilepsy, research on treatment has become increasingly most important, among which surgical resection, drug therapy, and non-drug therapy, including a ketogenic diet, nerve regeneration gene therapy, and neuromodulation, are the current focus.

On the one hand, nerve regeneration gene therapy may become a more promising treatment method, because it can produce GABAergic neurons through transcription factors, and functionally integrate the intermediate neurons transformed from astrocytes into the hippocampus, saving the abnormal behavior caused by intractable temporal lobe epilepsy. This behavior not only produces functional new neurons in the brain of rats, but also protects the existing neurons, and effectively treats epilepsy. On the other hand, we look forward to an innovative epilepsy treatment method, which can be used as a supplement to drug treatment. Neuromodulation is more adaptive, reversible, and usually safer than surgery. Epilepsy patients who are resistant to drug therapy also benefit from this treatment. In the future, neuromodulation will be non-invasive because of its acceptability, no requirement for surgery, and low cost. By understanding the pathogenesis of nerve stimulation and the regulation mechanism of the brain network, neuromodulation therapy will be more effective. In addition, predicting epileptic seizures can improve the timeliness and effectiveness of intervention measures, which is very important for improving patients’ health and quality of life. The introduction of dry electrodes, gel electrodes, wireless EEG devices, portable devices, and algorithms to identify epilepsy will greatly promote the prediction of epileptic seizures. Therefore, future research should focus on the molecular pathophysiology of epilepsy, including the temporal and spatial patterns of cell and molecular expression during seizures. In addition, it is very important to develop effective equipment and technology for epilepsy treatment and diagnosis. This paper provides a comprehensive overview of the opinions of numerous scholars, highlighting the benefits of neuroregulation in treating various forms of epilepsy. However, it is important to note that the applicability of neural regulation as a treatment method for all types of epileptic seizures remains uncertain. Current research primarily focuses on investigating the molecular-level pathogenesis through analyzing the clinical manifestations of patients and exploring relevant diagnostic and treatment approaches. It is worth mentioning that the lack of a consensus on the disease-related pathogenesis poses a significant challenge and will be a prominent area of future research.