Seizures arise from aberrant and coordinated neural activity within specific cerebral regions or across the entire brain, resulting from irregular neural network formation or disruptions induced by structural, infectious, or metabolic factors. Seizures constitute ephemeral phenomena characterized by precisely defined onsets and conclusions [3].

In the pediatric population, seizures predominantly stem from genetic factors, perinatal insults, or malformations of cortical development [3]. In adults without a familial history of epilepsy, common etiologies include conditions such as encephalitis, meningitis, head injuries, and the presence of brain tumors. In elderly individuals, seizures are often associated with head trauma and brain tumors [3].

As articulated by the International League Against Epilepsy (ILAE), epilepsy clinically manifests through (1) the occurrence of two or more unprovoked seizures with an interictal interval exceeding 24 h, or (2) a solitary unprovoked or reflex seizure when the likelihood of recurrence exceeds 60% in the subsequent decade, or (3) a clinical diagnosis of an epilepsy syndrome [3, 4]. The second criterion deems an individual at risk when brain imaging reveals epileptogenic potential or when epileptiform activity manifests on an electroencephalogram [1]. The classification of epilepsy involves three discernible strata: the nature of seizure onset, the specific epilepsy subtype, and the identification of epilepsy syndrome [1, 2].

Epileptogenic constitutes a nuanced phenomenon involving the conversion of a non-epileptic brain into a condition capable of generating spontaneous and recurrent seizures [5–7]. This process entails the development of brain tissue endowed with the capacity to generate seizures, culminating in the establishment of an epileptic condition and the progression of epilepsy even after diagnosis. In contrast to previous conceptions, palatogenesis extends beyond the timeframe between the epileptogenic trigger and diagnosis, encompassing mechanisms contributing to the ongoing progression of epilepsy [6, 7]. This intricate process is underscored by a delicate balance between excitatory and inhibitory activities within neural networks. Prolonged aberrant neural activity disrupts routine neural processing, exerting deleterious effects on other neural networks [1, 5].

In generalized epilepsies, the networks implicated in epileptogenic typically exhibit widespread distribution, involving bilateral thalamocortical structures [1, 2]. In contrast, focal epilepsies primarily implicate neural circuits within a single hemisphere, often situated in the limbic or neocortical regions [2]. In specific scenarios, such as the absence of seizures and limbic epilepsies in an immature brain, an abnormal increase in inhibition may contribute to a pro-epileptogenic state [1]. While numerous generalized epilepsies are postulated to possess a genetic basis, focal epilepsies have traditionally been associated with structural cerebral abnormalities, particularly in instances of drug-resistant epilepsy (DRE) [1, 8, 9].

The precise pathophysiological mechanisms underpinning the nexus between structural abnormalities and the onset of seizures remain incompletely understood. Seizures are primarily attributed to abnormal activity in cortical neurons, although there may be secondary involvement of glial cells and axons within the white matter [1].

Biomarkers associated with palatogenesis can serve multiple purposes, including predicting the emergence of epileptic conditions, identifying tissue capable of generating spontaneous seizures and determining its extent, measuring the progression of epilepsy post-establishment, and facilitating the cost-effective development of animal models for potential antiepileptogenic drugs and devices. Moreover, these biomarkers offer a means to reduce costs associated with clinical trials for potential antiepileptogenic interventions by including individuals at a heightened risk of developing epilepsy in the trial cohort [7, 10].

The connection between cerebral vascular malformations and epilepsy is still a topic of discussion [11]. In many cases, it is commonly observed that the specific area associated with the vascular malformation is closely linked to the occurrence of focal epileptogenic activity, and the removal of the malformation often results in better seizure control. There are exceptions to this pattern, including situations where vascular malformations are not directly linked to the seizure disorder or are part of a more extensive brain condition characterized by a broader influence on epileptogenesis [12]. Epileptic seizures frequently occur in patients with brain vascular malformations such as AVMs and cavernous malformations, significantly affecting their quality of life [12, 13].

Factors that increase the risk of seizures in patients with AVMs include male gender, younger age, frontal or temporal lobe AVMs, brain cortex-located AVMs, a superficial venous drainage, a superficial temporal lobe AVM nidus, fistulous AVMs, and AVMs with venous stenosis [14–19]. Some studies include a larger AVM (> 3 cm) as an independent predictor of seizures [14, 20].

Seizures in previously ruptured AVMs may result from defects and gliosis caused by intracerebral hemorrhage. However, seizures in unruptured AVMs have a more complex explanation [19]. The tissue surrounding AVMs is particularly vulnerable to hypoperfusion due to the “steal phenomenon” and a lack of vascular autoregulation [21]. The reduced cerebrovascular reserve in the perinidal area can lead to changes in the perinidal cortex induced by hypoperfusion, resulting in chronic ischemia and the development of new blood vessels. These changes promote excitatory pathways in neurons, easily developing epileptogenic mechanisms [20, 22–24]. Localized hypoxia resulting from hypoperfusion can lead to gliosis and trigger seizure activity [19].

Rajeswarie et al. [11] investigated angioarchitectural factors in cerebral vascular malformations related to changes in neuroglial and stromal components in the adjacent cortex to understand the physiopathology of epilepsy in these individuals. The study included 32 patients, with 24 having AVMs and 8 having cavernous malformations [11]. Among the AVM patients (consisting of 18 men and 6 women), 50% experienced seizures as their initial clinical presentation. The most affected areas were the frontal lobe (33.3%) and the temporal lobe (29.2%). The analysis revealed a significantly higher occurrence of hemosiderin deposition and gliosis in the tissue surrounding the lesion in patients with seizures compared to those without seizures (P < 0.05). These findings support the idea that in AVMs, vascular remodeling, chronic bleeding, and microhemorrhages lead to hemosiderin deposition in the surrounding tissue, subsequently triggering gliosis and promoting the development of seizures [11].

The diagnostic rate for AVMs is estimated at 1–1.42 cases per 100,000 people [44, 52–54]. This rate decreases to 0.84 cases per 100,000 individuals when AVMs are detected due to a hemorrhagic presentation [54]. Autopsies have shown a wide AVM prevalence ranging between 5 cases and 613 cases per 100,000 corpses [55, 56].

The average age at which AVMs are diagnosed ranges between 32 years and 39 years [25–36]. In pediatric patients, AVMs have an estimated diagnostic rate of 0.014–0.028% [40]. No gender differences are observed in either pediatric or adult series [36].

The estimated mortality rate in AVM patients is 0.7–2.9%. Untreated patients have a 50% higher risk of mortality 30 years after diagnosis compared to the general population. For patients with completely occluded AVMs, the risk of mortality 30 years after diagnosis is 13%, as compared to the general population [25].

Angioarchitectural factors assessed through digital subtraction angiography include the origin of feeding arteries from the external carotid artery, middle cerebral artery, or posterior cerebral artery, as well as the presence of superficial venous drainage with venous ectasia [18, 20, 59, 67, 69].

Fierstra et al. [20] proposed that a high-flow shunting vascular system within the AVM may result in the formation of aneurysms. This process generates sufficient wall shear stress to disrupt blood flow, potentially contributing to the occurrence of seizures [72–74]. Considering the characteristics of high-flow AVMs, including venous varix, venous ectasia, and intranidal fistulae, Shankar et al. [19] introduced a scoring system based on these angioarchitectural features. Their findings indicate that higher AVM flow correlates with an increased likelihood of seizures [75]. This scoring system relies on three robust predictors: the AVM’s location (considered cortical when the AVM blood supply originates from the cortical branches of the anterior cerebral artery, middle cerebral artery, or posterior cerebral artery), venous outflow stenosis (defined as a reduction of 50% or more in the diameter of the vein), and the presence of a long pial draining vein (defined as longer than a 3 cm superficial course of the draining vein). Each of these variables is assigned a score of 1 when present [75].

In this system, obtaining a score of 0 demonstrates excellent sensitivity and a high negative predictive value. Surpassing the threshold score of 2 results in overall high diagnostic efficacy, while achieving a perfect score of 3 out of 3 corresponds to the system’s highest positive predictive value and strongest specificity (98%) [75].

Benson et al. [76] observed that various factors correlated with the initial presentation of seizures in AVM patients, including perinidal edema, perinidal T2 MRI blooming, the presence of a venous pouch or varix, the existence of a long draining vein, and large AVM size. These findings align with angiography observations in AVM patients [59, 75, 76].

In terms of AVM location, seizures were associated with AVMs situated in the frontal lobe, motor cortex, and sensory cortex [17]. A significant correlation also exists between epilepsy incidence and damage to both the right precentral gyrus and the right superior longitudinal fasciculus [17]. Zhang et al. [70] emphasized that a high AVM heterogeneity heightened the risk of seizures but did not establish connections between radiomic attributes and the detailed angioarchitecture of AVM. Different authors did not take into account the perinidal regions, potentially overlooking epilepsy induced by gliosis or vascular steal in adjacent brain regions [59, 70, 74].

Blood oxygenation level-dependent (BOLD) cerebrovascular reactivity (CVR) mapping is an emerging hemodynamic imaging technique conducted through functional MRI [77–79]. This technique allows for the quantitative assessment of cerebrovascular reserve capacity on a voxel-by-voxel basis and is capable of quantitatively assessing arterial steal in cerebrovascular pathologies. Arterial steal is observed as an unexpected drop in BOLD signal during vasodilation, resulting in an adverse CVR outcome [77–79]. A recent study employing this BOLD-CVR technique revealed weakened perinidal CVR in conjunction with venous congestion observed in standard angiography. This finding suggests that patients with brain AVMs are more likely to experience seizures [59].

Individuals with epilepsy associated with brain AVMs demonstrate a notable decrease in whole-brain CVR compared to those without epilepsy. This reduction in CVR is observed in both gray and white matter [80].

Sebök et al. [81] reported that a decrease in overall brain CVR, coupled with pronounced hemodynamic alterations like the arterial steal phenomenon and venous outflow restriction, is associated with an increased risk of seizures in patients with brain AVMs. These findings align with current and prior research [59, 81, 82], suggesting that the non-invasive BOLD-CVR method could potentially serve as an imaging marker for predicting epilepsy in individuals with AVMs [81].

The role of microsurgery in the treatment of AVM-related epilepsy is a topic of ongoing debate, although recent series have shown promising results [59]. Some studies have reported high rates of seizure freedom, ranging from 77% to 91%, following microsurgical intervention, and in some cases, these positive outcomes persisted for up to 10 years [59, 86]. Patients with DRE often experience less favorable outcomes for seizure control after microsurgical resection [59]. Potential better outcomes seem to correlate with extended lesionectomy in comparison to standard AVM resection, but the significance of this finding is limited due to the small amount of available evidence [86].

Microsurgery aims to completely remove brain AVMs, resulting in a change in seizure response as its primary goal [69]. A meta-analysis by Baranoski et al. [87] determined that successful seizure control in AVM-related epilepsy is closely associated with the complete obliteration of AVMs, and microsurgery is more likely to accomplish this outcome. Various surgical series have documented a substantial level of seizure freedom following microsurgical AVM removal, with rates around 77% at 1 year, 79% at 2 years, and 84% at 10 years of follow-up [65, 66, 87–89].

In epilepsy-related AVM surgery, intraoperative neuromonitoring techniques, including electroencephalography, somatosensory-evoked potentials, and intraoperative electrocorticography, appear promising but have been sparingly described in the literature [90, 91]. These tools can identify epileptic foci resulting from secondary epileptogenesis and guide extended lesionectomies to enhance seizure control [90, 91]. Advances in microsurgical technology, neuroimaging, and intraoperative neuromonitoring enable safe microsurgical removal of AVMs located in eloquent motor areas or deep-seated regions. Nevertheless, the treatment-associated morbidity for high-grade AVMs remains elevated [92, 93].

SRS represents a secure alternative to microsurgery and EVE for achieving the elimination of seizures [59]. SRS involves the precise delivery of a single high dose of radiation to a specific intracranial target area, aiming to treat or eliminate a preexisting lesion. While this technology may result in reduced AVM obliteration and a longer time to achieve seizure control, it can lead to a decreased risk of experiencing an epileptic crisis and lower morbidity [85, 94].

A meta-analysis by van Beijnum et al. [92] revealed that, in terms of the case-fatality rate, SRS demonstrated the most favorable outcome with a rate of 0.5 per 100 person-years compared to other procedures. This rate outperformed the rates of 1.1 following microsurgery and 0.96 after EVE. SRS was also associated with the lowest complication rate at 5.1%, whereas microsurgery and EVE had rates of 7.4% and 6.6%, respectively. However, SRS exhibited a lower obliteration rate at 38% compared to microsurgery, which achieved a 96% obliteration rate. This raises questions about the effectiveness of SRS in achieving sufficient seizure control [91].

In other reports on SRS for AVMs, 69% of patients saw an improvement or complete absence of status epilepticus, and 48% experienced a total absence of seizures [92]. The effectiveness of SRS in managing AVM-related epilepsy seems to arise from mechanisms such as neuromodulation of the perinidal nervous tissue, the formation of a gliotic capsule, and vascular modifications [59, 92, 93].

EVE is a minimally invasive procedure in the field of interventional radiology used to intentionally block blood vessels as a treatment for diseased or injured vasculature [94]. Embolizing brain AVMs appear to be beneficial for seizure control by inducing hypoxia in epileptogenic brain tissue, although this might stimulate intranidal angiogenesis and lead to seizure recurrence. AVM patients treated solely with EVE, who continue to experience epileptic seizures, seem to have residual gliotic plaques, perinidal gliosis, and epileptic foci distant from the primary AVM site [95].

EVE serves as an adjunct procedure to other treatment approaches for AVMs. The exploration of EVE as an independent treatment method for AVMs is relatively limited [85]. A study led by Zhang et al. [71] evaluated seizure control in patients who underwent EVE with ethylene vinyl alcohol (Onyx). Complete obliteration and seizure freedom were attained in 51.4% of the AVM patients. Those with complete EVE achieved higher seizure control rates at 57.9% compared to those with partial EVE at 44.4% [71].

AVM-related seizures treated through AVM microsurgical excision exhibit favorable overall outcomes in terms of seizure freedom. Various AVM microsurgical series report mean overall rates of seizure-free outcomes (Engel scale class I) ranging between 61.54% and 96% [59, 97, 99–102, 125, 126, 128–130]. The average follow-up time across all series was 65.75 months. Microsurgery emerges as the AVM treatment with the highest seizure freedom rates [59, 75, 97, 99–102, 125, 126, 128–130]. A multimodal treatment meta-analyses conducted by Lak et al. [96] described higher rates of seizure control in patients treated with microsurgery compared to SRS and EVE. This meta-analysis included 49 studies involving 2,668 patients, reporting seizure control rates as follows: microsurgery, 65.0%; SRS, 58.5%; and EVE therapy, 48.0% [96]. Hyun et al. [101] reported 78% seizure-freedom outcomes in patients treated with microsurgery while seizure-free outcomes after SRS or EVE were 66% and 50%, respectively. The authors found no significant differences in Class I outcomes between microsurgery, SRS, and EVE (P = 0.1) [101].

Series that report seizure freedom outcomes in patients with partial and total obliteration of AVM volume treated by SRS show a wide range of results, varying from 18.2% to 89% [65, 104, 106, 107, 115–122, 127]. Some of these outcomes indicate a lower seizure control rate (mean of 61%) compared to AVM microsurgery (mean of 79%) [14, 86, 97–101, 112, 114, 129, 130]. The average follow-up time across all series was 58 months. SRS, introduced in the late 1980s as an alternative to microsurgical resection for AVM treatment, has been considered a safe option [14]. Despite its safety, the available literature indicates a lower rate of seizure control compared to microsurgical treatment [14].

Baranoski et al. [87] conducted a meta-analysis of 137 series involving 13,698 AVM patients. The authors concluded that SRS is a more effective treatment strategy than microsurgery and EVE for seizure control in AVM patients. However, this assertion raises questions as it is contingent on the condition that SRS is superior only when achieving complete obliteration of the AVM. The rate of complete obliteration of AVM volume with SRS was only 38%, in contrast to the 96% achieved with microsurgery [87]. Soldozy et al. [59] also highlighted this aspect in their reports.

Zhang et al. [71] reported seizure predictors and outcomes after Onyx embolization in AVM patients. Among 239 patients studied, 28.5% initially presented with seizures. The occurrence of seizures correlated to a history of cerebral hemorrhage, as well as the frontal-temporal and arterial borderzone locations. Among the 37 patients who experienced initial seizures and underwent Onyx embolization, 23 (62.2%) received ASM before the embolization. During the last follow-up, 19 (51.4%) of these 37 patients achieved a modified Engel class I outcome. Among the 23 patients previously treated with ASM, 12 (52.2%) were still using ASM at the last follow-up. Single-factor analysis revealed a significant correlation between arterial borderzone location and a higher modified Engel class outcome (P = 0.046) [71].

Lv et al. [94] reported their experience in addressing seizures associated with brain AVMs in 109 patients without a history of intracranial hemorrhage. The 30 patients with seizure disorders related to brain AVMs underwent endovascular treatment. Post-embolization seizure control after an average follow-up of 80 months was excellent in 21 patients, good in four, fair in two, and poor in three. The study concludes that EVE achieves successful seizure control in AVM patients [94].

Factors linked to pre-operative AVM-related seizures, as documented in the literature, encompass the presence of hemorrhage, gender, localization, normal neurological examination, AVM volume, middle cerebral artery feeding, varix in the venous drainage and the absence of deep venous drainage [17, 86, 98]. Englot et al. [86] identified deep artery perforators as an additional factor associated with post-operative seizures in a series of 416 patients. The univariate Cox regression analysis revealed a significant correlation between post-operative seizures and AVMs with deep artery perforators (Hazard ratio: 4.35) [86]. There was no observed connection between pre-operative seizure factors and post-operative seizure factors [86]. Ferlisi et al. [98] also found no relationship between pre-operative seizure factors and post-operative seizure factors.