Cortical and etiological determinants of stimulation‐induced seizures in children undergoing stereo‐EEG

Abstract

Objective

Electrical stimulation during stereo‐electroencephalography (SEEG) is used for functional mapping and seizure induction in epilepsy surgery candidates. However, its effects in pediatric populations remain underexplored. The aim of this study is to characterize stimulation‐induced seizures (SISs) in children undergoing SEEG and evaluate associations with stimulation parameters, cortical architecture, etiology, and surgical outcome.

Methods

We retrospectively analyzed 78 children with drug‐resistant focal epilepsy (median age at epilepsy onset=2.8 years ± SD=4.4) who underwent SEEG with low‐frequency (1–3 Hz) and/or high‐frequency (50 Hz) stimulation. SISs were categorized as low‐frequency seizures (LF‐ISs) or high‐frequency seizures (HF‐ISs) and analyzed by anatomic site, cytoarchitectonics, cortical hierarchy, and etiology. Post‐surgical outcomes were assessed using Engel classification.

Results

SISs occurred in 43.6% of patients (LF‐IS: 21.8%, HF‐IS: 32.1%). Focal cortical dysplasia type II (FCD II)—particularly FCD IIb—was significantly associated with SISs (LFS: p = .016; HFS: p = .001). Interictal SEEG biomarkers such as brushes and suppression predicted SISs (p = .002). LF‐ISs predominantly involved the orbitofrontal and insular regions, whereas HF‐ISs clustered in the cingulate and central areas (p = .029). Stimulation type correlated with cortical hierarchy (p = .005) but not cytoarchitecture. SISs were marginally associated with favorable surgical outcome (Engel class Ia; p = .046).

Significance

This is the first study to systematically assess SISs in a pediatric cohort. Findings suggest that SISs—especially in FCD II and heteromodal cortical areas—may serve as biomarkers of epileptogenicity and predictors of surgical outcome. These insights can guide SEEG planning and interpretation in pediatric epilepsy surgery.

Key points.

• Both low‐frequency stimulation (LFS) and high‐frequency stimulation (HFS) can induce typical electroclinical seizures in children.

• Focal cortical dysplasia type II (FCD II), particularly FCD IIb, is highly associated with stimulation‐induced seizures (SISs) from both LFS and HFS.

• Seizure induction varies by cortical region and hierarchical organization, with LFS linked to heteromodal and HFS to unimodal regions.

• SISs are marginally associated with favorable post‐surgical outcomes (Engel class Ia), suggesting prognostic value.

• The study integrates anatomic, etiological, and functional markers to refine SEEG‐guided surgical planning in children.

1. INTRODUCTION

Stereo‐electroencephalography (SEEG) was developed by Talairach and Bancaud at Hôpital Sainte Anne in Paris as a methodology utilized to delineate the epileptogenic zone (EZ). ¹ , ² This methodology is indicated for patients with drug‐resistant focal epilepsy who are candidates for surgical treatment but in whom non‐invasive diagnostic methods fail to localize the EZ accurately. Stereo‐EEG is particularly recommended in patients with focal epilepsy with discordant anatomo‐electro‐clinical findings, with complex malformation of cortical development, and with anatomic lesions close to eloquent cortical areas. ³ During SEEG monitoring, electrical stimulation is employed for functional mapping and seizure induction. ⁴ , ⁵ , ⁶ , ⁷ Stimulation is delivered between two close leads of each electrode, utilizing two primary modalities: high‐frequency stimulation (HFS, 50 Hz) and low‐frequency stimulation (LFS, 1 Hz). ⁷ , ⁸ , ⁹ Historically, HFS has been used for seizure induction, although it carries a significant risk of false positives, whereas LFS has been used primarily for mapping the cerebral cortex. ⁸ Over time, it has been demonstrated that LFS can also induce seizures, particularly in mesial temporal ⁸ and insular epilepsy, ¹⁰ as well as in focal cortical dysplasia type II (FCD II). ¹¹ Few studies have assessed the prognostic role of intracranial electrical stimulation in determining both the efficacy of stimulation and the long‐term surgical outcome. ⁹ , ¹⁰ , ¹² Induced seizures from both HFS ¹² and LFS ⁹ appeared to predict good surgical outcomes. As already established, the induced seizure could identify the EZ similarly to the spontaneous one, reducing the need to record it ¹² and the time of monitoring.

In the French guidelines the usual parameters for LFS and HFS are outlined, highlighting differences in pulse width, current, and stimulation duration. ⁵ Results from an international survey confirmed that there is variability across centers in stimulation parameters, particularly in intensity and pulse width, as well as in the assessment of patients and interpretation of stimulation results. ¹³ In addition, in Italian centers, the same difference in stimulation parameters exists. ³ Up to now, standardized practices are not available, and intracranial stimulation remains a time‐consuming procedure that requires the patient's cooperation.

To date no studies are available on intracranial stimulations focusing only on pediatric patients. The aim of this study is to investigate the effects of intracranial stimulation in pediatric patients undergoing SEEG at the Neurology, Epilepsy, and Movement Disorders Unit of Bambino Gesù Children's Hospital, focusing on the relationships between stimulation‐induced seizures (SISs) and age, prognosis, and etiology. Moreover, we delineated the potential role of cortical and etiological determinants in LF‐ISs (low frequency‐induced seizures) and HF‐ISs (high frequency‐induced seizures).

2. MATERIALS AND METHODS

We retrospectively analyzed 100 patients with drug‐resistant focal epilepsy, who underwent SEEG between January 2012 and April 2024. We included patients if: (i) age ≤ 18 years; (ii) typical focal seizures or epileptic spasms were recorded during SEEG monitoring; and with (iii) at least 6 months of follow‐up after surgery, including radiofrequency thermocoagulation (RF‐THC). Seventy‐eight patients fulfilled the inclusion criteria.

We collected information on gender, age at epilepsy onset, duration of epilepsy, etiology of epilepsy, brain magnetic resonance imaging (MRI), and age at the time of SEEG. We also gathered information on post‐surgical outcome as Engel classification. Seizures have been classified according to the International League Against Epilepsy (ILAE) classification. ¹⁴

All patients underwent presurgical assessment, including ictal video‐EEG recordings and brain MRI. Thirty‐one patients (39.7%) also underwent fluorodeoxyglucose–positron emission tomography (FDG‐PET) and 49 (62.8%) morphometric analysis program (MAP). Based on the latter data we designed SEEG planning. The technologies and methodologies employed have already been detailed. ³ Based on the SEEG findings, RF‐THC and/or surgery was proposed.

Etiology was defined based on both radiologic and histopathological findings according to ILAE consensus criteria ¹⁵ , ¹⁶ Cortical tuber in patients with tuberous sclerosis complex (TSC) was considered a subtype of FCD II, considering the similarities in histological, radiological, and molecular features. ¹⁷ More specifically, we included patients with FCD II–like features on brain MRI, such as increased cortical thickness, gray–white matter blurring, the transmantle sign, and calcification, ¹⁸ as well as T2‐hypointensity in the unmyelinated brain. ¹⁹ In these patients SEEG has a focal tuber organization. ²⁰ When histology was unavailable, we included patients in the group of FCD II if clinical, radiological, and specific SEEG patterns—both ictal and interictal—were detected, ¹¹ , ²¹ and we defined these patients in the group of FCD II pattern.

We systematically applied low‐frequency stimulation (or LFS) and/or high‐frequency stimulation (or HFS) between two adjacent contacts. LFS consisted of single bipolar pulses (pulse width 0.5 ms; current range 0.5–5 mA) delivered for 20–30 s at frequencies between 1 and 3 Hz. ³ We included 3 Hz in the LFS because it might correspond to the frequency of the sub‐continuous rhythmic spikes observed in FCD II ²¹ and because it was not considered an intermediate stimulation frequency. ²² , ²³ HFS was administered using a train of stimuli (frequency 50 Hz; pulse width 0.5 ms; current range 0.5–3 mA intensity) for 5 s. ³ Current intensity, pulse width (fixed at 0.5 ms), electrode contact surface, charge density (CD), and total CD per second (tCD/s) were recorded systematically for all stimulations and are detailed in Tables S1 and S2. Within each stimulation modality (LFS vs HFS), these parameters did not vary across anatomic regions. Stimulations at 1 Hz were applied at all contact points, whereas 3 Hz and 50 Hz stimulations were limited to gray matter. Particularly, we stimulated the pairs of leads selected in the synthetic montage, corresponding to the distinct anatomic structures sampled by each electrode. Additional lead pairs were stimulated when no seizure was elicited, and the presumed EZ was thought to be nearby. LFS was initiated at 3 mA. In the primary motor cortex, where clinical responses are expected at lower intensities, stimulations were initiated at 0.5 mA, with increments of 0.5 or 1 mA based on clinical response. In regions presumed to be close to the EZ, the initial intensity was set below 3 mA, adjusted based on the observed interictal activity. For HFS, stimulations were initiated at 1 mA and increased by 0.5/1‐mA increments if ineffective, up to a maximum of 3 mA. Induced seizure was defined as a typical electroclinical seizure with clinical semiology resembling spontaneous seizure or aura. ¹² Whenever an atypical seizure occurred, stimulation was interrupted until the trace returned to baseline and until the patient fully recovered. LFSs are performed on the second or third day after implantation, before any reduction in anti‐seizure medication (ASM). ³ HFSs are executed after spontaneous seizures have been recorded and ASM has returned to baseline levels. ³

We evaluated the potential relationships between stimulation‐induced seizures and age, prognosis, and etiology. To understand the influence of specific SEEG features on induced seizure in a subgroup of patients with FCD II, we also analyzed the electrographic pattern of interictal and ictal SEEG.

To discern differences between LFS and HFS and the possible role of cortical structures, each SIS was considered an independent variable. The overall goal was not to compare the absolute effectiveness of LFs vs HFs, but rather to identify which factors may better predict whether a seizure is more likely to be elicited by LFS or HFS at the lowest parameters. For both LF‐ISs and HF‐ISs, we retrieved stimulation parameters (frequency and current). As part of routine clinical practice, seizures induced are generally not re‐tested at higher stimulation intensities or frequencies. Because only a small subset of LFS‐ISs was subsequently re‐tested with HFS, these re‐tested cases were excluded from the analysis of LF‐ISs and HF‐ISs. This approach was adopted to avoid potential bias or errors arising from protocol modifications. The precise anatomic location of the contact pair was identified using 3D Slicer ²⁴ and then mapped onto Talairach's atlas. ²⁵ For each contact pairs, the cortical cytoarchitecture was classified according to Brodmann, ²⁶ and the cortical hierarchy was determined based on the classical Mesulam's classification. ²⁷ , ²⁸

2.1. Statistical analysis

Descriptive statistical analysis was computed to describe our cohort. Categorical variables were summarized using absolute frequencies and percentages, whereas continuous variables were described with all indices, including mean, standard deviation (SD), median, 1° and 3° interquartile range (IQR), and range min–max. Boxplots and bar plots were used for visual representation of the data. To assess the normality of continuous data, the Shapiro–Wilk test was performed.

For categorical variables, the chi‐square test was used to assess associations between categorical groups, and the Fisher's exact test was applied when frequencies were below 5, to ensure the validity of the results. For continuous variables, comparisons between independent groups were performed using the non‐parametric Mann–Whitney U test for two groups and the Kruskal–Wallis test for more than two groups, as these tests do not assume normality of the data. A significance level of 0.05 was used for all tests. All analyses are computed using Python and Jamovi 2.6.26.

3. RESULTS

Table 1 summarizes the demographic and clinical characteristics of the cohort. Median age at epilepsy onset was 2.8 years ±4.4 (0.7, 7.8), disease duration was 4.5 years ±4.2 (2.2, 8.0), and age at SEEG was 10.7 years ±5.5 (4.5, 15.3). LFS triggered seizures in 17 (21.8%), HFS in 25 (32.1%), and both types of stimulation (LFS and HFS) in 34 (43.6%) patients. Eight patients with seizures triggered by LFS also had HFS‐ISs. In 17 patients (21.8%), more than one seizure was induced by LFS (5 patients) and HFS (14 patients). No statistically significant differences were observed in the age of patients with SISs by LFS (10.3 ± 5.4 [3.9, 14.7]), HFS (10.9 ± 5.4 [5.3, 14.7]), or both (10.7 ± 5.4 [4.6, 14.6]). LFS (n = 5229) and HFS (n = 6342) were performed in all 78 patients (Table S1). Table S4 summarizes the demographic and clinical features of patients in whom seizures were not induced by stimulation. Eight atypical seizures (10.3%) were recorded, mostly induced by HFS (7/8), predominantly in the hippocampus (3/7; CD 9.95–19.89). A single atypical seizure was triggered by LFS (1 Hz) within the hippocampus (CD 29.84). A multivariate logistic regression analysis was performed both for SIS occurrence and for Engel outcome using the full set of clinical, imaging, histological, and surgical predictors. Full model results, including coefficients and confidence intervals (CIs), are reported in the Tables S6–S8. We also included these results in the Tables S1–S8.

TABLE 1.

Demographic and clinical features of patients with induced seizure by LFS, HFS, and LFS + HFS.

	N, [%]	LFS	HFS	LFS and HFS
		(n = 17)	(n = 25)	(n = 34)
Sex	n = 78
F	42 [53.8]
M	36 [46.2]
Type of seizures	n = 78	n = 17	n = 25	n = 34*
Focal	61 [78.2]	16 [94.1]	23 [92.0]	32 [94.1]
Spasms and focal	13 [16.7]	Only focal 1 [5.9]	Only focal 2 [8.0]	Only focal 2 [5.9]
Spasms	4 [5.1]	No	No	No
Brain MRI	n = 78	n = 17	n = 25	n = 34
FCD	38 [48.7]	10 [58.8]	18 [72.0]	22 [64.7]
FCD I	11 [28.9]	1 [10.0]	3 [16.7]	4 [18.2]
FCD II	23 [60.6]	8 [80.0]	14 [77.8]	17 [77.3]
TSC	4 [10.5]	1 [10.0]	1 [5.5]	1 [4.5]
PMG ± PNH	9 [11.5]	No	No	No
Ischemia	8 [10.3]	No	No	No
HS	3 [3.8]	1 [5.9]	1 [4.0]	2 [5.9]
LEAT	2 [2.6]	1 [5.9]	1 [4.0]	2 [5.9]
Normal	18 [23.1]	5 [29.4]	5 [20.0]	8 [23.5]
Surgery	n = 78	n = 17	n = 25	n = 34
Resective	46 [59.0]	10 [58.8]	15 [60.0]	22 [64.7]
RF‐THC	20 [25.6]	4 [23.5]	6 [24.0]	7 [20.6]
No	12 [15.4]	3 [17.7]	4 [16.0]	5 [14.7]
Etiology n = 60	n = 60	n = 12	n = 20	n = 27
FCD I	10 [16.8]	1 [8.3]	2 [10.0]	3 [11.1]
FCD II	26 [43.3]	9 [75.0]*	16 [80.0]*	20 [74.1]*
FCD IIa	9 [34.6]	4 [44.5]	4 [25.0]	7 [35.0]
FCD IIb	10 [38.5]	3 [33.3]	8 [50.0]	9 [45.0]
TSC	4 [15.4]	1 [11.1]	1 [6.3]	1 [5.0]
FCD II pattern	3 [11.5]	1 [11.1]	3 [18.7]	3 [15.0]
Gliosis (ischemia)	8 [13.3]	No	No	No
PMG ± PNH	8 [13.3]	No	No	No
PMG	3 [37.5]
PNH	4 [50.0]
PMG + PNH	1 [12.5]
Unknown	3 [5.0]	No	No	No
HS	3 [5.0]	1 [8.3]	1 [5.0]	2 [7.4]
LEAT	2 [3.3]	1 [8.3]	1 [5.0]	2 [7.4]
Engel	n = 66	n = 14	n = 21	n = 29*
Ia	24 [36.4]	8 [57.1]	10 [43.5]	15 [51.7]
I (b–d)	11 [16.7]	2 [14.3]	4 [19.0]	5 [17.2]
II–IV	31 [46.9]	4 [28.6]	6 [28.6]	7 [33.3]
Therapy	n = 66	n = 14	n = 21	n = 29
Ongoing	51 [77.3]	9 [64.3]	15 [71.4]	19 [65.5]
Stopped	15 [22.7]	5 [35.7]	6 [28.6]	10 [34.5]

Abbreviations: F, female; FCD, focal cortical dysplasia; HFS, high‐frequency stimulation; HS, hippocampal sclerosis; LEAT, long‐term epilepsy‐associated tumor; LFS, low‐frequency stimulation; M, male; brain MRI, magnetic resonance imaging; n, number; PMG, polymicrogyria; PNH, periventricular nodular heterotopia; TSC, tuberous sclerosis complex.

^* p < .05.

3.1. Type of induced seizures

Among the 78 patients, 61 (78.2%) experienced focal seizures, 4 (5.1%) had epileptic spasms only, and 13 (16.7%) exhibited both focal seizures and epileptic spasms. Electrical stimulation—both LFS and HFS—induced only focal seizures.

To evaluate whether the likelihood of SISs differed across clinical seizure types, we performed a chi‐square test comparing seizure induction (present/absent) across the three semiology groups (focal only; spasms only; focal + spasms). The association was statistically significant (χ² test, p = .004). In the focal‐only group, stimulation induced seizures in 32 of 61 patients (52.5%). Among patients with both focal seizures and epileptic spasms, stimulation elicited focal seizures in 2 of 13 patients, whereas 11 of 13 (84.6%) showed no response. Notably, none of the four patients with epileptic spasms‐only had stimulation‐induced seizures. These results indicate that seizure induction occurred predominantly in patients without epileptic spasms in their spontaneous seizure semiology, and that epileptic spasms could not be elicited by intracranial stimulation.

3.2. Focal cortical dysplasia

FCD II was the most prevalent etiology among patients with seizures induced by LFS (n = 9/12, 75.0%) and by HFS (n = 12/20, 80.0%). A statistically significant correlation was found between the presence of FCD II vs other etiologies and seizure induction by both LFS (p = .016) and HFS (p = .001), analyzed as separate binary variables, as reported in Figure 1.

FIGURE 1.

Association between etiology and stimulation‐induced seizures during stereo‐electroencephalography (SEEG). Bar plots show the proportion of patients with and without stimulation‐induced seizures (SISs) stratified by etiological diagnosis (focal cortical dysplasia type II [FCD II] vs other), separately for low‐frequency stimulation (LFS, left panel) and high‐frequency stimulation (HFS, right panel). For LFS, 75.0% of patients with SISs had FCD II compared to 32.6% in the non‐SIS group. For HFS, 80.0% of patients with SISs had FCD II vs 21.6% without SISs. Both stimulation types showed a significant association between seizure induction and FCD II histology (LFS: p = .016; HFS: p = .001), with FCD II demonstrating higher susceptibility to stimulation‐induced seizures. FCD, focal cortical dysplasia; HFS, high frequency stimulation; LFS, low frequency stimulation.

Considering patients with FCD II (n = 26, 43.3%), seizures were induced in 20 patients (74.1%) by both LFS and HFS, and 6 patients did not experience stimulation‐induced seizures. Among them, three had TSC, two had FCD IIa, and one had FCD IIb. Regarding FCD II subtypes, their distribution was relatively similar in seizures induced by LFS (4 FCD IIa vs 3 FCD IIb). However, in HFS‐induced seizures, FCD IIb was more prevalent (8 FCD IIb vs 4 FCD IIa).

The analysis of interictal SEEG activity revealed the absence of background rhythm in 22 patients (84.6%), sub‐continuous rhythmic spikes and poly‐spikes‐and‐waves in 21 (80.8%), and the presence of brushes and suppression of electrical activity in 19 (73.1%). During ictal discharges, the intensification of the interictal abnormalities, preceding the ictal onset, was observed in 25 (96.1%), whereas low‐voltage fast activity (LVFA) was seen in 20 patients (76.9%). Based on these features, 19 patients (73.1%) were classified as having an intralesional pattern. A significant correlation was found between induced seizures and the presence of brushes and suppression (p = .002) (Figure 2A), as well as the presence of intralesional patterns (p = .028). The only patient with brushes and suppression pattern in whom seizure induction was not possible had TSC (Figure 2B). Interictal brushes and suppression were absent in seven patients, and among them two experienced SIS (Figure 2C).

FIGURE 2.

Representative stereo‐electroencephalography (SEEG) recordings of stimulation‐induced seizures (SISs). (A) Patient with focal cortical dysplasia type IIb (FCD IIb). SEEG shows sub‐continuous rhythmic spike‐and‐waves with intermixed brushes followed by suppression (I). The brush‐suppression pattern becomes more prominent at seizure onset (II). During the post‐ictal phase, the background activity is flattened, and similar brushes reappear (III). A typical seizure was induced by low‐frequency stimulation (LFS) (IV). (B) Patient with tuberous sclerosis complex (TSC). The background activity is characterized by spike and some brushes followed by suppression (I). Interictal activity increases in rhythmicity prior to seizure onset, which is identified by a tonic discharge (II). During the post‐ictal phase, the background rhythm is similar to the interictal (III). No seizure was induced by stimulation. (C) Patient with FCD IIa. The background rhythm is characterized by slow waves with rare intermixed spikes (I). Spikes increase during the pre‐ictal phase, and the seizure onset is characterized by a rapid, not tonic, discharges (II). In the post‐ictal phase, spikes followed by slow waves are observed (III). A typical seizure was induced by high‐frequency stimulation (HFS) (IV).

3.3. Outcome

As reported in Table 1, 66 patients underwent intervention: 46 resective surgery and 20 RF‐THC. Resective surgery involved a single lobe in 30 patients (65.2%): frontal (n = 12), mesial temporal (n = 6), central (n = 5), parietal (n = 3), insula (n = 2), and occipital (n = 1). The remaining patients underwent more extensive resections: parieto‐occipital (n = 4), temporo‐parieto‐occipital (n = 4), temporo‐occipital (n = 1), and insular surgery extended to the anterior or posterior operculum or to the temporal lobe (n = 8). Regarding RF‐THC, the targeted structures were: frontal (n = 3), insula (n = 3), heterotopic nodule (n = 3), temporal (n = 3), central (n = 2), parietal (n = 2), fronto‐temporal (n = 1), mesial temporal (n = 1), occipital (n = 1), and insula plus adjacent regions (n = 1). The minimum follow‐up considered was 12 months, and the median follow‐up time was 4.4 years (± 4.1 years).

The occurrence of seizures induced by either LFS or HFS alone was not significantly associated with Engel class Ia (p > .05). However, when considering the two types of stimulations (LFS and HFS), a marginally significant association was found between induced seizure (29 patients with induced seizures vs 37 without) and surgical outcome, with a p‐value of 0.046. As illustrated in Figure 3, among patients with induced seizures, 15 had an Engel Ia outcome, compared to 9 in the group of patients with non‐induced seizures; for Engel Ib–Id, the distribution was 5 vs 6; and for Engel II–IV, 8 vs 18.

FIGURE 3.

Relationship between stimulation‐induced seizures (SISs) and post‐surgical outcome in pediatric stereo‐electroencephalography (SEEG). Bar plot showing the distribution of Engel surgical outcomes (Ia, Ib–d, II–IV) in patients with and without SISs during SEEG (combined low‐frequency stimulation [LFS] and high‐frequency stimulation [HFS]). Among patients with SISs, 55.2% achieved Engel class Ia outcome, compared to 29.7% in the non‐SIS group. SIS was marginally associated with favorable outcome (Engel Ia; p = .046), suggesting a potential prognostic value of seizure induction during electrical stimulation.

No significant association was found when comparing resective surgery with RF‐THC and Engel.

3.4. Seizures induced by LFS and HFS (LF‐ISs and HF‐ISs)

A total of 23 seizures were induced by LFS and 48 by HFS, with seizure induction efficiencies of 0.44% and 0.77%, respectively. Among LF‐ISs, the most effective stimulation frequency was 1 Hz, accounting for 17 seizures (73.9%), followed by 3 Hz (n = 6, 26.1%). The efficacy ratio for 1 Hz was 0.35% (17/4919) and for 3 Hz was 1.94% (6/310). According to our protocol, as detailed previously, we observed that the mean current was 2.85 ± 1.05 mA (median 3 mA; range 0.5–5.0). The CD was 28.33 ± 10.44 μC/cm² (median 28.33 μC/cm²; range 4.97–49.74) and tCD 46.49 ± 38.19 μC/cm²/s (median 29.84 μC/cm²/s; range 4.97–149.22). Considering HF‐IS, the mean current was 1.55 ± 0.75 mA (median 1 mA; range 0.5–3), the CD 15.44 ± 7.48 μC/cm² (median 9.95 μC/cm²; range 4.97–29.84) and tCD/s 771.91 ± 374.21 μC/cm²/s (median 497.50 μC/cm²/s; range 248.50–1492.00). A full quantitative description of stimulation parameters, including current intensity, CD, and tCD/s for both LFS and HFS, is provided in Tables S1–S3, confirming that within each modality these values were homogeneous across regions and therefore could not account for the observed anatomic clustering of LF‐ISs and HF‐ISs. Excluding all seizures for which the exact location could not be reconstructed in 3D Slicer and HF‐ISs already induced by LFS, we analyzed 46 of 54 SISs, particularly 22 LF‐ISs and 24 HF‐ISs. The precise anatomic location of the contact pair, the corresponding cortical cytoarchitecture, and the hierarchical organization of the cortex are reported in Table S5. In addition, we categorized the precise location into broader anatomic regions, as shown in Table S5. A statistically significant association was found between the type of stimulation (LF‐ISs vs HF‐ISs) and the anatomic region involved (p = .029). LF‐ISs occurred more frequently in the orbito‐frontal (22.7% vs 8.3%) and insula (18.3% vs 0%) regions, whereas HF‐ISs were more often associated with the cingulate gyrus (20.8% vs 0%) and central region (29.2% vs 13.6%). No significant correlation was observed between stimulation type and the cytoarchitectural organization of the cortex. In contrast, a significant association was revealed between stimulation type and the hierarchical organization of the cortex (p = .005). LF‐ISs were more frequent in heteromodal (40.9%), whereas HF‐ISs were more commonly associated with unimodal regions (41.7%) (Figure 4). Descriptive frequencies for each category are detailed in Table 2.

FIGURE 4.

Cortical distribution of stimulation‐induced seizures (SISs) according to hierarchical brain architecture. Three‐dimensional (3D) surface reconstructions of the pediatric brain showing cortical sites where seizures were induced by low‐frequency stimulation (LFS; cross markers) and high‐frequency stimulation (HFS; circle markers). Color code corresponds to Mesulam's cortical hierarchy: heteromodal (yellow), unimodal (blue), idiotypic (red), paralimbic (green), and limbic (olive). Low‐frequency induced seizures (LF‐ISs) occurred predominantly in heteromodal and paralimbic regions, whereas high‐frequency induced seizures (HF‐ISs) clustered in unimodal and idiotypic areas, consistent with observed statistical associations (p = .005).

TABLE 2.

Descriptive frequencies for each category related to localization.

	LF‐IS (n = 22)	HF‐IS (n = 24)
	n [%]	n [%]
Anatomy
Orbito‐frontal	5 [22.7]	2 [8.3]
Cingulate gyrus	0	5 [20.8]
Insula	4 [18.3]	0
Central	3 [13.6]	7 [29.2]
Mesial temporal	2 [9.1]	4 [16.7]
Temporal	3 [13.6]	4 [16.7]
Occipito‐parietal	5 [22.7]	2 [8.3]
Cytoarchitecture
Granular	7 [31.8]	5 [20.8]
Dysgranular	10 [45.5]	9 [37.5]
Agranular	3 [13.6]	7 [29.2]
Allocortex	2 [9.1]	3 [12.5]
Hierarchical organization
Idiotypic	1 [4.5]	3 [12.5]
Unimodal	4 [18.2]	10 [41.7]
Heteromodal	8 [36.4]	0
Paralimbic	8 [36.4]	8 [33.3]
Limbic	1 [4.5]	3 [12.5]

Abbreviations: HF‐IS, high frequency‐induced seizure; LF‐IS, low frequency‐induced seizure; n, number.

Finally, given the distinctive features of FCD II and its potential to confounding local cytoarchitecture, we also examined radiological and SEEG characteristics at the contact level. In patients with FCD II, the rates of SISs in radiologically lesional contacts were 91.7% in LF‐ISs (11/12) and 55.0% in HF‐ISs (11/20). In normal‐appearing gray matter, seizures were induced only once with LFS compared with nine times with HFS, demonstrating a significant association between stimulation type and radiological findings (p < .01). Analysis of contact‐level SEEG patterns showed that, in LF‐ISs, sub‐continuous activity intermixed with brushes and suppression was most common (n = 6, 50.0%), followed by brush–suppression (n = 4, 33.3%) and isolated sub‐continuous activity (n = 2, 16.7%). In HF‐ISs, the predominant pattern was again sub‐continuous activity intermixed with brushes and suppression (n = 8, 40.0%), followed by brush–suppression (n = 7, 35.0%) and sub‐continuous activity alone (n = 3, 15.0%); notably, normal SEEG activity was observed only in HF‐ISs (n = 2, 10%).

3.5. Multivariable predictors of stimulation‐induced seizures

To further investigate independent determinants of seizure inducibility, we performed a multivariable logistic regression analysis (Table 3). We found that the probability of belonging to the LFS + HFS group (vs non‐SISs) was neither explained by demographic variables (age: odds ratio [OR] 0.87, 95% CI 0.58–1.31; p = .512) nor by epilepsy duration (OR 1.09, 95% CI 0.69–1.71; p = .722) and was not driven by MRI positivity (MRI positive vs normal: OR 0.48, 95% CI 0.01–23.70; p = .712). Conversely, site‐ and disease‐related factors emerged as independent determinants: seizure semiology showed a strong association (e.g., spasms and focal vs focal: OR 326.15, 95% CI 3.54–30061.59; p = .012), and etiology was highly informative, with other vs FCD II yielding increased odds (OR 126.51, 95% CI 4.59–3488.36; p = .004), consistent with a substantially higher susceptibility of FCD II substrates to stimulation‐induced seizures. In addition, anatomic location of the stimulated site remained independently associated with seizure inducibility: compared with temporal sites (reference), parieto‐occipital (PO–T: OR 237.48, 95% CI 4.44–12690.48; p = .007) and central regions (central–T: OR 54.46, 95% CI 1.13–2628.36; p = .043) showed significantly different odds of SISs, whereas frontal sites showed a similar trend that did not reach significance (F–T: OR 14.67, 95% CI 0.60–356.42; p = .099).

TABLE 3.

Multivariable logistic regression comparing LFS + HFS vs non‐SISs.

	Estimate	SE	Z	p	Odds ratio	95% CI
						Inferior	Superior
Age	−0.1353	0.207	−0.65531	0.512	0.87342	0.58267	1.309
Epilepsy duration	0.0823	0.231	0.35599	0.722	1.08573	0.69032	1.708
MRI
Positive – normal	−0.7344	1.990	−0.36910	0.712	0.47977	0.00971	23.700
Type of seizure
s_f – focal	5.7873	2.308	2.50748	0.012	326.14600	3.53844	30061.590
Spasms – focal	16.5701	3577.285	0.00463	0.996	1.57e+7	0.00000	Inf
Surgery
THC – yes	−0.6827	2.057	−0.33193	0.740	0.50523	0.00897	28.464
No – yes
Etiology short
Other – FCD II	4.8403	1.692	2.86022	0.004	126.51301	4.58827	3488.358
Engel class
I(b–d) – Ia	0.9833	1.336	0.73590	0.462	2.67338	0.19482	36.684
II–IV – Ia	1.4078	1.390	1.01275	0.311	4.08699	0.26801	62.324
Surgery_site
F – T	2.6859	1.628	1.65013	0.099	14.67125	0.60392	356.417
PNH – T	21.2762	3643.136	0.00584	0.995	1.74e+9	0.00000	Inf
PO – T	5.4701	2.030	2.69474	0.007	237.47526	4.44384	12690.475
Central – T	3.9974	1.978	2.02103	0.043	54.45910	1.12838	2628.355

Note: The values in bold are statistically significant (p < 0.05).

Abbreviations: Central, including surgeries involving the insula; F, frontal; PNH, periventricular nodular heterotopia; PO, parieto‐occipital (including patients who underwent surgery in the posterior temporal cortex); s_f, spasms and focal seizures.

4. DISCUSSION

This is the first study to investigate stimulation‐induced seizures (or SISs) in an exclusively pediatric population. Our findings indicate that induced seizures could be influenced by different factors, as the type of stimulation, seizure semiology, and underlying etiology but also by intrinsic cortical features, including anatomic localization and hierarchical organization. Notably, this is the first report to evaluate intrinsic cortical characteristics as independent factors contributing to seizure susceptibility.

We successfully induced typical electro‐clinical seizures in 43.6% of patients using both LFS and HFS; seizures are more frequently induced with HFS (32.1%) than LFS (21.8%). Age did not appear to be correlated with the ability to induce seizures using LFS, HFS, or both. Our data confirm what was already reported in adult cohorts where the HFS rate ranges from 54.9% to 40.8% and LFS from 18.2% to 6.6%. ⁹ , ¹⁴ In our study, HFS resulted in a higher occurrence of seizures than LFS, although the difference was less prominent than in the adult population. This observation may likewise be influenced by the exclusion of HF‐IS that had already been elicited by LFS.

As a pediatric cohort, some of our patients also exhibited epileptic spasms, either alone or in combination with focal seizures. No data regarding induced epileptic spasms is available. In our study, we were unable to induce epileptic spasms, although we successfully induced focal seizures also in patients who experienced both types of seizures. This observation further supports the hypothesis that the pathophysiological mechanism underlying epileptic spasms is complex and involves interactions between subcortical structures and either focal or diffuse cortical abnormalities. ²⁹ Focal seizures were not further characterized; therefore, additional differences within this group may exist but were not evaluated.

We found a statistically significant correlation between both LFS‐ and/or HFS‐induced seizures and etiology, with FCD II being the most prevalent. Of interest, FCD IIa and IIb appeared to have the same distribution in seizures triggered by LFS, whereas FCD IIb seemed more sensitive to HFS. However, this observation is limited by the small number of cases analyzed. These findings were not reported by Cuello Oderiz et al., who did not find any correlation between FCD II and SISs. ¹² The ability of electrical stimulation to induce seizures in FCD II was already detailed by Chassoux et al.: among 27 patients, HFS induced 20 typical seizures, highlighting the effectiveness of this stimulation method. ³⁰ Seizures induced by LFS were also observed in patients with FCD II located in the central or parietal areas. ¹¹ The ability of LFS to induce seizures in FCD II suggests a low epileptogenic threshold, ³¹ especially considering that this type of stimulation was historically reported to be ineffective in inducing seizures outside the hippocampal formation. ³²

In our cohort of patients with FCD II, the presence of brushes and suppression, rather than LVFA, showed a statistically significant correlation with seizure induction. In these patients, seizures were induced mostly by HFS or by both types of stimulation. These findings were not reported previously, as only LVFA was investigated as a possible marker of high epileptogenicity in LFS‐induced seizures. ¹² , ³³ Two neurophysiological mechanisms have been proposed for burst suppression pattern. The cortical hypersensitivity hypothesis, described by Kroeger et al., assumes that excitation–inhibition balance is shifted toward excitation. ³⁴ The other main explanation, derived from mathematical modeling, suggests that reduced neural metabolic activity can generate burst‐suppression pattern. ³⁵ Ching et al. proposed that when ATP production falls below consumption demand, intracellular ATP levels decline, leading to activation of ATP‐sensitive potassium channels. ³⁶ This process triggers a suppression phase that persists until ATP concentrations return to baseline. Restoration of baseline ATP levels terminates the suppression and allows initiation of a new burst. During each burst, ATP is rapidly depleted, thereby triggering the next suppression phase. Looking at SEEG pattern (see also Figure 2A), when burst suppression precedes a seizure, the final burst is not followed by a suppression phase but by a tonic discharge, as if the recovery mechanism after a burst were inhibited. This observation could suggest increased cortical excitability, potentially linked both to the burst‐suppression pattern itself and to the intrinsic characteristics of FCD, and may explain the greater susceptibility to stimulation‐induced seizure.

In our study, we found a statistically significant, albeit marginal, correlation between seizures induced by LFS and HFS and Engel Ia. Seizure‐freedom rates were highest in patients with LFS‐induced seizures (57.1%), followed by LFS and HFS (51.7%) and HFS alone (43.5%), suggesting a trend where patients with LFS‐induced seizures may have a better post‐surgical outcome. These findings support the importance of seizures induced by cortical stimulation in predicting a good post‐surgical outcome, especially when LFS is obtained. ¹² In patients without induced seizures, it is possible that the electrodes did not sufficiently explore the EZ or that the cortical areas have a high activation threshold, making it difficult to trigger the epileptic network. ⁹

With the hypothesis that a single factor was not enough for seizure induction we evaluated if cortical site, architectonic organization of the brain, and hierarchical cortical organization may influence the occurrence of seizure induced by cortical stimulation. Previously it has been suggested that brain regions may have different activation thresholds, but this was not analyzed in depth. In our study we found a statistically significant association between the type of stimulation (LF‐IS vs HF‐IS) and the anatomic region involved (p = .029). LF‐ISs occurred more frequently in the orbito‐frontal (22.7% vs 8.3%) and insular (18.3% vs 0%) regions, whereas HF‐ISs were more often associated with the cingulate gyrus (20.8% vs 0%) and central region (29.2% vs 13.6%). A recent study confirmed that LFS can induce seizures not only in mesial temporal structures, as previously reported, ³² but also in the insula. ¹⁰ The susceptibility of the central region more to HFS is, on the other hand, is somewhat unexpected. In adults, the central region was traditionally reported as having a low excitability threshold. ⁵ , ¹¹ This different finding may be explained with the pediatric age of our study population, potentially reflecting age‐related differences in cortical excitability.

We also considered the intrinsic organization of the cortex, including both its structural (architectonic) characteristics ²⁶ and the functional connections. ²⁷ Although correlations exist between the hierarchical organization and the cytoarchitectonics of the cortex, ²⁷ , ²⁸ we found conflicting results with the two sub‐classifications. The cytoarchitectonics of the cortex (Brodmann areas) does not seem to influence SISs. No previous studies have considered this variable; a larger population may confirm these findings. On the other hand, we found a statistically significant association between LF‐ISs and heteromodal areas, as well as HF‐ISs and unimodal areas. We observed, also, that LF‐IS is more frequent in the insula (olfactocentric subdivision of the paralimbic belt), whereas HF‐IS is more commonly found in the cingulate gyrus (hippocampocentric subdivision of paralimbic belt). We did not find a difference in the limbic sub‐structures. It appears that LFS is more likely to induce seizures in heteromodal areas, probably because the complex organization of these structures requires a lower frequency but longer‐lasting stimulus to activate the network. Unimodal areas appear to respond more effectively to HFS, likely because they have a simpler network. The limbic system was studied extensively and was postulated to be more sensible to LFS ³² ; this result was not confirmed in our population and it may be explained by the different etiology between adults and children (hippocampal sclerosis vs cortical dysplasia), and by the length of the disease, again highlighting the potential multifactorial nature of responses to stimulation. In our cohort, only three patients had hippocampal sclerosis. The overall number of atypical seizures induced by hippocampal stimulation is also lower than what is reported in adults, suggesting age‐dependent modulation of intrinsic threshold. Regarding the central regions, we observed three LF‐ISs elicited by CDs ranging from 29.84 to 39.79 μC/cm². Lower CDs (9.95–19.89 μC/cm²) induced seven HF‐IS events. In our cohort, as in our clinical practice, children do not appear to require higher current intensities but rather higher stimulation frequencies. Indeed, children often require HFS, even in regions that exhibit low intrinsic thresholds and that, in adults, typically respond to LFS. ⁷ , ³² Under our stimulation parameters, the currents, and consequently the CDs, delivered during HFS were significantly lower than those used for LFS, and remained below the established safe upper limit of charge density to protect against injuries. ¹³ The reasons that these regions in children are more easily activated by HFS are likely multifactorial, involving age, degree of myelination, and underlying etiology. These hypotheses, however, need confirmation by direct comparison. Finally, we sought to explore independent determinants of seizure inducibility through a multivariable logistic regression analysis. The results indicate that seizure induction is determined predominantly by intrinsic characteristics of the stimulated cortical substrate rather than by nonspecific stimulation parameters.

4.1. Limitations

The main limitations of this study are its retrospective nature, moderate sample size, and variability in stimulation parameters. These limitations were reduced by a standardized stimulation protocol during the study period and by the same investigators performing stimulations for all patients studied. To reduce stimulation variability, we excluded from the analysis the HFS delivered on contacts that elicited seizures during LFS. This approach, however, decreases the apparent seizure‐induction efficacy of HFS. The lack of standardized repetition of HFS that triggered seizures during LFS makes it impossible to fully compare the two groups, which is not the primary aim of our study. Moreover, in this study we did not report all current intensities, CDs, and the exact topography of stimulations that did not elicit seizures.

This restricted our multivariate analysis and prevented us from fully accounting the rate of stimulations in each region, as well as the location of underlying lesions. We addressed this partially for FCD II. Finally, the absence of an adult population represents a limitation, as it prevents the identification of potential differences between the children and adults.

5. CONCLUSION

This is the first study examining stimulation‐induced seizures (or SISs) in a pediatric population. We confirm that induced seizures are associated with both good post‐surgical outcomes and with FCD II. Given the high sensitivity of FCD II to stimulation, we identified brushes and suppression as potential biomarkers of epileptogenicity. In addition, induced seizures appear to be influenced by anatomic location, particularly by the hierarchical organization of the cortex. Epileptic spasms were not induced by cortical stimulation, confirming the distinct and more extensive network associated with this seizure type. Further studies are needed to confirm these findings, to have a wider comprehensive comparison across stimulation types, and to enable to compare pediatric with adult population, considering the age differences in myelination and connectivity.

AUTHOR CONTRIBUTIONS

Daniela Chiarello: conceptualization, data curation, formal analysis, investigation, methodology, writing – original draft preparation, and writing—review and editing. Mattia Mercier: data curation, formal analysis, investigation, methodology, writing – original draft preparation, and writing—review and editing. Giusy Carfì‐Pavia: data curation andinvestigation. Concetta Luisi: investigation and conceptualization. Chiara Pepi: investigation. Alessandro De Benedictis: investigation. Carlo Efisio Marras: investigation. Nicola Specchio: conceptualization, methodology, and writing—review and editing. Luca de Palma: conceptualization, methodology, investigation, validation, and writing—review and editing.

FUNDING INFORMATION

This study was supported by #NEXTGENERATIONEU (NGEU) and funded by the Ministry of University and Research (MUR), National Recovery and Resilience Plan (NRRP), project MNESYS (PE0000006)A Multiscale integrated approach to the study of the nervous system in health and disease (DN. 1553 11.10.2022). This study was also funded by Ministry of Health with Current Research Funds.

CONFLICT OF INTEREST STATEMENT

N.S. has served on scientific advisory boards for GW Pharma, BioMarin, Arvelle, Marinus, and Takeda; has received speaker honoraria from Eisai, Biomarin, Livanova, and Sanofi; and has served as an investigator for Zogenix, Marinus, Biomarin, UCB, Roche. the remaining authors have no conflicts of interest.

ETHICS STATEMENT

We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Supporting information

Figure S1.

Table S1.

DATA AVAILABILITY STATEMENT

The authors confirm that the data supporting the findings of this study are available within the article and its supplementary material.