Ictal central apnea is predictive of mesial temporal seizure onsets: an intracranial investigation

Abstract

Ictal central apnea (ICA) is a semiological sign of focal epilepsy, associated with temporal and frontal lobe seizures. In this study, using qualitative and quantitative approaches, we aimed to assess the localizational value of ICA. We also aimed to compare ICA clinical utility in relation to other seizure semiological features of focal epilepsy. We analyzed seizures in patients with medically refractory focal epilepsy undergoing intracranial stereotactic electroencephalography (SEEG) evaluations with simultaneous multimodal cardiorespiratory monitoring. A total of 179 seizures in 72 patients with reliable artifact-free respiratory signal were analyzed. ICA was seen in 55/179 (30.7%) seizures. Presence of ICA predicted a mesial temporal seizure onset compared to those without ICA with an OR=3.8, 95% CI [1.3, 11.6] (p-value = 0.01). ICA specificity was 0.82. ICA onset was correlated with increased high frequency broadband gamma (60–150 Hz) activity in specific mesial or basal temporal regions, including amygdala, hippocampus, fusiform and lingual gyri. Based on our results, ICA has an almost four-fold greater association with mesial temporal seizure onset zones, compared to those without ICA, and is highly specific for mesial temporal seizure onset zones. As evidence of symptomatogenic areas, onset-synchronous increase in high gamma activity in mesial or basal temporal structures was seen in early onset ICA, likely representing anatomical substrates for ICA generation. ICA recognition may help anatomo-electro-clinical localization of clinical seizure onset to specific mesial and basal temporal brain regions, and the inclusion of these regions in SEEG evaluations may help accurately pinpoint seizure onset zones for resection.

INTRODUCTION

Ictal central apnea (ICA) is a seizure semiological sign seen in patients with focal epilepsy. It is frequently seen in temporal lobe epilepsy ^1–3, where ICA is commonly the first ictal clinical sign². In a small series of mesial temporal lobe seizures, simultaneous use of respiratory monitoring revealed ICA in 70% of seizures ². Seizure semiological signs, especially early ictal clinical signs, are crucial to localize the epileptogenic focus for epilepsy surgery. Early ICA may indicate seizure origin in, or in close proximity to cortical regions involved in breathing modulation. The clinical significance and utility of ICA as a semiological sign in focal seizures is unknown. In this study, we aimed to define the clinical value of ICA and to compare it to other ictal semiological signs in epilepsy surgery evaluations. In addition, we aimed to identify cortical regions directly correlated with ICA onset with specific quantitative spatiotemporal SEEG signal features. ICA recognition may help anatomo-electro-clinical localization of clinical seizure onset to specific brain regions, and the inclusion of these regions in SEEG evaluations may help accurately pinpoint seizure onset zones for resection.

METHODS

Experimental design.

We prospectively enrolled 85 consecutive persons with epilepsy (PwE), undergoing standard of care stereotactic electroencephalography (SEEG) evaluations in the Epilepsy Monitoring Unit (EMU) from June 2016 to April 2023 after they provided informed consent under a study approved by the local committee on the protection of human subjects (HSC-MS-19–0010). Intracerebral electrodes (8 to 25 depth electrodes; PMT Corporation, Chanhassen, MN, USA; platinum contact, diameter=0.8 mm, inter-contact spacing = 2.75– 3.75 mm) were implanted using stereotactic robotic guidance (ROSA robot -Zimmer Biomet, Warsaw IN)⁴ stereotactically in various cortical areas depending on the implantation hypothesis of the putative epileptogenic zone⁵. Cortical reconstruction and volumetric segmentation of T1 structural magnetic resonance imaging (MRI) scans were performed using the FreeSurfer image analysis suite (http://surfer.nmr.mgh.harvard.edu). Electrode localization and three-dimensional (3D) reconstructions were performed using 3D slicer (http://www.slicer.org). Patient characteristics compiled included gender, age at the time of EMU admission, electroclinical epilepsy and seizure phenotype, and cardiorespiratory comorbidities. Seizure semiological signs were classified according to the ILAE seizure classification.⁶ Seizure onset zone (the cortical area from where seizures start), and epileptogenic zone (the region resected to produce seizure freedom⁷), type of surgery and outcomes (Engel classification)⁸ were collected.

Multimodal cardio-respiratory and video SEEG monitoring⁹.

Real-time cardiorespiratory data and video SEEG were obtained simultaneously. SEEG and electrocardiogram (EKG) were obtained using the EEG-1200 recording system (Nihon-Kohden, Tokyo, Japan). Thoracic and abdominal excursions were monitored using respiratory inductance plethysmography (Piezo Effort Belt Systems, Cadwell, Kennewick, USA). Peripheral capillary oxygen saturation (SpO₂) and heart rate (HR) were monitored using pulse oximetry (Nihon-Kohden, Tokyo, Japan). Electrical cardiac activity was monitored using a four-channel EKG (E501RASR Monitoring Electrode RT Snap repositionable, Cardinal Health, Ohio, USA). ICA was defined as a cessation of breathing for at least 5 seconds ^{2, 3}, confirmed by thoracic and abdominal signals (flattening of respiratory excursions and with visible cardio ballistic pulse artifact ⁹) in the absence of generalized tonic or clonic movements, since such movements invariably produced movement artifact in breathing channels. ICA was classified as early-onset ICA (≤5 seconds after SEEG onset) and late-onset ICA (>5 seconds after SEEG onset). ICA occurrence and onset were identified by two expert physicians (N.L and S.L). ICA features including duration, and relationships to ictal EEG and clinical onsets, were collected.

Ictal Central Apnea onset data processing.

We used Brainstorm suite (MATLAB Mathworks) ¹⁰ and an in-house MATLAB tool to post-process SEEG recording data in seizures with ICA to identify brain areas with high frequency broadband gamma changes coincident (60–150 Hz) at ICA onset¹¹. We analyzed brain regions consistently sampled across all seizures/patients, including amygdala, hippocampus, fusiform and lingual gyri, orbitofrontal, temporal pole and insula, frontal and parietal cortices. Grey matter electrode contacts most representative of the brain region studied were selected and bipolar montages were applied ¹¹. SEEG data were separated into two groups (early-onset ICA and late-onset ICA). SEEG signal at ICA onset was compared to pre-ICA baseline. Baseline normalization for ICA onset was relative to a pre-ICA baseline window comprising a full respiratory cycle (inspiration and expiration). The ICA window length analyzed was five seconds before and after ICA onset, for a total of ten seconds. Baseline and ICA windowed neural data were transformed into the time-frequency domain using a Morse wavelet transform¹². A quantitative analysis was performed via a z transform base normalization between the time-frequency power plots from the baseline and the apnea window. This type of normalization was defined as follows:

$$z_{tf}=\frac{Apne{a}_{tf}-\overline{Baselin{e}_f}}{\sqrt{n^{-1}{\sum }_{i=1}^n{\left(Baselin{e}_{if}-\overline{Baselin{e}_f}\right)}^2}}$$

where $t$ represents the time points, $f$ the frequency range and $n$ the number of time points in the baseline. This type of normalization enables correction of the $1/f$ power scaling intrinsic to intracranial SEEG signals and also enables reasonable comparisons across subjects and trials relatively unbiased by the individual variations in baseline power ¹³. Normalized time-frequency maps were then averaged across each group (early and late-onset ICA).

Statistical analysis.

Descriptive statistics (mean, standard deviation [SD], median, interquartile range, and frequency distribution) were used to summarize participants’ demographic, epilepsy, follow-up, seizure outcome and ICA data. We analyzed predictive associations with the seizure onset zone outcome among all patients, and a subgroup analysis was performed on patients with an Engel 1 outcome at follow-up was performed for epileptogenic zone determination. Predictive factors were entered as a single predictor using separate simple logistic regression models. For semiological signs that were statistically significant in simple logistic regression, we also assessed their predictive power using various metrics including sensitivity, specificity, negative predictive value, positive predictive value, and accuracy. To assess co-occurrence effects of ICA (presence and characteristics) and one other semiological sign, statistically significant semiological signs were also analyzed as a second predictor to the ICA presence logistic regression model using multiple logistic regression. Due to sample size and statistical power, co-occurrence of ICA and multiple signs was not considered. Fisher’s exact test was used to test the association between ICA presence and MTS. Two-sided p-values < 0.05 were considered as statistically significant. Due to the discovery nature of this study, multiple corrections were not performed on reported p-values. All predictive analyses were performed using R 4.2.2¹⁴. For quantitative analysis of gamma activity at ICA onset, a z-score of ≥4 was considered significant. By choosing a z-score of ≥4 we followed the Chebyshev inequality to guarantee significance of our results.

RESULTS

Eighty-five adult patients were enrolled and 220 seizures were prospectively collected. Thirty-nine seizures in 13 patients were excluded for further analysis due to unreliable or artifactual breathing signals. A total of 179 seizures in 72 patients (35 [48.6%] females, mean age 35.8 [18–63]) years with a total of 1,196 intracranial depth electrodes (mean number of electrodes per patient 16.6 ± 3.0) were analyzed. Patients and epilepsy characteristics are summarized in Table 1. Collected seizure semiological signs for analysis are shown in supplemental material table 1. The first clinical sign was focal onset aware (FOA) (aura) in 73/179 (40.8%), ICA in 49/179 (27.4%), focal onset impaired awareness (FOIA) non-motor onset (dialepsis) in 24/179 (13.4%), FOIA or FOA motor onset tonic in 14/179 (7.8%), FOIA/ FOA motor onset with automatisms in 11/179 (6.1%), gelastic in 3/179 (1.7%), aphasia in 3/179 (1.7%), and FOIA or FOA motor onset hyperkinetic in 2/179 (1.1%). Fifty-nine patients (81.9%) underwent epilepsy surgery, including resective surgery or laser interstitial thermal therapy (LITT) in 51/59 (70.8%) and neuromodulation treatments (RNS-responsive neurostimulation or DBS-deep brain stimulation) in eight (11.1%). In the group of patients who underwent resective surgery or LITT, 31/51 (60.8%) had Engel class I, 6/51 (11.8%) had II, 8/51 (15.7%) had III, and 5/51 (9.8%) had IV outcomes. Of the 31 patients with Engel class I, 30 had at least six months follow up (mean follow up of 22.8 months ± SD 13.1 [6–50]) and one patient did not. All seizures were classified according to seizure onset zones (table 1). For epileptogenic zone classification, only patients who underwent resective surgery or LITT and had outcome Engel class I with a minimum follow up of six months were classified. Epileptogenic zones were mesial temporal in 10/30 (33.3%) patients, temporal pole and anterior basal temporal in 1/30 (3.3%), posterior basal temporal in 1/30 (3.3%), insula 3/30 (9%), frontal in 3/30 (9%), occipital in 5/30 (1.7%), parietal in 6/30 (20%), and central in 1/30 (3.3%).

Table 1.

Patient and epilepsy characteristics.

Age, years	35.8 ± SD 12.4 (18–63)
Male, n (%)	37/72 (52.4%)
Age at epilepsy onset, years	18.0 ± SD 12.4 (1–57)
Epilepsy duration, years	17.8 ± SD 12.6 (0–59)
Cardiorespiratory comorbidities, n (%)	Obstructive sleep apnea: 8/72 (11.1%)
	Hypertension: 10/72 (13.9%)
	Asthma: 5/72 (6.9%)
Epilepsy etiology, n (%)	Structural: 50/72 (69.4%)-mesial temporal sclerosis 13/72 (18.1%)
	Unknown: 17/72 (23.6%)
	Infectious: 4/72 (5.6 %)
	Genetic: 1/72 (1.4%)
Seizure onset zone, n (%)	Mesial temporal in 81/179 (45.3%)
	Frontal in 25/179 (14.0%)
	Insula in 24/179 (13.4%)
	Occipital in 14/179 (7.8%)
	Parietal in 11/179 (6.1%)
	Temporal pole and anterior basotemporal in 10/179 (5.6%)
	Posterior basotemporal in 7/179 (3.9%)
	Lateral temporal in 6/179 (3.3%)
	Central in 1/179 (0.6%)

Abbreviations: %: percentage, s: seconds, n: number, SD: standard deviation.

Ictal central apnea presence and characteristics (table 2).

Table 2.

Patient and epilepsy characteristics in seizures with ICA

Age, years	33.3 ± SD deviation (SD) 12.8
Female, n (%)	13/ 23 patients (56.5%)
Ictal Central Apnea duration, s	26.4 ± 19.1
Time from SEEG onset to ICA onset, s	6.4 ± 9.5
Time from ICA onset to clinical onset, s	0.6 ± SD 3.5
Cardiorespiratory comorbidities, n (%)	Obstructive sleep apnea: 1/23 (4.4%)
	Hypertension: 0/23 (0%)
	Asthma: 1/23 (4.3%)
Epilepsy etiology, n (%)	Structural: 15/23 (65.2%)-mesial temporal sclerosis 7/23 (30.4%)
	Unknown: 7/23 (30.4%)
	Infectious: 1/23 (4.4%)
Seizure onset zones with ICA, n (%)	Mesial temporal in 35/81 (43.2%)  • Amygdala involvement at ICA onset 30/30 (100%)  • Hippocampus involvement at ICA onset 28/30 (93.3%)
	Frontal in 3/25 (12%)
	Insula in 0/24 (0%)
	Occipital in 6/14 (42.8%)
	Parietal in 3/11 (27.3%)
	Temporal pole and anterior basotemporal in 3/10 (30%)
	Posterior basotemporal in 5/7 (71.4%)
	Lateral temporal in 0/6 (0%)
	Central in 0/1 (0%)
Epileptogenic zones with ICA, n (%)	Mesial temporal in 16/34 (47.1%)
	Frontal in 1/8 (12.5%)
	Insula in 0/8 (0%)
	Occipital in 2/7 (28.6%)
	Parietal in 1/6 (16.7%)
	Temporal pole and anterior basotemporal in 1/4 (25%)
	Posterior basotemporal in 2/2 (100%)
	Lateral temporal in 0/3 (0%)
	Central in 0/1 (0%)

Abbreviations: %: percentage, n: number, s: seconds, SD: standard deviation.

Fifty-five out of 179 (30.7%) seizures in 23/72 (31.5%) patients had ICA. ICA was the first clinical feature in 49/55 (89.1%). In the remaining 6/55 (10.9%) seizures, aura was the first clinical symptom. ICA was the only clinical sign in 10/179 (5.6%) of all seizures and in 5/81 (6.8%) mesial temporal seizures. In 10/55 (18.2%) seizures with ICA, ICA was the only clinical sign. Some examples of a SEEG recorded seizures showing ictal EEG onset and ICA onset are shown in figures 1 (late-onset ICA) and figure 2 (early-onset ICA). In this cohort of 23 patients with ICA, 13/23 (56.5%) underwent resective surgery or LITT and 4/23 (17.4%) neuromodulation. Out of 13 patients who underwent resective surgery or LITT, 8/13 (61.5%) had Engel I, 1/13 (7.7%) had II, 2/13 (15.4%) had III and 1/13 (7.7%) had IV outcomes, and the remaining one patient was lost to follow-up. The relationship between ICA presence and seizure onset zones and epileptogenic zones are shown in figure 3.

Figure 1.

A) Example of a stereo-electroencephalographically recorded seizure showing ictal EEG onset and late-onset ictal central apnea (ICA) onset-complete cessation of breathing movements (THOR and ABD signals) replaced by cardioballistic artifact. Patient 35, seizure 1, right implantation; bipolar montage; high-frequency filter at 300 Hz; TC 0.03Hz. Note the right anterobasal low-amplitude fast activity at EEG onset (electrode ABT), followed by rapid spread to the temporal pole (electrode TP) and the emergence of ICA after mesial temporal (amygdala, anterior and posterior hippocampus) involvement (electrodes AH, PH and AMY) 21 seconds after ictal EEG onset B) A schema showing the electrode placement from the lateral view. C) T1 magnetic resonance imaging sagittal slice showing the placement of the mesial temporal electrodes.</P/>EKG: electrocardiogram, THOR: thoracic belt, ABD: abdominal belt.

Figure 2.

A) Example of a stereo-electroencephalographically recorded seizure showing ictal EEG onset and early-onset ictal central apnea (ICA) onset-complete cessation of respiratory movements. Patient 25, seizure 1, bilateral implantation; bipolar montage; high-frequency filter at 300 Hz; TC 0.03Hz). Note the left mesial temporal low-amplitude fast activity at EEG onset (electrodes LAH-left anterior hippocampus, LPH-left posterior hippocampus LAMY-left amygdala) and the emergence of ICA within two seconds after ictal EEG onset. B) A schema showing the electrode placement from axial view. C) T1 magnetic resonance imaging axial slice showing the placement of the left mesial temporal electrodes. </P/>EKG: electrocardiogram, ABD: abdominal belt.

Figure 3.

These figures represent percentages of seizures with ICA classified based on seizure onset zones (A) and epileptogenic zones (B). Legend: Abbreviations: ICA=Ictal central apnea.

Visual analysis of SEEG data (table 3).

Table 3.

Analysis of ICA and other seizure semiological signs predicting mesial temporal seizure onset.

	OR, 95%CI (p value)	sensitivity	specificity	PPV	NPV
ICA	3.8 [1.3, 11.6] p=0.01	0.45	0.82	0.68	0.64
FOIA non-motor onset	6.2 [2.1, 20.0] p= 0.001	0.60	0.77	0.69	0.70
FOIA/ FOA motor onset with automatisms	3.0 [1.1, 9.2] p= 0.04	0.45	0.79	0.65	0.63
ICA + FOIA non-motor onset	30.0 [5.5, 221.7] p < 0.001	0.85	0.64	0.67	0.83
ICA + FOIA/ FOA motor onset with automatisms	10.9 [2.49, 57.15] p= 0.001	0.69	0.66	0.63	0.72

ICA: ictal central apnea, FOIA: focal onset impaired awareness, FOA: focal onset aware, PPV: positive predictive value, NPV: negative predictive value

ICA predicted mesial temporal seizure onset with an OR=3.8, 95% CI [1.3, 11.6] (p-value = 0.01). ICA sensitivity was 0.45, specificity 0.82, with positive predictive value (PPV) 0.68, negative predictive value (NPPV) 0.64 (table 3). In addition to ICA, there were two other semiological signs associated with mesial temporal seizure onset, including FOIA and FOIA/ FOA motor onset with automatisms (table 3). The co-occurrence of ICA and FOIA or FOIA/ FOA motor onset with automatisms increased the odds ratio of mesial temporal seizure (table 3). When we examined the occurrence of ICA with either FOIA or FOIA/ FOA motor onset with automatisms, the specificity of ICA increased up to 0.95 and sensitivity decreased down to 0.21. ICA relationship with mesial temporal epileptogenic zone was not statistically significant (OR=2.7, 95% CI [0.5, 15.0], p-value = 0.25). Presence of MTS was associated with both, mesial temporal seizure onset (OR=24.70, 95% CI [4.44, 464.99], p-value = 0.003) and mesial temporal epileptogenic zone (OR=19.00, 95% CI [2.39, 412.02], p-value = 0.01). The only semiological sign associated with mesial temporal epileptogenic zone was FOIA (OR=14.79, 95% CI [2.47, 134.54], p-value = 0.006). There was no association between ICA features, including ICA total duration, or time from ICA onset to next clinical seizure sign. No significant association was found between time interval between SEEG seizure onset and ICA onset, and mesial temporal seizure onset zone (p>0.99). Similarly, no association was found between time interval from SEEG seizure onset to ICA onset, and mesial temporal epileptogenic zone (p=0.43), or between mesial temporal sclerosis (MTS) and ICA presence (p-value = 0.34).

Quantitative Analysis of SEEG Data (Fig 4).

Figure 4.

Quantitative analysis of breathing correlated brain regions. Analysis was carried out with the purpose of detecting high gamma band activity (60 to 150 Hz). We used base normalization between baseline and ICA windows to observe significant neural activity (Z score ≥4, “y” axis). The upper panel represents the group with early-onset ICA, and the lower panel represents the group with late-onset ICA. The “x” axis represents the time (in seconds) in reference to ICA onset. In the early-onset ICA group (27 seizures in 13 patients), significant high gamma activity compared to baseline (z score >4, p<0.05) was identified in amygdala, hippocampus, posterior fusiform and lingual gyri at ICA onset. In the group with late-onset ICA (12 seizures in seven patients), new increased high gamma activity was identified in hippocampus, anterior fusiform, temporal pole, lingual, and frontal opercular cortex (z score >4, p<0.05) at ICA onset. The latter had increased gamma activity before ICA onset in 8/12 seizures as well as at SEEG onset, and hence subtraction of baseline from SEEG onset led to the appearance of an apparent absence of high gamma activity. In 2/12 seizures amygdala was not sampled.

In a subset of 20 patients (39 seizures with ICA) with SEEG implantation structures that included all regions of breathing interest, grouped quantitative analysis of high gamma activity was performed. SEEG targeted regions included amygdala, hippocampus, anterior and posterior fusiform and lingual gyri, orbitofrontal, temporal pole and insula, frontal and parietal cortices. In the early-onset ICA group (27 seizures in 13 patients), significant high gamma activity compared to baseline (z score ≥4, p<0.05) was identified in amygdala, hippocampus, posterior fusiform and lingual gyri at ICA onset. Regions that did not show increased high gamma activity were temporal pole, anterior fusiform, orbitofrontal, anterior cingulate, anterior insula, frontal opercular and inferior parietal cortices. In the group with late-onset ICA (12 seizures in seven patients), new increased high gamma activity was identified in hippocampus, anterior fusiform, temporal pole, lingual, and frontal opercular cortex (z score ≥4, p<0.05) at ICA onset. Regions that did not show any significant new onset activation were the posterior fusiform, orbitofrontal, anterior cingulate, anterior insula, inferior parietal cortex or amygdala.

DISCUSSION

We have shown that seizures with ICA have an almost four-fold greater association with mesial temporal seizure onset zones, compared to those without ICA, and thus ICA is highly specific for mesial temporal seizure onset zones. If ICA is the first clinical feature, then followed by FOIA or FOIA/ FOA motor onset with automatisms, it further strengthens the odds and specificity for mesial temporal onsets. As evidence of symptomatogenic areas, onset-synchronous increase in high gamma activity in mesial or basal temporal structures (amygdala, hippocampus, fusiform and lingual gyrus), was seen in early onset ICA, likely representing anatomical substrates for ICA generation.

The localizing value of ictal central apnea.

One of our most striking findings was the localizing value of ICA for mesial temporal seizure onset zones, suggesting that ICA is an important semiological sign in the presurgical assessment of focal epilepsy. Prior studies have alluded to this association between ICA and temporal lobe seizures, although it is rarely looked for in semiological analyses ^{1, 3, 15–17}. ICA has high specificity for mesial temporal seizure onset zones (0.82), and has relatively low sensitivity (0.45). Thus, if ICA is identified in video EEG evaluations, a mesial temporal seizure onset zone should be highly suspected. On the other hand, lack of ICA does not rule out a mesial temporal onset. We were not able to establish differences between mesial and lateral temporal lobe seizures as far as ICA frequency, duration, or correlation with EEG and clinical onsets are concerned, probably due to a small sample size of lateral temporal seizures ¹⁷. However, in our study, no lateral temporal or insular seizure onsets had ICA. Thus, ICA may provide more value than abdominal auras for example, which though commonly seen in mesial temporal lobe seizures, are not specific to mesial temporal lobe epilepsy and have been reported in electrical stimulations of extratemporal areas, as well as temporo-limbic regions^{18, 19}. ICA incidence in our study was similar to previous reports of 36–44%^{1, 3, 15} of seizures in focal epilepsy and ICA duration (average = 26 seconds) was also similar^{1–3, 16, 17}. ICA was the only clinical sign in 5.5% of all seizures and in 6.8% mesial temporal seizures, which is consistent with previous studies (2, 3, 15). Thus, ICA recognition may prove crucial for identification of the true region of clinical seizure onset in SEEG evaluations. Currently, intracranial SEEG demonstration of the seizure onset zone (the cortical areas from where seizures start) offers a well-accepted approximation of the epileptogenic zone²⁰. However, we used a more strict definition of epileptogenic zone as the minimum amount of cortex that was resected to produce seizure freedom⁷. ICA association with a mesial temporal epileptogenic zone trended towards statistically significant results, probably due to the smaller sample size of this cohort (n=30 patients). This relationship is worth exploring in a larger cohort of patients.

Seizure evolution features further increase odds of mesial temporal ictal onset zones.

ICA followed by FOIA was associated with a 30-fold increase in the odds of mesial temporal seizure onsets while ICA followed by FOIA/ FOA motor onset with automatisms was associated with a 10-fold increase. The co-occurrence of ICA and FOIA and/or FOIA/ FOA motor onset with automatisms increased specificity for mesial temporal seizure onset to 0.95, and decreased sensitivity to 0.21, suggesting that if ICA is identified in video EEG evaluations followed by FOIA or FOIA/ FOA motor onset with automatisms, it is 95% likely that the seizure onset zone is mesial temporal. Focal onset impaired awareness and FOIA/ FOA motor onset with automatisms are most commonly seen in temporal lobe seizures, but can also occur in temporal plus, “pseudotemporal” and other extratemporal seizure types. Our data suggests that these seizure semiologies following ICA, can be highly predictive of mesial temporal rather than extra-temporal seizures, and should be actively sought in video EEG semiological analysis in building implantation hypotheses for SEEG.

Quantitative analysis of video-SEEG recordings identified mesial and additional basal temporal structures activation at ICA onset.

Electrocorticography recordings in patients undergoing epilepsy surgery have suggested that functional activation of cortex is consistently associated with a broadband increase in signal power at higher frequencies greater than 60 Hz^{21, 22}. Gamma responses have been observed in a variety of functional domains^{23, 24} and in cortical breathing mapping²⁵. Based on previous studies it is reasonable to speculate that high gamma activity reflects neural activity tightly related to cortical function and may be exploited for functional mapping of respiration^{23, 24}. In our study, we identified increased high gamma activity suggesting neuronal activation in several cortical structures within the mesial and basal temporal lobe. Early-onset ICA occurred synchronously with high gamma ictal discharges in mesial amygdala, hippocampus, fusiform and lingual gyri. This is consistent with the central apnea elicited by electrical stimulation (the gold standard technique for the mapping of cognitive, emotional and other brain functions²⁶) of mesial temporal regions and are likely to represent anatomical substrates for ICA generation^{1, 27–30}. It follows that early ICA likely indicates seizure discharges in, or in close proximity to, these regions, and is more indicative of the seizure onset zone. This observation, while under recognized in seizure analysis, is not surprising since these structures are part of the limbic/paralimbic non-volitional breathing modulatory network. Some hippocampal and amygdala neurons phase-lock with the respiratory cycle in humans, suggesting that these structures are involved in breathing regulation ^31–33. Single-pulse stimulation of the central nucleus of the amygdala in cats paces the inspiratory cycle³⁴. Neuroimaging breathing studies have implicated limbic-paralimbic regions in the control of breathing^{35, 36}. Additionally, air hunger is known to activate fusiform and lingual gyri, suggesting their involvement in breathing³⁷. The more likely mechanism of temporal brain region activation inducing central apnea is the inhibition or disruption of brainstem neuronal respiratory function either through the thalamus or through direct connections. Late-onset ICA was characterized by ictal discharges in these same regions (except in amygdala), and in two additional regions, including the temporal pole and frontal opercular regions, suggesting these areas potentially contribute to apnea generation, and maybe part of a wider suprapontine breathing network. A possible role of temporal pole in breathing has been suggested in several stimulation studies in animals ³⁸ and in humans ^39–41 as a part of the limbic/paralimbic non volitional respiratory modulation. The temporal pole connects indirectly to brainstem nuclei through the fasciculus amygdalo-temporalis ⁴² and receives hippocampal afferents via the subiculum ⁴². The posterior frontal operculum carries primary motor representation of the larynx ^43–45, and is involved in a variety of orofacial⁴⁶, pharyngeal and laryngeal movements⁴⁷, some of which may contribute to volitional control of respiration⁴⁸. Since these regions were activated late in the seizure, the localizing value of late ICA in relation to the seizure onset zone is limited. However, its association with new gamma activity in the same regions as early ICA suggest consistency in the anatomical substrates of ICA.

Our study has several limitations. Electrode locations were standard of clinical care and selected based on clinical diagnosis and surgical hypothesis, thus limiting our sampling of other potential ICA anatomical substrates. Although SEEG is a unique opportunity to understand ictal semiological signs in epilepsy and to map cortical human functions, the spatial sampling of the epileptogenic network is an important issue in SEEG⁴⁹. The number of electrodes and the selection of sampled regions in each patient is necessarily restricted, which may constitute a limitation in comparison with more global brain network approaches (using MRI or magnetoencephalography [MEG], for example). There was limited sample size, accounting for large confidence intervals in some of the significant findings, although this is so far the largest SEEG study analyzing the value of ICA. Quantitative methods have a limited capacity for identifying unique areas of ICA generation due to sampling biases inherent to SEEG studies, and because multiple sampled and non-sampled areas may be simultaneously involved in a semiological sign especially during seizure propagation.

In conclusion, we propose that ICA is a useful semiological sign in SEEG evaluations and epilepsy surgery. However, ICA diagnosis requires additional multimodal respiratory monitoring, with thoracic-abdominal belts. ICA is frequently the first clinical signal in focal seizures, and sometimes the only clinical manifestation. ICA predicts seizure onset in mesial temporal structures with a high specificity. High gamma activity in mesial or basal temporal regions (amygdala, hippocampus, fusiform and lingual gyrus) likely represents anatomo-electrical substrates for ICA generation. Recognition of ICA is therefore important for guiding SEEG implantations, which should include these structures as possible ictal onset sites. Additionally, prolonged ICA is of SUDEP interest and when associated with severe hypoxemia has been posited as a potential SUDEP biomarker, and hence may identify those PwE with a higher risk of SUDEP. These findings link observations in previous stimulation studies that found temporal limbic/paralimbic regions, to be possible symptomatogenic zones of ICA, to direct qualitative and quantitative confirmation of ICA arising from seizure discharges in these regions in a prospective SEEG cohort.

Summary for Social Media.

1. If you and/or a co-author has a Twitter handle that you would like to be tagged, please enter it here. (format: @AUTHORSHANDLE).

   @SLhatoo

2. What is the current knowledge on the topic?

3. Ictal central apnea (ICA) is a recently recognized semiological seizure sign seen in patients with focal epilepsy. Its value in localizing seizures to parts of the brain as an aid to identifying areas for surgical intervention is under-recognized.

4. What question did this study address?

   This study provides evidence of the clinical value of recognizing ICA as a semiological sign in patients undergoing epilepsy surgery evaluations. This has the potential to aid the success of epilepsy surgery, improving outcomes in patients with intractable epilepsy.

5. What does this study add to our knowledge?

   ICA has an almost four-fold greater association with mesial temporal seizure onset zones, compared to those without ICA, and thus ICA is highly specific for mesial temporal seizure onset zones. If ICA is the first clinical feature, followed by certain other seizure types, it further strengthens odds and specificity for mesial temporal onsets by up to 30-fold increases.

6. How might this potentially impact on the practice of neurology?

   ICA recognition may help anatomo-electro-clinical localization of clinical seizure onsets to specific mesial and basal temporal brain regions, and the inclusion of these regions in SEEG evaluations may help accurately pinpoint seizure onset zones for resection.

DATA AVAILABILITY:

Data from the study is available upon the corresponding author’s request.