Machine learning-based classification of temporal lobe epilepsy subtypes and surgical prognosis evaluation using PET metabolic networks

This study evaluated the association between temporal piriform cortex (tPC) resection and long-term postoperative outcomes in temporal lobe epilepsy (TLE). We conducted a retrospective analysis of patients with TLE undergoing surgery between 2012 and 2022, with a minimum follow-up of 2 years. The extent of resection, including the tPC, hippocampal head/body, and amygdala, was assessed using magnetic resonance imaging scans. These measurements, along with demographic data and other presurgical evaluations, were analyzed for their associations with Engel class postoperative outcomes and antiseizure medications (ASMs) withdrawal across different TLE subtypes. Among 216 included patients, 158 had mesial TLE (MTLE). At the final follow-up, 131 MTLE and 35 neocortical TLE (NTLE) patients were seizure-free. Multivariable logistic regression revealed that unresected tPC significantly predicted seizure recurrence (p < .001, odds ratio [OR] = 4.415, 95% confidence interval [CI] = 2.032-9.594), along with older age at surgery (p = .018, OR = 1.034, 95% CI = 1.006-1.063) and nonspecific pathology (p = .017, OR = 3.899, 95% CI = 1.278-11.894). In NTLE, tPC resection did not significantly affect outcomes. However, in MTLE, unresected tPC was strongly associated with poorer seizure outcomes at both 2-year (p = .012, OR = 3.362, 95% CI = 1.311-8.621) and 5-year (p = .014, OR = 5.750, 95% CI = 1.423-23.242) follow-ups. Among seizure-free MTLE patients, those with tPC resection had higher rates of ASMs reduction and withdrawal after 5 years (p < .05). In hippocampal sclerosis (HS) patients, unresected tPC correlated with seizure recurrence (p < .05), whereas no such association was observed in non-HS cases. Our findings suggest that tPC involvement in the epileptogenic zone varies by TLE subtype, with tPC resection strongly linked to favorable outcomes in MTLE, particularly in HS patients. These results reinforce the understanding of the hippocampus-amygdala-piriform complex as crucial to the epileptogenic zone in MTLE.

AimsDifferentiating mesial temporal lobe epilepsy (MTLE) and neocortical temporal lobe epilepsy (NTLE) remains challenging. Our study characterized the metabolic profiles between MTLE and NTLE and their correlation with surgical prognosis using 18F‐FDG‐PET.MethodsA total of 137 patients with intractable temporal lobe epilepsy (TLE) and 40 age‐matched healthy controls were recruited. Patients were divided into the MTLE group (N = 91) and the NTLE group (N = 46). 18F‐FDG‐PET was used to measure the metabolism of regional cerebra, which was analyzed using statistical parametric mapping. The volume of abnormal metabolism in cerebral regions and their relationship with surgical prognosis were calculated for each surgical patient.ResultsThe cerebral hypometabolism of MTLE was limited to the ipsilateral temporal and insular lobes (p < 0.001, uncorrected). The NTLE patients showed hypometabolism in the ipsilateral temporal, frontal, and parietal lobes (p < 0.001, uncorrected). The MTLE patients showed extensive hypermetabolism in cerebral regions (p < 0.001, uncorrected). Hypermetabolism in NTLE was limited to the contralateral temporal lobe and cerebellum, ipsilateral frontal lobe, occipital lobe, and bilateral thalamus (p < 0.001, uncorrected). Among patients who underwent resection of epileptic lesions, 51 (67.1%) patients in the MTLE group and 10 (43.5%) in the NTLE group achieved Engel class IA outcome (p = 0.041). The volumes of metabolic increase for the frontal lobe or thalamus in the MTLE group were larger in non‐Engel class IA patients than Engel class IA patients (p < 0.05).ConclusionsThe spatial metabolic profile discriminated NTLE from MTLE. Hypermetabolism of the thalamus and frontal lobe in MTLE may facilitate preoperative counseling and surgical planning.

BackgroundMesial temporal lobe epilepsy (TLE) is one of the most widespread neurological network disorders. Computational anatomy MRI studies demonstrate a robust pattern of cortical volume loss. Most statistical analyses provide information about localization of significant focal differences in a segregationist way. Multivariate Bayesian modeling provides a framework allowing inferences about inter‐regional dependencies. We adopt this approach to answer following questions: Which structures within a pattern of dynamic epilepsy‐associated brain anatomy reorganization best predict TLE pathology. Do these structures differ between TLE subtypes?MethodsWe acquire clinical and MRI data from TLE patients with and without hippocampus sclerosis (n = 128) additional to healthy volunteers (n = 120). MRI data were analyzed in the computational anatomy framework of SPM12 using classical mass‐univariate analysis followed by multivariate Bayesian modeling.ResultsAfter obtaining TLE‐associated brain anatomy pattern, we estimate predictive power for disease and TLE subtypes using Bayesian model selection and comparison. We show that ipsilateral para‐/hippocampal regions contribute most to disease‐related differences between TLE and healthy controls independent of TLE laterality and subtype. Prefrontal cortical changes are more discriminative for left‐sided TLE, whereas thalamus and temporal pole for right‐sided TLE. The presence of hippocampus sclerosis was linked to stronger involvement of thalamus and temporal lobe regions; frontoparietal involvement was predominant in absence of sclerosis.ConclusionsOur topology inferences on brain anatomy demonstrate a differential contribution of structures within limbic and extralimbic circuits linked to main effects of TLE and hippocampal sclerosis. We interpret our results as evidence for TLE‐related spatial modulation of anatomical networks.

Background/Objectives: Temporal lobe epilepsy (TLE) responds well to surgical treatment, although a considerable percentage of patients experience seizure recurrence after resection. Relapse from the contralateral mesial temporal lobe, extratemporal lobe epilepsy mimicking TLE, or temporal plus epilepsy might account for surgical failures. Methods: We included patients with a pre-implantation hypothesis suggesting TLE, who underwent stereo-EEG (SEEG) evaluation at our institution and had an individual SEEG exploration paradigm with at least twelve stereo-electrodes placed to sixteen brain regions allowing exploration of limbic and paralimbic networks. We analyzed the prevalence of TLE subtypes based on ictal onset localization with SEEG and response to resective surgery. Results: Twenty-four subjects met the inclusion criteria. Seven patients had unilateral mesial temporal epilepsy (UMTE), five had bilateral mesial temporal epilepsy (BMTE), five had unilateral neocortical temporal epilepsy (UNTE), six had temporal-plus epilepsy (TPE), one had extratemporal epilepsy (ETE). The number of patients who underwent destructive surgeries and surgical outcomes are as follows: UMTE-all seven patients, Engel I; BMTE- three out of five, Engel I, III, and IV, respectively; UNTE-three out of five, Engel I; TLE mimic (ETE)-one, Engel I; TPE-all six patients, Engel I-three, Engel III-two, Engel IV-one. Conclusions: In our study, UMTE was the most frequent TLE subtype (29%), and all patients proceeded to resective surgery with good outcomes. TPE comprised a substantial component (25%) of this cohort with initially presumed TLE, who had a notable proportion of unfavorable outcomes. Larger studies are needed to create guidelines for rational counseling of patients with presumed TLE regarding surgical outcomes.

This study aims to evaluate the glymphatic system (GS) in different temporal lobe epilepsy (TLE) subtypes using diffusion tensor imaging analysis along the perivascular space (DTI-ALPS) and to explore its correlation with clinical factors and memory performance. The study encompassed 112 TLE patients with hippocampal sclerosis (TLE-HS), 73 TLE patients with no lesions on magnetic resonance imaging (TLE-NL), and 55 healthy controls. The DTI-ALPS index was calculated based on 3.0T diffusion tensor image sequences, and the memory performance was assessed using the Wechsler Memory Scale-Revised. The DTI-ALPS index was compared among the three groups, and its relationships with clinical characteristics and memory performance were explored. TLE-HS group showed a significantly lower DTI-ALPS index compared with healthy controls in both hemispheres (ipsilateral: p < 0.001; contralateral: p = 0.002). By contrast, TLE-NL group exhibited a reduced DTI-ALPS index solely in the ipsilateral hemisphere (p < 0.001). Within TLE-NL cohort, those with a history of focal to bilateral tonic-clonic seizures showed reduced DTI-ALPS indices in both hemispheres (ipsilateral: p = 0.037; contralateral: p = 0.004). In the TLE-HS group, DTI-ALPS index positively correlated with memory performance (ps < 0.05). A multiple regression analysis indicated that the average DTI-ALPS index was significantly associated with memory quotient (β = 0.309, p < 0.001; R2 = 0.226), independent of the ipsilateral hippocampal volume. The patterns of reduced DTI-ALPS index differed between TLE-HS and TLE-NL patients. The extent of GS impairment in TLE-HS patients correlated with memory decline, suggesting its potential as a therapeutic target for memory enhancement. This study employed the DTI-ALPS index, a neuroimaging marker, to assess glymphatic system function in distinct subtypes of temporal lobe epilepsy (TLE). Glymphatic impairment was observed in both TLE with hippocampal sclerosis (TLE-HS) and nonlesional TLE (TLE-NL), exhibiting distinct patterns. Notably, this dysfunction was associated with memory deficits, suggesting that targeting glymphatic clearance may represent a novel therapeutic strategy for memory improvement in epilepsy.

Seizures of temporal lobe epilepsy: identification of subtypes by coherence analysis using stereo-electro-encephalography

This study compared the metabolic regional alterations, characterized by proton magnetic spectroscopic imaging ((1)H-MRSI), with electrophysiological abnormalities recorded by using depth electrodes and with structural lesions, in patients with several subtypes of temporal lobe epilepsy (TLE). Twenty-five subjects were investigated, including 15 controls and 10 patients with drug-resistant unilateral TLE, nine of whom had structural abnormalities identified by MRI. All patients underwent noninvasive presurgical evaluation and then stereoelectroencephalography (SEEG). We performed an original metabolic exploration combining two (1)H-MRS imaging acquisitions associated with two single-voxel acquisitions (temporal poles) to map the most informative regions of interest (ROIs) including mesial and neocortical localizations. The N-acetyl aspartate/(choline+creatine) ratio was chosen as a metabolic index. SEEG analysis allowed the classification of each ROI as electrically normal or abnormal (i.e., involved in ictal and/or interictal discharges). Groups were compared by using a nonparametric Mann-Whitney U test. N-Acetyl aspartate/(choline+creatine) was significantly lower in all regions involved in SEEG electrophysiological epileptic abnormalities than in controls (p < 0.05). In contrast, the regions without any electrophysiological abnormalities were not metabolically different from those in controls (p > 0.05) except in one ROI. No differences between the metabolic profiles of epileptogenic and irritative zones were found. The metabolic alterations included, but also extended beyond, the lesions. The presence of metabolic abnormalities in mesial structures was not specific for the mesial subtype and generally extended outside the mesial structures. These results indicate that metabolic abnormalities are linked to ictal and interictal epileptiform activities rather than to structural alterations in TLE.

Recent studies highlight the amygdala's crucial role in temporal lobe epilepsy (TLE), particularly in magnetic resonance imaging-negative cases and new TLE subtypes with structural amygdala changes. This study aims to investigate the electrophysiological properties and connectivity patterns of the amygdaloid complex in TLE patients using stereoelectroencephalography (SEEG). From March 2020 to December 2023, we collected data from nine patients with drug-resistant TLE who underwent SEEG with dual amygdala trajectories: dorsal amygdala (DA) targeting medial and central nuclei, and ventral amygdala (VA) targeting basal and lateral nuclei. We analyzed interictal and ictal activities, focusing on power spectral density, interictal epileptiform discharges (IEDs), and coherence between DA and VA regions and ipsilateral regions of the temporal-limbic network, including the hippocampus, superior temporal gyrus, entorhinal cortex, fusiform gyrus, orbitofrontal cortex, temporal pole, and ventral insula. IEDs were significantly higher in the DA, particularly in patients with epileptogenic zones involving the amygdala. We identified frequency-dependent connectivity patterns between DA and VA with ipsilateral cortical areas of interest during ictal activity. The VA exhibited higher connectivity with the limbic network in a greater number of seizures, particularly with the hippocampal head, entorhinal cortex, fusiform gyrus, and temporal pole in delta and theta frequencies. In contrast, DA connectivity was mainly confined to the hippocampus in theta and high-gamma frequency ranges. Our findings reveal distinct functional connectivity patterns and potentially divergent roles of DA and VA in TLE. These insights could refine intracranial sampling protocols of the amygdaloid complex, guiding more precise strategies for resection and neuromodulation in TLE patients.

Temporal lobe epilepsy is the most common focal epilepsy in adults. While temporal lobe epilepsy was historically perceived to have a largely acquired aetiology, growing evidence points to important genetic contributions. There are several temporal lobe epilepsy subtypes, including mesial temporal lobe epilepsy with or without hippocampal sclerosis, but the relative genetic contributions to each of these subtypes have not been directly studied.In this study, we use the classical twin model in 80 twin pairs where at least one twin had temporal lobe epilepsy. We assessed the genetic contribution to various subtypes [lesional temporal lobe epilepsy, non-lesional temporal lobe epilepsy, mesial temporal lobe epilepsy (with or without hippocampal sclerosis), lateral temporal lobe epilepsy and non-localized temporal lobe epilepsy], by analysing the concordance for temporal lobe epilepsy in monozygotic twins compared with dizygotic twins. In the 10 monozygotic pairs where at least one twin had hippocampal sclerosis, we searched for within-pair acquired differences between affected and unaffected individuals.There was an excess of monozygotic pairs concordant for temporal lobe epilepsy compared with dizygotic pairs (17/47 concordant monozygotic versus 0/33 concordant dizygotic, P < 0.05). This supports a genetic contribution to temporal lobe epilepsy, but notably this concordance was driven by non-lesional temporal lobe epilepsy cases, particularly mesial temporal lobe epilepsy without hippocampal sclerosis (14/22 concordant monozygotic versus 0/11 concordant dizygotic, P < 0.05). No concordant monozygotic or dizygotic pairs were observed in the lesional temporal lobe epilepsy (n = 8) and non-localized temporal lobe epilepsy (n = 15) groups.The concordance for temporal lobe epilepsy in monozygotic twins with mesial temporal lobe epilepsy with hippocampal sclerosis was much lower (2/10 concordant monozygotic versus 0/9 concordant dizygotic, P = 1), suggesting a lesser contribution from germline genetic causes to mesial temporal lobe epilepsy with hippocampal sclerosis. Eight monozygotic twin pairs were discordant for hippocampal sclerosis. In four of these pairs, both twins had febrile seizures, but hippocampal sclerosis was only present in the twin who had prolonged seizures.The two monozygotic twin pairs concordant for hippocampal sclerosis had clinical neurofibromatosis type 1 with pathogenic germline NF1 variants.Our findings confirm a germline genetic component in temporal lobe epilepsy, strongest in mesial temporal lobe epilepsy without hippocampal sclerosis and present in lateral temporal lobe epilepsy but absent in lesional and non-localized temporal lobe epilepsy. In our mesial temporal lobe epilepsy with hippocampal sclerosis twins, we found both genetic factors (NF1) and prolonged febrile seizures contributed to the aetiology of hippocampal sclerosis.

The aim of this study was to investigate the different ictal onset stereoelectroencephalography patterns (IOPs) in patients with drug-resistant temporal lobe epilepsy (TLE). We examined whether the IOPs relate to different TLE subtypes, MRI findings, and underlying pathologies, and we evaluated their prognostic value for predicting the surgical outcome. We retrospectively analyzed data from patients with TLE who underwent stereoelectroencephalography (SEEG) monitoring followed by surgical resection between January 2018 and January 2020. The SEEG recordings were independently analyzed by two epileptologists. Forty-five patients were included in the study, and 61seizures were analyzed. Five IOPs were identified: low voltage fast activity (LVFA; 44.3%), spike-and-wave activity (16.4%), low frequency high-amplitude periodic spikes (LFPS; 18%), a burst of high-amplitude polyspikes (8.2%), and rhythmic sharp activity at ≤ 13Hz (13.1%). Thirty-two patients were found to have a single IOP, while the other 13 patients had two or more IOPs. All five IOPs were found to occur in the medial temporal lobe epilepsy (MTLE), while four IOPs occurred in the lateral temporal lobe epilepsy (LTLE). The LFPS was a common IOP that could distinguish MTLE from LTLE (x2 = 7.046, p = 0.011). Among the MTLE patients, the LFPS was exclusively seen in cases of hippocampal sclerosis (x2 = 5.058, p = 0.038), while the LVFA was associated with nonspecific histology (x2 = 6.077, p = 0.023). The IOPs were not found to differ according to whether the MRI scans were positive or negative. After surgery, patients achieved the higher seizure-free rate at 81.8% and 77.8%, respectively, if the LFPS and LVFA were the predominant patterns. Multiple IOPs or a negative MRI did not indicate a poor prognosis. Five distinct IOPs were identified in the patients with TLE. The differences found have important clinical implications and could provide complementary information for surgical decision-making, especially in MRI-negative patients.

Typical semiology with associated 5As (Antecedent, Aura, Arrest, Automatisms, and Amnesia) is the prototypical seizure manifestation of temporal lobe epilepsy (TLE) in adults. However, patients with TLE can present with seizure phenomenology referred to as "hypermotor TLE" (HMS-TLE) which can mimic extratemporal focal epilepsy. We studied the clinico-electrographic, imaging profile, and surgical outcomes of drug-resistant HMS-TLE patients. We analyzed the ictal signs and electroencephalography data of HMS-TLE patients who became seizure free following epilepsy surgery with a minimum follow-up of 1 year and compared them with patients with typical TLE (1:2 ratio). Hypermotor seizures were identified based on the presence of stereotyped ictal complex movements at seizure onset including body rocking, pelvic thrusting, and bimanual and bipedal movements with/without affective components. Out of 684 patients with drug-resistant TLE who underwent surgery during 2010-2020, 16 patients (2.34%) met criteria for HMS-TLE and were compared with 32 patients with typical TLE. Predominant "hypermotor" components included motor agitation, bicycling/kicking movements, body gyration followed by violent automatisms, and pelvic thrusting. Age at epilepsy onset, age at video telemetry recording, age at surgery, duration of epilepsy at surgery, gender distribution, and presence of aura were comparable between the two groups. History of febrile seizures (31.3% vs. 65.6%; p-value = .024), presence of hippocampal sclerosis on MRI, and histopathology (56.3% vs. 90.6%; p-value = .006 and 50% vs. 96.9%; p-value = .002, respectively) were lower, and temporal neocortical abnormalities on MRI (87.5% vs. 59.4%, p-value = .048), dysplasia on histopathological evaluation (25% vs. 0%, p-value = .003) were higher in HMS-TLE. HMS-TLE group had lower number of patients with focal-onset ictal patterns (37.5% vs. 90.6%, p-value = .0001). Hypermotor seizures are rare in TLE, however it does not preclude non-invasive selection for surgery. Hence, physicians should be aware of this subtype of drug-resistant TLE as early intervention can have good surgical outcomes, akin to typical TLE.

Monoamine neurotransmitters in resected hippocampal subparcellations from neocortical and mesial temporal lobe epilepsy patients:: In situ microvoltammetric studies

We studied cerebral perfusion patterns in the various subtypes of TLE, as determined by pathology and good outcome after temporal lobectomy (as confirmation of temporal origin). We studied clinical features and ictal technetium 99m hexamethyl-propyleneamineoxime (99mTc-HM-PAO) single-photon emission-computed tomography (SPECT) in four subgroups of patients with intractable temporal lobe epilepsy (TLE) treated with surgery: hippocampal sclerosis (group 1, n = 10), foreign-tissue lesion in mesial temporal lobe (group 2, n = 8), foreign-tissue lesion in lateral temporal lobe (group 3, n = 7), and normal temporal lobe tissue with good surgical outcome (group 4, n = 5). No major clinical differences in auras, complex partial seizures or postictal states were identified among the groups. Ictal SPECT showed distinct patterns of cerebral perfusion in these subtypes of TLE. In groups 1 and 2, hyperperfusion was seen in the ipsilateral mesial and lateral temporal regions. In group 3, hyperperfusion was seen bilaterally in the temporal lobes with predominant changes in the region of the lesion. Hyperperfusion was restricted to the ipsilateral anteromesial temporal region in group 4. Ipsilateral temporal hyperperfusion in mesial onset seizures can be explained by known anatomic projections between mesial structures and ipsilateral temporal neocortex. Bilateral temporal hyperperfusion in lateral onset seizures can be explained by the presence of anterior commissural connections between lateral temporal neocortex and the contralateral amygdala. We conclude that the perfusion patterns seen on ictal SPECT are helpful for subclassification of temporal lobe seizures, whereas clinical features are relatively unhelpful. These perfusion patterns provide an insight into preferential pathways of seizure propagation in the subtypes of TLE.

MRI and PET imaging enables subgroups of temporal lobe epilepsy (TLE) to be defined on the basis of structural pathology. Few studies have examined the variation in electroclinical seizure spread patterns based on imaging findings. We performed a retrospective cohort study to investigate the electroclinical differences among 3 specific groups of TLE: MRI-negative PET-positive TLE (MRI-negative TLE), temporal lobe lesion TLE (lesional TLE), and unilateral hippocampal sclerosis TLE (HS-TLE). Patients with an electroclinical diagnosis of TLE who had video-scalp EEG recordings of seizures were identified from the retrospective database of the Austin Comprehensive Epilepsy Program between 2005 and 2019. The cohort was further selected into the 3 defined groups based on imaging findings, using MRI and FDG-PET. Timings of clinical and electrographic seizure progression were measured, considering the onset, ipsilateral lobar spread, contralateral spread, and termination. Durations were compared between groups using linear mixed models with inclusion of demographic and clinical covariates. A total of 105 patients (137 seizures) were included, comprising 36 with MRI-negative TLE (54 seizures), 36 with lesional TLE (18 lateral vs 16 mesial lesions; 44 seizures), and 33 with HS-TLE (39 seizures). Seizure duration was similar between MRI-negative TLE and lesional TLE (mean 75.9 vs 71.7 seconds; p = 0.91). Further dividing lesional TLE into medial vs lateral temporal revealed no timing difference. However, the HS-TLE group had longer total seizure duration (114 seconds) compared with both MRI-negative TLE (p < 0.001) and lesional TLE (p < 0.001). Progression of electrographic spread also reflected this pattern, with involvement of extratemporal regions and then the contralateral hemisphere each taking significantly longer in HS-TLE. MRI-negative TLE appears electrographically similar to lesional TLE, whether mesial or lateral, in the duration of seizures and the timing of electrographic spread. Both appear electrographically different from HS-TLE, where propagation is slower, suggesting engagement of different epileptogenic networks or seizure suppression mechanisms. This study provides Class II evidence that the electroclinical features of seizures in HS-TLE are different than MRI-negative TLE and lesional TLE.

Clinical features that may help to differentiate medial temporal lobe epilepsy (MTLE) from neocortical temporal lobe epilepsy (NTLE) are lacking. To investigate the localizing and lateralizing value of the association of ipsilateral motor automatisms and contralateral dystonic posturing in patients with medically refractory temporal lobe epilepsy. Videotapes of 60 patients with well-defined MTLE, NTLE, or both were reviewed to assess the presence and the localizing value of unilateral dystonic posturing associated with motor automatisms. Twenty-eight of the 60 patients exhibited unilateral dystonic posturing. This sign was observed in patients with MTLE and NTLE. It was mostly contralateral to the seizure focus in patients with MTLE and exclusively ipsilateral in patients with NTLE. Unilateral motor automatisms occurred in 26 of the 60 patients with MTLE or NTLE. It was predominantly ipsilateral to the seizure focus in patients with MTLE and exclusively contralateral in patients with NTLE. The association of ipsilateral motor automatisms and contralateral dystonic posturing was found in 14 patients with MTLE but in none of the patients with NTLE. Two patients who had medial and neocortical seizure onset also exhibited this clinical feature. This association was not significantly correlated with the postoperative outcome in patients with MTLE. The association of ipsilateral motor automatisms and contralateral dystonic posturing may help to differentiate MTLE from NTLE with a reliable lateralizing value. This clinical association may reflect a specific pattern in the spread of the ictal discharge.