Objective: Research in seizure prediction from intracranial EEG has highlighted the usefulness of bivariate measures of brainwave synchronization. Spatio-temporal bivariate features are very high-dimensional and cannot be analyzed with conventional statistical methods. Hence, we propose state-of-the-art machine learning methods that handle high-dimensional inputs. Methods: We computed bivariate features of EEG synchronization (cross-correlation, nonlinear interdependence, dynamical entrainment or wavelet synchrony) on the 21-patient Freiburg dataset. Features from all channel pairs and frequencies were aggregated over consecutive time points, to form patterns. Patient-specific machine learning-based classifiers (support vector machines, logistic regression or convolutional neural networks) were trained to discriminate interictal from preictal patterns of features. In this explorative study, we evaluated out-of-sample seizure prediction performance, and compared each combination of feature type and classifier. Results: Among the evaluated methods, convolutional networks combined with wavelet coherence successfully predicted all ouf-of-sample seizures, without false alarms, on 15 patients, yielding 71% sensitivity and 0 false positives. Conclusions: Our best machine learning technique applied to spatio-temporal patterns of EEG synchronization outperformed previous seizure prediction methods on the Freiburg dataset. Significance: By learning spatio-temporal dynamics of EEG synchronization, pattern recognition could capture patient-specific seizure precursors. Further investigation on additional datasets should include the seizure prediction horizon.

Summary:Multicenter, retrospective analysis of 70 subjects with TSC following surgery for relief of epilepsy revealed significant associations between younger age at seizure onset, present/prior history of infantile spasms, interictal focality (bilateral versus unilateral), and absence of residual postoperative predominant tuber, and poorer postoperative outcome (p < 0.01). Ictal multifocality, mental retardation, and discordant EEG and MRI data showed a negative trend toward outcome, but were not significant.

No abstract

Objective To describe seizure outcomes in patients with medically refractory epilepsy who had evidence of bilateral mesial temporal lobe (MTL) seizure onsets and underwent MTL resection based on chronic ambulatory intracranial EEG (ICEEG) data from a direct brain‐responsive neurostimulator (RNS) system. Methods We retrospectively identified all patients at 17 epilepsy centers with MTL epilepsy who were treated with the RNS System using bilateral MTL leads, and in whom an MTL resection was subsequently performed. Presumed lateralization based on routine presurgical approaches was compared to lateralization determined by RNS System chronic ambulatory ICEEG recordings. The primary outcome was frequency of disabling seizures at last 3‐month follow‐up after MTL resection compared to seizure frequency 3 months before MTL resection. Results We identified 157 patients treated with the RNS System with bilateral MTL leads due to presumed bitemporal epilepsy. Twenty‐five patients (16%) subsequently had an MTL resection informed by chronic ambulatory ICEEG (mean = 42 months ICEEG); follow‐up was available for 24 patients. After MTL resection, the median reduction in disabling seizures at last follow‐up was 100% (mean: 94%; range: 50%‐100%). Nine patients (38%) had exclusively unilateral electrographic seizures recorded by chronic ambulatory ICEEG and all were seizure‐free at last follow‐up after MTL resection; eight of nine continued RNS System treatment. Fifteen patients (62%) had bilateral MTL electrographic seizures, had an MTL resection on the more active side, continued RNS System treatment, and achieved a median clinical seizure reduction of 100% (mean: 90%; range: 50%‐100%) at last follow‐up, with eight of fifteen seizure‐free. For those with more than 1 year of follow‐up (N = 21), 15 patients (71%) were seizure‐free during the most recent year, including all eight patients with unilateral onsets and 7 of 13 patients (54%) with bilateral onsets. Significance Chronic ambulatory ICEEG data provide information about lateralization of MTL seizures and can identify additional patients who may benefit from MTL resection.

In patients with tuberous sclerosis complex (TSC), the high rates of mental retardation are associated with cortical tubers, seizure activity, and genetic factors. The goal of the study was to investigate the relationship between bilateral cortical tubers and seizure variables and mental retardation in individuals with TSC. The records of 27 patients with TSC (age 6 months to 33 years) undergoing neuropsychological assessment and the following clinical variables were examined: bilateral versus non-bilateral cortical tubers, the age of seizure onset, and presence of infantile spasms. Results were statistically analyzed. Bilateral cortical tubers (p=0.02) and early age of seizure onset (p=0.04) were significantly related to impaired cognitive functioning. Only one of the seven patients with normal cognitive functioning had bilateral tubers, whereas 13/21 patients with intellectual impairment had bilateral tubers. Patients with normal cognitive functioning experienced a mean age of seizure onset after 6 years. A trend was observed between infantile spasms and cognitive functioning (p=0.06); the lack of statistical significance likely reflects the small sample size. Neither age nor gender was related to cognitive status. Further investigation incorporating additional neuroimaging factors, antiepileptic treatment effects, and genetic variables, is needed.