Early Identification of Candidates for Epilepsy Surgery: A Multicenter, Machine Learning, Prospective Validation Study.

BACKGROUND AND OBJECTIVES: Epilepsy surgery is often delayed. We previously developed machine learning (ML) models to identify candidates for resective epilepsy surgery earlier in the disease course. In this study, we report the prospective validation. METHODS: In this multicenter, prospective, longitudinal cohort study, random forest models were validated at a pediatric epilepsy center consisting of 2 hospitals and 14 outpatient neurology clinic sites and an adult epilepsy center with 2 hospitals and 27 outpatient neurology clinic sites. The models used neurology visit notes, EEG and MRI reports, visit patterns, hospitalizations, and medication, laboratory, and procedure orders to identify candidates for surgery. The models were trained on historical data up to May 10, 2019. Patients with an ICD-10 diagnosis of epilepsy who visited from May 11, 2019, to May 10, 2020, were screened by the algorithm and assigned surgical candidacy scores. The primary outcome was area under the curve (AUC), which was calculated by comparing scores from patients who underwent epilepsy surgery before November 10, 2020, against scores from nonsurgical patients. Nonsurgical patients' charts were reviewed to determine whether patients with high scores were more likely to be missed surgical candidates. Delay to surgery was defined as the time between the first visit that a surgical candidate was identified by the algorithm and the date of the surgery. RESULTS: A total of 5,285 pediatric and 5,782 adult patients were included to train the ML algorithms. During the study period, 41 children and 23 adults underwent resective epilepsy surgery. In the pediatric cohort, AUC was 0.91 (95% CI 0.87-0.94), positive predictive value (PPV) was 0.08 (0.05-0.10), and negative predictive value (NPV) was 1.00 (0.99-1.00). In the adult cohort, AUC was 0.91 (0.86-0.97), PPV was 0.07 (0.04-0.11), and NPV was 1.00 (0.99-1.00). The models first identified patients at a median of 2.1 years (interquartile range [IQR]: 1.2-4.9 years, maximum: 11.1 years) before their surgery and 1.3 years (IQR: 0.3-4.0 years, maximum: 10.1 years) before their presurgical evaluations. DISCUSSION: ML algorithms can identify surgical candidates earlier in the disease course. Even at specialized epilepsy centers, there is room to shorten the time to surgery. CLASSIFICATION OF EVIDENCE: This study provides Class II evidence that a machine learning algorithm can accurately distinguish patients with epilepsy who require resective surgery from those who do not.

Automated, machine learning-based alerts increase epilepsy surgery referrals: A randomized controlled trial.

OBJECTIVE: To determine whether automated, electronic alerts increased referrals for epilepsy surgery. METHODS: We conducted a prospective, randomized controlled trial of a natural language processing-based clinical decision support system embedded in the electronic health record (EHR) at 14 pediatric neurology outpatient clinic sites. Children with epilepsy and at least two prior neurology visits were screened by the system prior to their scheduled visit. Patients classified as a potential surgical candidate were randomized 2:1 for their provider to receive an alert or standard of care (no alert). The primary outcome was referral for a neurosurgical evaluation. The likelihood of referral was estimated using a Cox proportional hazards regression model. RESULTS: Between April 2017 and April 2019, at total of 4858 children were screened by the system, and 284 (5.8%) were identified as potential surgical candidates. Two hundred four patients received an alert, and 96 patients received standard care. Median follow-up time was 24 months (range: 12-36 months). Compared to the control group, patients whose provider received an alert were more likely to be referred for a presurgical evaluation (3.1% vs 9.8%; adjusted hazard ratio [HR] = 3.21, 95% confidence interval [CI]: 0.95-10.8; one-sided p = .03). Nine patients (4.4%) in the alert group underwent epilepsy surgery, compared to none (0%) in the control group (one-sided p = .03). SIGNIFICANCE: Machine learning-based automated alerts may improve the utilization of referrals for epilepsy surgery evaluations.

Prospective validation of a machine learning model that uses provider notes to identify candidates for resective epilepsy surgery.

OBJECTIVE: Delay to resective epilepsy surgery results in avoidable disease burden and increased risk of mortality. The objective was to prospectively validate a natural language processing (NLP) application that uses provider notes to assign epilepsy surgery candidacy scores. METHODS: The application was trained on notes from (1) patients with a diagnosis of epilepsy and a history of resective epilepsy surgery and (2) patients who were seizure-free without surgery. The testing set included all patients with unknown surgical candidacy status and an upcoming neurology visit. Training and testing sets were updated weekly for 1 year. One- to three-word phrases contained in patients' notes were used as features. Patients prospectively identified by the application as candidates for surgery were manually reviewed by two epileptologists. Performance metrics were defined by comparing NLP-derived surgical candidacy scores with surgical candidacy status from expert chart review. RESULTS: The training set was updated weekly and included notes from a mean of 519 ± 67 patients. The area under the receiver operating characteristic curve (AUC) from 10-fold cross-validation was 0.90 ± 0.04 (range = 0.83-0.96) and improved by 0.002 per week (P < .001) as new patients were added to the training set. Of the 6395 patients who visited the neurology clinic, 4211 (67%) were evaluated by the model. The prospective AUC on this test set was 0.79 (95% confidence interval [CI] = 0.62-0.96). Using the optimal surgical candidacy score threshold, sensitivity was 0.80 (95% CI = 0.29-0.99), specificity was 0.77 (95% CI = 0.64-0.88), positive predictive value was 0.25 (95% CI = 0.07-0.52), and negative predictive value was 0.98 (95% CI = 0.87-1.00). The number needed to screen was 5.6. SIGNIFICANCE: An electronic health record-integrated NLP application can accurately assign surgical candidacy scores to patients in a clinical setting.

Long-term treatment of epilepsy with everolimus in tuberous sclerosis.

OBJECTIVE: To evaluate the long-term benefit and safety of everolimus for the treatment of medically refractory epilepsy in patients with tuberous sclerosis complex (TSC). METHODS: Everolimus was titrated over 4 weeks and continued an additional 8 weeks in a prospective, open-label, phase I/II clinical trial design. Participants demonstrating initial benefit continued treatment until study completion (48 months). The primary endpoint was percentage of patients with a ≥50% reduction in seizure frequency compared to baseline. Secondary endpoints assessed absolute seizure frequency, adverse events (AEs), behavior, and quality of life. RESULTS: Of the 20 participants who completed the initial study phase, 18 continued extended treatment. Fourteen of 18 (78%) participants completed the study, all but 1 of whom reported ≥50% reduction in seizure frequency at 48 months. All participants reported at least 1 AE, the vast majority (94%) of which were graded mild or moderate severity. Improvements in behavior and quality of life were also observed, but failed to achieve statistical significance at 48 months. CONCLUSIONS: Improved seizure control was maintained for 4 years in the majority of patients with TSC with medically refractory epilepsy treated with everolimus. Long-term treatment with everolimus is safe and well-tolerated in this population. Everolimus may be a therapeutic option for refractory epilepsy in TSC. CLASSIFICATION OF EVIDENCE: This study provides Class IV evidence that for patients with TSC with medically refractory epilepsy everolimus improves seizure control.

Resection of ictal high frequency oscillations is associated with favorable surgical outcome in pediatric drug resistant epilepsy secondary to tuberous sclerosis complex.

Resective epilepsy surgery can improve seizures when the epileptogenic zone (EZ) is limited to a well-defined region. High frequency oscillations (HFO) have been recognized as having a high association with the seizure onset zone. Therefore, we retrospectively identified ictal HFOs and determined their relationship to specific intracranial features of cortical tubers in children with TSC who underwent resective surgery. We identified 14 patients with drug resistant epilepsy secondary to TSC who underwent subdural grid and strip implantation for presurgical evaluation and subsequent resection with adequate post-surgical follow-up. We aimed to determine the relationship between ictal HFOs, post-resection outcome and neuroimaging features in this population. The largest tuber was identified in all 14 patients (100%). Four patients (29%) had unusual tubers. HFOs were observed at ictal onset in all 14 patients. Seven of 10 patients with complete resection of HFOs were seizure free. The better seizure outcome (ILAE=1-3) was achieved with complete HFO resection regardless of the unique TSC structural features (p=0.0140). Our study demonstrates the presence of ripple and fast ripple range HFOs at ictal onset in children with TSC. Our study showed that complete HFO resection led to the better surgical outcome, independent of MR imaging findings.

Methodological Issues in Predicting Pediatric Epilepsy Surgery Candidates Through Natural Language Processing and Machine Learning.

OBJECTIVE: We describe the development and evaluation of a system that uses machine learning and natural language processing techniques to identify potential candidates for surgical intervention for drug-resistant pediatric epilepsy. The data are comprised of free-text clinical notes extracted from the electronic health record (EHR). Both known clinical outcomes from the EHR and manual chart annotations provide gold standards for the patient's status. The following hypotheses are then tested: 1) machine learning methods can identify epilepsy surgery candidates as well as physicians do and 2) machine learning methods can identify candidates earlier than physicians do. These hypotheses are tested by systematically evaluating the effects of the data source, amount of training data, class balance, classification algorithm, and feature set on classifier performance. The results support both hypotheses, with F-measures ranging from 0.71 to 0.82. The feature set, classification algorithm, amount of training data, class balance, and gold standard all significantly affected classification performance. It was further observed that classification performance was better than the highest agreement between two annotators, even at one year before documented surgery referral. The results demonstrate that such machine learning methods can contribute to predicting pediatric epilepsy surgery candidates and reducing lag time to surgery referral.

Clinical and histopathological outcomes in patients with SCN1A mutations undergoing surgery for epilepsy.

OBJECT: Mutations in the sodium channel alpha 1 subunit gene (SCN1A) have been associated with a wide range of epilepsy phenotypes including Dravet syndrome. There currently exist few histopathological and surgical outcome reports in patients with this disease. In this case series, the authors describe the clinical features, surgical pathology, and outcomes in 6 patients with SCN1A mutations and refractory epilepsy who underwent focal cortical resection prior to uncovering the genetic basis of their epilepsy. METHODS: Medical records of SCN1A mutation-positive children with treatment-resistant epilepsy who had undergone resective epilepsy surgery were reviewed retrospectively. Surgical pathology specimens were reviewed. RESULTS: All 6 patients identified carried diagnoses of intractable epilepsy with mixed seizure types. Age at surgery ranged from 18 months to 20 years. Seizures were refractory to surgery in every case. Surgical histopathology showed evidence of subtle cortical dysplasia in 4 of 6 patients, with more neurons in the molecular layer of the cortex and white matter. CONCLUSIONS: Cortical resection is unlikely to be beneficial in these children due to the genetic defect and the unexpected neuropathological finding of mild diffuse malformations of cortical development. Together, these findings suggest a diffuse pathophysiological mechanism of the patients' epilepsy which will not respond to focal resective surgery.

Everolimus for subependymal giant cell astrocytoma: 5-year final analysis.

OBJECTIVE: To analyze the cumulative efficacy and safety of everolimus in treating subependymal giant cell astrocytomas (SEGA) associated with tuberous sclerosis complex (TSC) from an open-label phase II study (NCT00411619). Updated data became available from the conclusion of the extension phase and are presented in this ≥5-year analysis. METHODS: Patients aged ≥ 3 years with a definite diagnosis of TSC and increasing SEGA lesion size (≥2 magnetic resonance imaging scans) received everolimus starting at 3mg/m(2) /day (titrated to target blood trough levels of 5-15ng/ml). The primary efficacy endpoint was reduction from baseline in primary SEGA volume. RESULTS: As of the study completion date (January 28, 2014), 22 of 28 (78.6%) initially enrolled patients finished the study per protocol. Median (range) duration of exposure to everolimus was 67.8 (4.7-83.2) months; 12 (52.2%) and 14 (60.9%) of 23 patients experienced SEGA volume reductions of ≥50% and ≥30% relative to baseline, respectively, after 60 months of treatment. The proportion of patients experiencing daily seizures was reduced from 7 of 26 (26.9%) patients at baseline to 2 of 18 (11.1%) patients at month 60. Most commonly reported adverse events (AEs) were upper respiratory tract infection and stomatitis of mostly grade 1 or 2 severity. No patient discontinued treatment due to AEs. The frequency of emergence of most AEs decreased over the course of the study. INTERPRETATION: Everolimus continues to demonstrate a sustained effect on SEGA tumor reduction over ≥5 years of treatment. Everolimus remained well-tolerated, and no new safety concerns were noted.

Everolimus treatment of refractory epilepsy in tuberous sclerosis complex.

OBJECTIVE: Epilepsy is a major manifestation of tuberous sclerosis complex (TSC). Everolimus is an mammalian target of rapamycin complex 1 inhibitor with demonstrated benefit in several aspects of TSC. We report the first prospective human clinical trial to directly assess whether everolimus will also benefit epilepsy in TSC patients. METHODS: The effect of everolimus on seizure control was assessed using a prospective, multicenter, open-label, phase I/II clinical trial. Patients≥2 years of age with confirmed diagnosis of TSC and medically refractory epilepsy were treated for a total of 12 weeks. The primary endpoint was percentage of patients with a ≥50% reduction in seizure frequency over a 4-week period before and after treatment. Secondary endpoints assessed impact on electroencephalography (EEG), behavior, and quality of life. RESULTS: Twenty-three patients were enrolled, and 20 patients were treated with everolimus. Seizure frequency was reduced by ≥50% in 12 of 20 subjects. Overall, seizures were reduced in 17 of the 20 by a median reduction of 73% (p<0.001). Seizure frequency was also reduced during 23-hour EEG monitoring (p=0.007). Significant reductions in seizure duration and improvement in parent-reported behavior and quality of life were also observed. There were 83 reported adverse events that were thought to be treatment-related, all of which were mild or moderate in severity. INTERPRETATION: Seizure control improved in the majority of TSC patients with medically refractory epilepsy following treatment with everolimus. Everolimus demonstrated additional benefits on behavior and quality of life. Treatment was safe and well tolerated. Everolimus may be a therapeutic option for refractory epilepsy in this population.

Focal Peak Activities in Spread of Interictal-Ictal Discharges in Epilepsy with Beamformer MEG: Evidence for an Epileptic Network?

Non-invasive studies to predict regions of seizure onset are important for planning intracranial grid locations for invasive cortical recordings prior to resective surgery for patients with medically intractable epilepsy. The neurosurgeon needs to know both the seizure onset zone (SOZ) and the region of immediate cortical spread to determine the epileptogenic zone to be resected. The immediate zone of spread may be immediately adjacent, on a nearby gyrus, in a different lobe, and sometimes even in the contralateral cerebral hemisphere. We reviewed consecutive simultaneous EEG/MEG recordings on 162 children with medically intractable epilepsy. We analyzed the MEG signals in the bandwidth 20-70 Hz with a beamformer algorithm, synthetic aperture magnetometry, at a 2.5 mm voxel spacing throughout the brain (virtual sensor locations, VSLs) with the kurtosis statistic (g 2) to determine presence of excess kurtosis (γ2) consistent with intermittent increased high frequency spikiness of the background. The MEG time series was reconstructed (virtual sensor signals) at each of these VSLs. The VS signals were further examined with a relative peak amplitude spike detection algorithm. The time of VS spike detection was compared to the simultaneous EEG and MEG sensor signals for presence of conventional epileptiform spike morphology in the latter signals. The time of VS spike detection was compared across VSLs to determine earliest and last VSL to show a VS spike. Seven subjects showed delay in activation across VS locations detectable on visual examination. We compared the VS locations that showed earliest and later VS spikes with the locations on intracranial grid locations by electrocorticography (ECoG) that showed spikes and both onset and spread of seizures. We compared completeness of resection of VS locations to postoperative outcome. The VS locations for spike onset and spread were similar to locations for ictal onset and spread by ECoG.

Resection of ictal high-frequency oscillations leads to favorable surgical outcome in pediatric epilepsy.

PURPOSE: Intracranial electroencephalography (EEG) is performed as part of an epilepsy surgery evaluation when noninvasive tests are incongruent or the putative seizure-onset zone is near eloquent cortex. Determining the seizure-onset zone using intracranial EEG has been conventionally based on identification of specific ictal patterns with visual inspection. High-frequency oscillations (HFOs, >80 Hz) have been recognized recently as highly correlated with the epileptogenic zone. However, HFOs can be difficult to detect because of their low amplitude. Therefore, the prevalence of ictal HFOs and their role in localization of epileptogenic zone on intracranial EEG are unknown. METHODS: We identified 48 patients who underwent surgical treatment after the surgical evaluation with intracranial EEG, and 44 patients met criteria for this retrospective study. Results were not used in surgical decision making. Intracranial EEG recordings were collected with a sampling rate of 2,000 Hz. Recordings were first inspected visually to determine ictal onset and then analyzed further with time-frequency analysis. Forty-one (93%) of 44 patients had ictal HFOs determined with time-frequency analysis of intracranial EEG. KEY FINDINGS: Twenty-two (54%) of the 41 patients with ictal HFOs had complete resection of HFO regions, regardless of frequency bands. Complete resection of HFOs (n = 22) resulted in a seizure-free outcome in 18 (82%) of 22 patients, significantly higher than the seizure-free outcome with incomplete HFO resection (4/19, 21%). SIGNIFICANCE: Our study shows that ictal HFOs are commonly found with intracranial EEG in our population largely of children with cortical dysplasia, and have localizing value. The use of ictal HFOs may add more promising information compared to interictal HFOs because of the evidence of ictal propagation and followed by clinical aspect of seizures. Complete resection of HFOs is a favorable prognostic indicator for surgical outcome.

Ictal MEG onset source localization compared to intracranial EEG and outcome: improved epilepsy presurgical evaluation in pediatrics.

PURPOSE: Magnetoencephalography (MEG) has been shown a useful diagnostic tool for presurgical evaluation of pediatric medically intractable partial epilepsy as MEG source localization has been shown to improve the likelihood of seizure onset zone (SOZ) sampling during subsequent evaluation with intracranial EEG (ICEEG). We investigated whether ictal MEG onset source localization further improves results of interictal MEG in defining the SOZ. METHODS: We identified 20 pediatric patients with one habitual seizure during MEG recordings between October 2007 and April 2011. MEG was recorded with sampling rates of 600Hz and 4000Hz for 10 and 2min respectively. Continuous head localization (CHL) was applied. Source localization analyses were applied using multiple algorithms, both at the beginning of ictal onset and for interictal MEG discharges. Ictal MEG onsets were identified by visual inspection and power spectrum using short-time Fourier transform (STFT). Source localizations were compared with ICEEG, surgical procedure and outcome. KEY FINDINGS: Eight patients met all inclusion criteria. Five of the 8 patients (63%) had concordant ictal MEG onset source localization and interictal MEG discharge source localizations in the same lobe, but the source of ictal MEG onset was closer to the SOZ defined by ICEEG. SIGNIFICANCE: Although the capture of seizures during MEG recording is challenging, the source localization for ictal MEG onset proved to be a useful tool for presurgical evaluation in our pediatric population with medically intractable epilepsy.