Diffuse Optical Imaging (DOI) allows the recovery of the hemodynamic response associated with evoked brain activity. The signal is contaminated with systemic physiological interference which occurs in the superficial layers of the head as well as in the brain tissue. The back-reflection geometry of the measurement makes the DOI signal strongly contaminated by systemic interference occurring in the superficial layers. A recent development has been the use of signals from small source-detector separation (1 cm) optodes as regressors. Since those additional measurements are mainly sensitive to superficial layers in adult humans, they help in removing the systemic interference present in longer separation measurements (3 cm). Encouraged by those findings, we developed a dynamic estimation procedure to remove global interference using small optode separations and to estimate simultaneously the hemodynamic response. The algorithm was tested by recovering a simulated synthetic hemodynamic response added over baseline DOI data acquired from 6 human subjects at rest. The performance of the algorithm was quantified by the Pearson R2 coefficient and the mean square error (MSE) between the recovered and the simulated hemodynamic responses. Our dynamic estimator was also compared with a static estimator and the traditional adaptive filtering method. We observed a significant improvement (two-tailed paired t-test, p < 0.05) in both HbO and HbR recovery using our Kalman filter dynamic estimator compared to the traditional adaptive filter, the static estimator and the standard GLM technique.

No abstract

Although magnetoencephalography (MEG) and electroencephalography (EEG) have been available for decades, their relative merits are still debated. We examined regional differences in signal-tonoise-ratios (SNRs) of cortical sources in MEG and EEG. Data from four subjects were used to simulate focal and extended sources located on the cortical surface reconstructed from highresolution magnetic resonance images. The SNR maps for MEG and EEG were found to be complementary. The SNR of deep sources was larger in EEG than in MEG, whereas the opposite was typically the case for superficial sources. Overall, the SNR maps were more uniform for EEG than for MEG. When using a noise model based on uniformly distributed random sources on the cortex, the SNR in MEG was found to be underestimated, compared with the maps obtained with noise estimated from actual recorded MEG and EEG data. With extended sources, the total area of cortex in which the SNR was higher in EEG than in MEG was larger than with focal sources. Clinically, SNR maps in a patient explained differential sensitivity of MEG and EEG in detecting epileptic activity. Our results emphasize the benefits of recording MEG and EEG simultaneously.

Machine learning leverages statistical and computer science principles to develop algorithms capable of improving performance through interpretation of data rather than through explicit instructions. Alongside widespread use in image recognition, language processing, and data mining, machine learning techniques have received increasing attention in medical applications, ranging from automated imaging analysis to disease forecasting. This review examines the parallel progress made in epilepsy, highlighting applications in automated seizure detection from electroencephalography (EEG), video, and kinetic data, automated imaging analysis and pre‐surgical planning, prediction of medication response, and prediction of medical and surgical outcomes using a wide variety of data sources. A brief overview of commonly used machine learning approaches, as well as challenges in further application of machine learning techniques in epilepsy, is also presented. With increasing computational capabilities, availability of effective machine learning algorithms, and accumulation of larger datasets, clinicians and researchers will increasingly benefit from familiarity with these techniques and the significant progress already made in their application in epilepsy.

For unknown reasons, placebos reduce seizures in clinical trials in many patients. It is also unclear why some drugs showing statistical superiority to placebo in one trial may fail to do so in another. Using Seizuretracker.com, a patient-centered database of 684,825 seizures, we simulated “placebo” and “drug” trials. These simulations were employed to clarify the sources of placebo effects in epilepsy, and to identify methods of diminishing placebo effects. Simulation 1 included 9 trials with a 6-week baseline and 6-week test period, starting at time 0, 3, 6…24 months. Here, “placebo” reduced seizures regardless of study start time. Regression-to-the-mean persisted only for 3–6 months. Simulation 2 comprised a 6-week baseline and then 2 years of follow-up. Seizure-frequencies continued to improve throughout follow-up. Although the group improved, individuals switched from improvement to worsening and back. Simulation 3 involved a placebo-controlled “drug” trial, to explore methods of placebo-response reduction. An efficacious “drug” failed to demonstrate a significant effect compared with “placebo” (p=0.12), though modifications either in study start-time (p=0.025) or baseline population reduction (p=0.0028) allowed the drug to achieve a statistically significant effect compared with placebo. In epilepsy clinical trials, some seizure reduction traditionally attributed to placebo effect may reflect the natural course of the disease itself. Understanding these dynamics will allow future investigations into optimal clinical trial design and may lead to identification of more effective therapies.

SummaryObjectiveTo analyze data from Seizure Tracker, a large electronic seizure diary, including comparison of seizure characteristics among different etiologies, temporal patterns in seizure fluctuations, and specific triggers.MethodsZero‐inflated negative binomial mixed‐effects models were used to evaluate temporal patterns of seizure events (during the day or week), as well as group differences in monthly seizure frequency between children and adults and between etiologies. The association of long seizures with seizure triggers was evaluated using a mixed‐effects logistic model with subject as the random effect. Incidence rate ratios (IRRs) and odds ratios were reported for analyses involving zero‐inflated negative binomial and logistic mixed‐effects models, respectively.ResultsA total of 1,037,909 seizures were logged by 10,186 subjects (56.7% children) from December 2007 to January 2016. Children had more frequent seizures than adults did (median monthly seizure frequency 3.5 vs. 2.7, IRR 1.26; p < 0.001). Seizures demonstrated a circadian pattern (higher frequency between 07:00 a.m. and 10:00 a.m. and lower overnight), and seizures were reported differentially across the week (seizure rates higher Monday through Friday than Saturday or Sunday). Longer seizures (>5 or >30 min) had a higher proportion of the following triggers when compared with shorter seizures: “Overtired or irregular sleep,” “Bright or flashing lights,” and “Emotional stress” (p < 0.004).SignificanceThis study explored a large cohort of patients with self‐reported seizures; strengths and limitations of large seizure diary databases are discussed. The findings in this study are consistent with those of prior work in smaller validated cohorts, suggesting that patient‐recorded databases are a valuable resource for epilepsy research, capable of both replication of results and generation of novel hypotheses.

Objective There are no validated methods for predicting the timing of seizures. Using machine learning, we sought to forecast 24‐hour risk of self‐reported seizure from e‐diaries. Methods Data from 5,419 patients on http://SeizureTracker.com (including seizure count, type, and duration) were split into training (3,806 patients/1,665,215 patient‐days) and testing (1,613 patients/549,588 patient‐days) sets with no overlapping patients. An artificial intelligence (AI) program, consisting of recurrent networks followed by a multilayer perceptron (“deep learning” model), was trained to produce risk forecasts. Forecasts were made from a sliding window of 3‐month diary history for each day of each patient's diary. After training, the model parameters were held constant and the testing set was scored. A rate‐matched random (RMR) forecast was compared to the AI. Comparisons were made using the area under the receiver operating characteristic curve (AUC), a measure of binary discrimination performance, and the Brier score, a measure of forecast calibration. The Brier skill score (BSS) measured the improvement of the AI Brier score compared to the benchmark RMR Brier score. Confidence intervals (CIs) on performance statistics were obtained via bootstrapping. Results The AUC was 0.86 (95% CI = 0.85–0.88) for AI and 0.83 (95% CI = 0.81–0.85) for RMR, favoring AI (p < 0.001). Overall (all patients combined), BSS was 0.27 (95% CI = 0.23–0.31), also favoring AI (p < 0.001). Interpretation The AI produced a valid forecast superior to a chance forecaster, and provided meaningful forecasts in the majority of patients. Future studies will be needed to quantify the clinical value of these forecasts for patients. ANN NEUROL 2020;88:588–595

BackgroundThere is currently no formal method for predicting the range expected in an individual's seizure counts. Having access to such a prediction would be of benefit for developing more efficient clinical trials, but also for improving clinical care in the outpatient setting.MethodsUsing three independently collected patient diary datasets, we explored the predictability of seizure frequency. Three independent seizure diary databases were explored: SeizureTracker (n = 3016), Human Epilepsy Project (n = 93), and NeuroVista (n = 15). First, the relationship between mean and standard deviation in seizure frequency was assessed. Using that relationship, a prediction for the range of possible seizure frequencies was compared with a traditional prediction scheme commonly used in clinical trials. A validation dataset was obtained from a separate data export of SeizureTracker to further verify the predictions.ResultsA consistent mathematical relationship was observed across datasets. The logarithm of the average seizure count was linearly related to the logarithm of the standard deviation with a high correlation (R 2 > 0.83). The three datasets showed high predictive accuracy for this log–log relationship of 94%, compared with a predictive accuracy of 77% for a traditional prediction scheme. The independent validation set showed that the log–log predicted 94% of the correct ranges while the RR50 predicted 77%.ConclusionReliably predicting seizure frequency variability is straightforward based on knowledge of mean seizure frequency, across several datasets. With further study, this may help to increase the power of RCTs, and guide clinical practice.