Sridhar Sunderam

Automated seizure abatement in humans using electrical stimulation

The need for novel, efficacious, antiseizure therapies is widely acknowledged. This study investigates in humans the feasibility, safety, and efficacy of high‐frequency electrical stimulation (HFES; 100–500Hz) triggered by automated seizure detections. Eight patients were enrolled in this study, which consisted of a control and an experimental phase. HFES was delivered directly to the epileptogenic zone (local closed‐loop) in four patients and indirectly, through anterior thalami (remote closed‐loop), to the other four patients for every other automated seizure detection made by a validated algorithm. Interphase (control vs experimental phase) and intraphase (stimulated vs nonstimulated) comparisons of clinical seizure rate and relative severity (clinical and electrographic) were performed, and differences were assessed using effect size. Patients were deemed “responders” if seizure rate was reduced by at least 50%; the remaining patients were deemed “nonresponders.” All patients completed the study; rescue medications were not required. There were 1,491 HFESs (0.2% triggered after‐discharges). Mean change in seizure rate in the local closed‐loop group was −55.5% (−100 to +36.8%); three of four responders had a mean change of −86% (−100 to −58.8%). In the remote closed‐loop, the mean change of seizure rate was −40.8% (−72.9 to +1.4%); two of four responders had a mean change of −74.3% (−75.6 to −72.9%). Mean effect size was zero in the local closed‐loop (responders: beneficial and medium to large in magnitude) and negligible in the remote closed‐loop group (responders: beneficial and medium to large). HFES effects on epileptogenic tissue were immediate and also outlasted the stimulation period. This study demonstrates the feasibility and short‐term safety of automated HFES for seizure blockage, and also raises the possibility that it may be beneficial in pharmaco‐resistant epilepsies. Ann Neurol 2005;57:258–268

An introduction to contingent (closed-loop) brain electrical stimulation for seizure blockage, to ultra-short-term clinical trials, and to multidimensional statistical analysis of therapeutic efficacy.

Summary Automated seizure blockage is a top research priority of the American Epilepsy Society. This delivery modality (referred to herein as contingent or closed loop) requires for implementation a seizure detection algorithm for control of delivery of therapy via a suitable device. The authors address the many potential advantages of this modality over conventional alternatives (periodic or continuous), and the challenges it poses in the design and analysis of trials to assess efficacy and safety—in the particular context of direct delivery of electrical stimulation to brain tissue. The experimental designs of closed-loop therapies are currently limited by ethical, technical, medical, and practical considerations. One type of design that has been used successfully in an in-hospital “closed-loop” trial using subjects undergoing epilepsy surgery evaluation as their own controls is discussed in detail. This design performs a two-way comparison of seizure intensity, duration, and extent of spread between the control (surgery evaluation) versus the experimental phase, and, within the experimental phase, between treated versus untreated seizures. The proposed statistical analysis is based on a linear model that accounts for possible circadian effects, changes in treatment protocols, and other important factors such as change in seizure probability. The analysis is illustrated using seizure intensity as one of several possible end points from one of the subjects who participated in this trial. In-hospital ultra-short-term trials to assess safety and efficacy of closed-loop delivery of electrical stimulation for seizure blockage are both feasible and valuable.

Toward rational design of electrical stimulation strategies for epilepsy control

Electrical stimulation is emerging as a viable alternative for patients with epilepsy whose seizures are not alleviated by drugs or surgery. Its attractions are temporal and spatial specificity of action, flexibility of waveform parameters and timing, and the perception that its effects are reversible unlike resective surgery. However, despite significant advances in our understanding of mechanisms of neural electrical stimulation, clinical electrotherapy for seizures relies heavily on empirical tuning of parameters and protocols. We highlight concurrent treatment goals with potentially conflicting design constraints that must be resolved when formulating rational strategies for epilepsy electrotherapy, namely, seizure reduction versus cognitive impairment, stimulation efficacy versus tissue safety, and mechanistic insight versus clinical pragmatism. First, treatment markers, objectives, and metrics relevant to electrical stimulation for epilepsy are discussed from a clinical perspective. Then the experimental perspective is presented, with the biophysical mechanisms and modalities of open-loop electrical stimulation, and the potential benefits of closed-loop control for epilepsy.

Rapid Eye Movement Sleep and Hippocampal Theta Oscillations Precede Seizure Onset in the Tetanus Toxin Model of Temporal Lobe Epilepsy

Improved understanding of the interaction between state of vigilance (SOV) and seizure onset has therapeutic potential. Six rats received injections of tetanus toxin (TeTX) in the ventral hippocampus that resulted in chronic spontaneous seizures. The distribution of SOV before 486 seizures was analyzed for a total of 19 d of recording. Rapid eye movement sleep (REM) and exploratory wake, both of which express prominent hippocampal theta rhythm, preceded 47 and 34%, for a total of 81%, of all seizures. Nonrapid eye movement sleep (NREM) and nonexploratory wake, neither of which expresses prominent theta, preceded 6.8 and 13% of seizures. We demonstrate that identification of SOV yields significant differentiation of seizure susceptibilities, with the instantaneous seizure rate during REM nearly 10 times higher than baseline and the rate for NREM less than half of baseline. Survival analysis indicated a shorter duration of preseizure REM bouts, with a maximum transition to seizure at ∼90 s after the onset of REM. This study provides the first analysis of a correlation between SOV and seizure onset in the TeTX model of temporal lobe epilepsy, as well as the first demonstration that hippocampal theta rhythms associated with natural behavioral states can serve a seizure-promoting role. Our findings are in contrast with previous studies suggesting that the correlations between SOV and seizures are primarily governed by circadian oscillations and the notion that hippocampal theta rhythms inhibit seizures. The documentation of significant SOV-dependent seizure susceptibilities indicates the potential utility of SOV and its time course in seizure prediction and control.

Improved sleep-wake and behavior discrimination using MEMS accelerometers.

State of vigilance is determined by behavioral observations and electrophysiological activity. Here, we improve automatic state of vigilance discrimination by combining head acceleration with EEG measures. We incorporated biaxial dc-sensitive microelectromechanical system (MEMS) accelerometers into head-mounted preamplifiers in rodents. Epochs (15s) of behavioral video and EEG data formed training sets for the following states: Slow Wave Sleep, Rapid Eye Movement Sleep, Quiet Wakefulness, Feeding or Grooming, and Exploration. Multivariate linear discriminant analysis of EEG features with and without accelerometer features was used to classify behavioral state. A broad selection of EEG feature sets based on recent literature on state discrimination in rodents was tested. In all cases, inclusion of head acceleration significantly improved the discriminative capability. Our approach offers a novel methodology for determining the behavioral context of EEG in real time, and has potential application in automatic sleep-wake staging and in neural prosthetic applications for movement disorders and epileptic seizures.

Stochastic modeling and prediction of experimental seizures in Sprague-Dawley rats.

Summary: Most seizure prediction methods are based on nonlinear dynamic techniques, which are highly computationally expensive, thus limiting their clinical usefulness. The authors propose a different approach for prediction that uses a stochastic Markov chain model. Seizure (Ts) and interictal (Ti) durations were measured from 11 rats treated with 3‐mercaptopropionic acid. The duration of a seizure Ts was used to predict the time (Ti2) to the next one. Ts and Ti were distributed bimodally into short (S) and long (L), generating four probable transitions: S → S, S → L, L → S, and L → L. The joint probability density f (Ts, Ti2) was modeled, and was used to predict Ti2 given Ts. An identical model predicted Ts given the duration Ti1 of the preceding interictal interval. The median prediction error was 3.0 ± 3.5 seconds for Ts (given Ti1) and 6.5 ± 2.0 seconds for Ti2 (given Ts). In comparison, ranges for observed values were 2.3 seconds < Ts < 120 seconds and 6.6 seconds < Ti < 782 seconds. These results suggest that stochastic models are potentially useful tools for the prediction of seizures. Further investigation of the probable temporal interdependence between the ictal and interictal states may provide valuable insight into the dynamics of the epileptic brain.

A Novel Approach to the Study of Hypoxia-Ischemia-Induced Clinical and Subclinical Seizures in the Neonatal Rat

Perinatal hypoxic-ischemic encephalopathy (HIE) is a major cause of acute mortality and chronic neurologic morbidity in infants and children. HIE is the most common cause of neonatal seizures, and seizure activity in neonates can be clinical, with both EEG and behavioral symptoms, subclinical with only EEG activity, or just behavioral. The accurate detection of these different seizure manifestations and the extent to which they differ in their effects on the neonatal brain continues to be a concern in neonatal medicine. Most experimental studies of the interaction between hypoxia-ischemia (HI) and seizures have utilized a chemical induction of seizures, which may be less clinically relevant. Here, we expanded our model of unilateral cerebral HI in the immature rat to include video EEG and electromyographic recording before, during and after HI in term-equivalent postnatal-day-12 rats. We observed that immature rats display both clinical and subclinical seizures during the period of HI, and that the total number of seizures and time to first seizure correlate with the extent of tissue damage. We also tested the feasibility of developing an automated seizure detection algorithm for the unbiased detection and characterization of the different types of seizure activity observed in this model.

Seizure entrainment with polarizing low-frequency electric fields in a chronic animal epilepsy model

Neural activity can be modulated by applying a polarizing low-frequency (<<100 Hz) electric field (PLEF). Unlike conventional pulsed stimulation, PLEF stimulation has a graded, modulatory effect on neuronal excitability, and permits the simultaneous recording of neuronal activity during stimulation suitable for continuous feedback control. We tested a prototype system that allows for simultaneous PLEF stimulation with minimal recording artifact in a chronic tetanus toxin animal model (rat) of hippocampal epilepsy with spontaneous seizures. Depth electrode local field potentials recorded during seizures revealed a characteristic pattern of field postsynaptic potentials (fPSPs). Sinusoidal voltage-controlled PLEF stimulation (0.5-25 Hz) was applied in open-loop cycles radially across the CA3 of ventral hippocampus. For stimulated seizures, fPSPs were transiently entrained with the PLEF waveform. Statistical significance of entrainment was assessed with Thomsons harmonic F-test, with 45/132 stimulated seizures in four animals individually demonstrating significant entrainment (p < 0.04). Significant entrainment for multiple presentations at the same frequency (p < 0.01) was observed in three of four animals in 42/64 stimulated seizures. This is the first demonstration in chronically implanted freely behaving animals of PLEF modulation of neural activity with simultaneous recording.

Quasi-supervised scoring of human sleep in polysomnograms using augmented input variables

The limitations of manual sleep scoring make computerized methods highly desirable. Scoring errors can arise from human rater uncertainty or inter-rater variability. Sleep scoring algorithms either come as supervised classifiers that need scored samples of each state to be trained, or as unsupervised classifiers that use heuristics or structural clues in unscored data to define states. We propose a quasi-supervised classifier that models observations in an unsupervised manner but mimics a human rater wherever training scores are available. EEG, EMG, and EOG features were extracted in 30s epochs from human-scored polysomnograms recorded from 42 healthy human subjects (18-79 years) and archived in an anonymized, publicly accessible database. Hypnograms were modified so that: 1. Some states are scored but not others; 2. Samples of all states are scored but not for transitional epochs; and 3. Two raters with 67% agreement are simulated. A framework for quasi-supervised classification was devised in which unsupervised statistical models-specifically Gaussian mixtures and hidden Markov models--are estimated from unlabeled training data, but the training samples are augmented with variables whose values depend on available scores. Classifiers were fitted to signal features incorporating partial scores, and used to predict scores for complete recordings. Performance was assessed using Cohens Κ statistic. The quasi-supervised classifier performed significantly better than an unsupervised model and sometimes as well as a completely supervised model despite receiving only partial scores. The quasi-supervised algorithm addresses the need for classifiers that mimic scoring patterns of human raters while compensating for their limitations.

Naive scoring of human sleep based on a hidden Markov model of the electroencephalogram.

Clinical sleep scoring involves tedious visual review of overnight polysomnograms by a human expert. Many attempts have been made to automate the process by training computer algorithms such as support vector machines and hidden Markov models (HMMs) to replicate human scoring. Such supervised classifiers are typically trained on scored data and then validated on scored out-of-sample data. Here we describe a methodology based on HMMs for scoring an overnight sleep recording without the benefit of a trained initial model. The number of states in the data is not known a priori and is optimized using a Bayes information criterion. When tested on a 22-subject database, this unsupervised classifier agreed well with human scores (mean of Cohens kappa > 0.7). The HMM also outperformed other unsupervised classifiers (Gaussian mixture models, k-means, and linkage trees), that are capable of naive classification but do not model dynamics, by a significant margin (p <; 0.05).