Quantitative Electroencephalogram Trends Predict Recovery in Hypoxic-Ischemic Encephalopathy

OBJECTIVES

Electroencephalogram features predict neurologic recovery following cardiac arrest. Recent work has shown that prognostic implications of some key electroencephalogram features change over time. We explore whether time dependence exists for an expanded selection of quantitative electroencephalogram features and whether accounting for this time dependence enables better prognostic predictions.

DESIGN

Retrospective.

SETTING

ICUs at four academic medical centers in the United States.

PATIENTS

Comatose patients with acute hypoxic-ischemic encephalopathy.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

We analyzed 12,397 hours of electroencephalogram from 438 subjects. From the electroencephalogram, we extracted 52 features that quantify signal complexity, category, and connectivity. We modeled associations between dichotomized neurologic outcome (good vs poor) and quantitative electroencephalogram features in 12-hour intervals using sequential logistic regression with Elastic Net regularization. We compared a predictive model using time-varying features to a model using time-invariant features and to models based on two prior published approaches. Models were evaluated for their ability to predict binary outcomes using area under the receiver operator curve, model calibration (how closely the predicted probability of good outcomes matches the observed proportion of good outcomes), and sensitivity at several common specificity thresholds of interest. A model using time-dependent features outperformed (area under the receiver operator curve, 0.83 ± 0.08) one trained with time-invariant features (0.79 ± 0.07; p < 0.05) and a random forest approach (0.74 ± 0.13; p < 0.05). The time-sensitive model was also the best-calibrated.

CONCLUSIONS

The statistical association between quantitative electroencephalogram features and neurologic outcome changed over time, and accounting for these changes improved prognostication performance.

Reactivation of Motor-Related Gamma Activity in Human NREM Sleep

Models of memory consolidation posit a central role for reactivation of brain activity patterns during sleep, especially in non-Rapid Eye Movement (NREM) sleep. While such "replay" of recent waking experiences has been well-demonstrated in rodents, electrophysiological evidence of reactivation in human sleep is still largely lacking. In this intracranial study in patients with epilepsy (N = 9) we explored the spontaneous electroencephalographic reactivation during sleep of spatial patterns of brain activity evoked by motor learning. We first extracted the gamma-band (60-140 Hz) patterns underlying finger movements during a tapping task and underlying no-movement during a short rest period just prior to the task, and trained a binary classifier to discriminate between motor movements vs. rest. We then used the trained model on NREM sleep data immediately after the task and on NREM sleep during a control sleep period preceding the task. Compared with the control sleep period, we found, at the subject level, an increase in the detection rate of motor-related patterns during sleep following the task, but without association with performance changes. These data provide electrophysiological support for the reoccurrence in NREM sleep of the neural activity related to previous waking experience, i.e. that a basic tenet of the reactivation theory does occur in human sleep.

Artificial intelligence in sleep medicine: background and implications for clinicians

None

Polysomnography remains the cornerstone of objective testing in sleep medicine and results in massive amounts of electrophysiological data, which is well-suited for analysis with artificial intelligence (AI)-based tools. Combined with other sources of health data, AI is expected to provide new insights to inform the clinical care of sleep disorders and advance our understanding of the integral role sleep plays in human health. Additionally, AI has the potential to streamline day-to-day operations and therefore optimize direct patient care by the sleep disorders team. However, clinicians, scientists, and other stakeholders must develop best practices to integrate this rapidly evolving technology into our daily work while maintaining the highest degree of quality and transparency in health care and research. Ultimately, when harnessed appropriately in conjunction with human expertise, AI will improve the practice of sleep medicine and further sleep science for the health and well-being of our patients.

Excess brain age in the sleep electroencephalogram predicts reduced life expectancy

The brain age index (BAI) measures the difference between an individual's apparent "brain age" (BA; estimated by comparing EEG features during sleep from an individual with age norms), and their chronological age (CA); that is BAI = BA-CA. Here, we evaluate whether BAI predicts life expectancy. Brain age was quantified using a previously published machine learning algorithm for a cohort of participants ≥40 years old who underwent an overnight sleep electroencephalogram (EEG) as part of the Sleep Heart Health Study (n = 4877). Excess brain age (BAI >0) was associated with reduced life expectancy (adjusted hazard ratio: 1.12, [1.03, 1.21], p = 0.002). Life expectancy decreased by -0.81 [-1.44, -0.24] years per standard-deviation increase in BAI. Our findings show that BAI, a sleep EEG-based biomarker of the deviation of sleep microstructure from patterns normal for age, is an independent predictor of life expectancy.

Electrographic predictors of successful weaning from anaesthetics in refractory status epilepticus

Intravenous third-line anaesthetic agents are typically titrated in refractory status epilepticus to achieve either seizure suppression or burst suppression on continuous EEG. However, the optimum treatment paradigm is unknown and little data exist to guide the withdrawal of anaesthetics in refractory status epilepticus. Premature withdrawal of anaesthetics risks the recurrence of seizures, whereas the prolonged use of anaesthetics increases the risk of treatment-associated adverse effects. This study sought to measure the accuracy of features of EEG activity during anaesthetic weaning in refractory status epilepticus as predictors of successful weaning from intravenous anaesthetics. We prespecified a successful anaesthetic wean as the discontinuation of intravenous anaesthesia without developing recurrent status epilepticus, and a wean failure as either recurrent status epilepticus or the resumption of anaesthesia for the purpose of treating an EEG pattern concerning for incipient status epilepticus. We evaluated two types of features as predictors of successful weaning: spectral components of the EEG signal, and spatial-correlation-based measures of functional connectivity. The results of these analyses were used to train a classifier to predict wean outcome. Forty-seven consecutive anaesthetic weans (23 successes, 24 failures) were identified from a single-centre cohort of patients admitted with refractory status epilepticus from 2016 to 2019. Spectral components of the EEG revealed no significant differences between successful and unsuccessful weans. Analysis of functional connectivity measures revealed that successful anaesthetic weans were characterized by the emergence of larger, more densely connected, and more highly clustered spatial functional networks, yielding 75.5% (95% confidence interval: 73.1-77.8%) testing accuracy in a bootstrap analysis using a hold-out sample of 20% of data for testing and 74.6% (95% confidence interval 73.2-75.9%) testing accuracy in a secondary external validation cohort, with an area under the curve of 83.3%. Distinct signatures in the spatial networks of functional connectivity emerge during successful anaesthetic liberation in status epilepticus; these findings are absent in patients with anaesthetic wean failure. Identifying features that emerge during successful anaesthetic weaning may allow faster and more successful anaesthetic liberation after refractory status epilepticus.

Sensitivity of Continuous Electroencephalography to Detect Ictal Activity After Cardiac Arrest

IMPORTANCE

Epileptiform electroencephalographic (EEG) patterns are common after resuscitation from cardiac arrest, are associated with patient outcome, and may require treatment. It is unknown whether continuous EEG monitoring is needed to detect these patterns or if brief intermittent monitoring is sufficient. If continuous monitoring is required, the necessary duration of observation is unknown.

OBJECTIVE

To quantify the time-dependent sensitivity of continuous EEG for epileptiform event detection, and to compare continuous EEG to several alternative EEG-monitoring strategies for post-cardiac arrest outcome prediction.

DESIGN, SETTING, AND PARTICIPANTS

This observational cohort study was conducted in 2 academic medical centers between September 2010 and January 2018. Participants included 759 adults who were comatose after being resuscitated from cardiac arrest and who underwent 24 hours or more of EEG monitoring.

MAIN OUTCOMES AND MEASURES

Epileptiform EEG patterns associated with neurological outcome at hospital discharge, such as seizures likely to cause secondary injury.

RESULTS

Overall, 759 patients were included in the analysis; 281 (37.0%) were female, and the mean (SD) age was 58 (17) years. Epileptiform EEG activity was observed in 414 participants (54.5%), of whom only 26 (3.4%) developed potentially treatable seizures. Brief intermittent EEG had an estimated 66% (95% CI, 62%-69%) to 68% (95% CI, 66%-70%) sensitivity for detection of prognostic epileptiform events. Depending on initial continuity of the EEG background, 0 to 51 hours of monitoring were needed to achieve 95% sensitivity for the detection of prognostic epileptiform events. Brief intermittent EEG had a sensitivity of 7% (95% CI, 4%-12%) to 8% (95% CI, 4%-12%) for the detection of potentially treatable seizures, and 0 to 53 hours of continuous monitoring were needed to achieve 95% sensitivity for the detection of potentially treatable seizures. Brief intermittent EEG results yielded similar information compared with continuous EEG results when added to multivariable models predicting neurological outcome.

CONCLUSIONS AND RELEVANCE

Compared with continuous EEG monitoring, brief intermittent monitoring was insensitive for detection of epileptiform events. Monitoring EEG results significantly improved multimodality prediction of neurological outcome, but continuous monitoring appeared to add little additional information compared with brief intermittent monitoring.

Automatic Detection of General Anesthetic-States using ECG-Derived Autonomic Nervous System Features

Electroencephalogram (EEG)-based prediction systems are used to target anesthetic-states in patients undergoing procedures with general anesthesia (GA). These systems are not widely employed in resource-limited settings because they are cost-prohibitive. Although anesthetic-drugs induce highly-structured, oscillatory neural dynamics that make EEG-based systems a principled approach for anesthetic-state monitoring, anesthetic-drugs also significantly modulate the autonomic nervous system (ANS). Because ANS dynamics can be inferred from electrocardiogram (ECG) features such as heart rate variability, it may be possible to develop an ECG-based system to infer anesthetic-states as a low-cost and practical alternative to EEG-based anesthetic-state prediction systems. In this work, we demonstrate that an ECG-based system using ANS features can be used to discriminate between non-GA and GA states in sevoflurane, with a GA F₁ score of 0.834, [95% CI, 0.776, 0.892], and in sevoflurane-plus-ketamine, with a GA F₁ score of 0.880 [0.815, 0.954]. With further refinement, ECG-based anesthetic-state systems could be developed as a fully automated system for anesthetic-state monitoring in resource-limited settings.

Soluble ST2 Is Associated With New Epileptiform Abnormalities Following Nontraumatic Subarachnoid Hemorrhage

Background and Purpose- We evaluated the association between 2 types of predictors of delayed cerebral ischemia after nontraumatic subarachnoid hemorrhage, including biomarkers of the innate immune response and neurophysiologic changes on continuous electroencephalography. Methods- We studied subarachnoid hemorrhage patients that had at least 72 hours of continuous electroencephalography and blood samples collected within the first 5 days of symptom onset. We measured inflammatory biomarkers previously associated with delayed cerebral ischemia and functional outcome, including soluble ST2 (sST2), IL-6 (interleukin-6), and CRP (C-reactive protein). Serial plasma samples and cerebrospinal fluid sST2 levels were available in a subgroup of patients. Neurophysiologic changes were categorized into new or worsening epileptiform abnormalities (EAs) or new background deterioration. The association of biomarkers with neurophysiologic changes were evaluated using the Wilcoxon rank-sum test. Plasma and cerebrospinal fluid sST2 were further examined longitudinally using repeated measures mixed-effects models. Results- Forty-six patients met inclusion criteria. Seventeen (37%) patients developed new or worsening EAs, 21 (46%) developed new background deterioration, and 8 (17%) developed neither. Early (day, 0-5) plasma sST2 levels were higher among patients with new or worsening EAs (median 115 ng/mL [interquartile range, 73.8-197]) versus those without (74.7 ng/mL [interquartile range, 44.8-102]; P=0.024). Plasma sST2 levels were similar between patients with or without new background deterioration. Repeated measures mixed-effects modeling that adjusted for admission risk factors showed that the association with new or worsening EAs remained independent for both plasma sST2 (β=0.41 [95% CI, 0.09-0.73]; P=0.01) and cerebrospinal fluid sST2 (β=0.97 [95% CI, 0.14-1.8]; P=0.021). IL-6 and CRP were not associated with new background deterioration or with new or worsening EAs. Conclusions- In patients admitted with subarachnoid hemorrhage, sST2 level was associated with new or worsening EAs but not new background deterioration. This association may identify a link between a specific innate immune response pathway and continuous electroencephalography abnormalities in the pathogenesis of secondary brain injury after subarachnoid hemorrhage.

Burst Suppression: Causes and Effects on Mortality in Critical Illness

BACKGROUND

Burst suppression in mechanically ventilated intensive care unit (ICU) patients is associated with increased mortality. However, the relative contributions of propofol use and critical illness itself to burst suppression; of burst suppression, propofol, and critical illness to mortality; and whether preventing burst suppression might reduce mortality, have not been quantified.

METHODS

The dataset contains 471 adults from seven ICUs, after excluding anoxic encephalopathy due to cardiac arrest or intentional burst suppression for therapeutic reasons. We used multiple prediction and causal inference methods to estimate the effects connecting burst suppression, propofol, critical illness, and in-hospital mortality in an observational retrospective study. We also estimated the effects mediated by burst suppression. Sensitivity analysis was used to assess for unmeasured confounding.

RESULTS

The expected outcomes in a "counterfactual" randomized controlled trial (cRCT) that assigned patients to mild versus severe illness are expected to show a difference in burst suppression burden of 39%, 95% CI [8-66]%, and in mortality of 35% [29-41]%. Assigning patients to maximal (100%) burst suppression burden is expected to increase mortality by 12% [7-17]% compared to 0% burden. Burst suppression mediates 10% [2-21]% of the effect of critical illness on mortality. A high cumulative propofol dose (1316 mg/kg) is expected to increase burst suppression burden by 6% [0.8-12]% compared to a low dose (284 mg/kg). Propofol exposure has no significant direct effect on mortality; its effect is entirely mediated through burst suppression.

CONCLUSIONS

Our analysis clarifies how important factors contribute to mortality in ICU patients. Burst suppression appears to contribute to mortality but is primarily an effect of critical illness rather than iatrogenic use of propofol.

EEG Functional Connectivity is a Weak Predictor of Causal Brain Interactions

In recent years there has been an explosion of research evaluating resting-state brain functional connectivity (FC) using different modalities. However, the relationship between such measures of FC and the underlying causal brain interactions has not been well characterized. To further characterize this relationship, we assessed the relationship between electroencephalography (EEG) resting state FC and propagation of transcranial magnetic stimulation (TMS) evoked potentials (TEPs) at the sensor and source level in healthy participants. TMS was applied to six different cortical regions in ten healthy individuals (9 male; 1 female), and effects on brain activity were measured using simultaneous EEG. Pre-stimulus FC was assessed using five different FC measures (Pearson's correlation, mutual information, weighted phase lag index, coherence and phase locking value). Propagation of the TEPs was quantified as the root mean square (RMS) of the TEP voltage and current source density (CSD) at the sensor and source level, respectively. The relationship between pre-stimulus FC and the spatial distribution of TEP activity was determined using a generalized linear model (GLM) analysis. On the group level, all FC measures correlated significantly with TEP activity over the early (15-75 ms) and full range (15-400 ms) of the TEP at the sensor and source level. However, the predictive value of all FC measures is quite limited, accounting for less than 10% of the variance of TEP activity, and varies substantially across participants and stimulation sites. Taken together, these results suggest that EEG functional connectivity studies in sensor and source space should be interpreted with caution.

Abstract TMP22: Dynamic Worsening of Epileptiform Abnormalities Following Subarachnoid Hemorrhage is Associated With Poor 3-Month Outcomes

Objective: Worsening epileptiform abnormalities (EAs) and deteriorating background activity are common continuous electroencephalography (cEEG) patterns that predict subsequent clinical deterioration following subarachnoid hemorrhage (SAH). While worsening EAs and background deterioration both imply cortical dysfunction, we sought to clarify if these patterns have a different association with clinical outcomes.

Methods: We enrolled patients with SAH undergoing > 3 days of cEEG monitoring enrolled in a prospective outcome study with a modified Rankin Scale (mRS) assessment at 3 months. Worsening EAs included new or increasing burden of sporadic epileptiform discharges, lateralized rhythmic delta activity (LRDA), lateralized periodic discharges (LPD), or generalized periodic discharges (GPD). Background deterioration was defined as decreasing Alpha Delta Ratio (ADR), Relative Alpha Variability (RAV) or worsening focal slowing. We evaluated the association between these cEEG patterns and 3-month mRS >3 and examined whether the influence on outcome was independent of delayed cerebral ischemia (DCI).

Results: Of 59 patients meeting inclusion criteria (3-month mRS median 3 [IQR 1-5]), worsening EAs developed in 23 (39%) and new background deterioration in 24 (41%), whereas 24 patients (41%) developed neither finding and 12 (20%) developed both. Patients with worsening EAs were more likely to have a poor 3-month mRS compared to those without worsening EAs (OR 6.44; 95%CI 1.99-20.9; p=0.001). Developing new background deterioration was not significantly associated with poor 3-month outcome (OR 1.56, 95%CI 0.53-4.59; p=0.42). There was no additive effect on poor outcome of developing both findings. In a multivariate logistic regression, the effect of worsening EAs on 3-month mRS was independent of DCI.

Interpretation: While both worsening EAs and new background deterioration have previously been associated with DCI, only worsening EAs influences poor long-term outcome. Further investigation may clarify if distinct mechanisms underlie these differences.

Early Prediction of Sepsis From Clinical Data: The PhysioNet/Computing in Cardiology Challenge 2019

OBJECTIVES

Sepsis is a major public health concern with significant morbidity, mortality, and healthcare expenses. Early detection and antibiotic treatment of sepsis improve outcomes. However, although professional critical care societies have proposed new clinical criteria that aid sepsis recognition, the fundamental need for early detection and treatment remains unmet. In response, researchers have proposed algorithms for early sepsis detection, but directly comparing such methods has not been possible because of different patient cohorts, clinical variables and sepsis criteria, prediction tasks, evaluation metrics, and other differences. To address these issues, the PhysioNet/Computing in Cardiology Challenge 2019 facilitated the development of automated, open-source algorithms for the early detection of sepsis from clinical data.

DESIGN

Participants submitted containerized algorithms to a cloud-based testing environment, where we graded entries for their binary classification performance using a novel clinical utility-based evaluation metric. We designed this scoring function specifically for the Challenge to reward algorithms for early predictions and penalize them for late or missed predictions and for false alarms.

SETTING

ICUs in three separate hospital systems. We shared data from two systems publicly and sequestered data from all three systems for scoring.

PATIENTS

We sourced over 60,000 ICU patients with up to 40 clinical variables for each hour of a patient's ICU stay. We applied Sepsis-3 clinical criteria for sepsis onset.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

A total of 104 groups from academia and industry participated, contributing 853 submissions. Furthermore, 90 abstracts based on Challenge entries were accepted for presentation at Computing in Cardiology.

CONCLUSIONS

Diverse computational approaches predict the onset of sepsis several hours before clinical recognition, but generalizability to different hospital systems remains a challenge.

Effects of Cholinergic Neuromodulation on Thalamocortical Rhythms During NREM Sleep: A Model Study

It has been suggested that cholinergic neurons shape the oscillatory activity of the thalamocortical (TC) network in behavioral and electrophysiological experiments. However, theoretical modeling demonstrating how cholinergic neuromodulation of thalamocortical rhythms during non-rapid eye movement (NREM) sleep might occur has been lacking. In this paper, we first develop a novel computational model (TC-ACH) by incorporating a cholinergic neuron population (CH) into the classical thalamo-cortical circuitry, where connections between populations are modeled in accordance with existing knowledge. The neurotransmitter acetylcholine (ACH) released by neurons in CH, which is able to change the discharge activity of thalamocortical neurons, is the primary focus of our work. Simulation results with our TC-ACH model reveal that the cholinergic projection activity is a key factor in modulating oscillation patterns in three ways: (1) transitions between different patterns of thalamocortical oscillations are dramatically modulated through diverse projection pathways; (2) the model expresses a stable spindle oscillation state with certain parameter settings for the cholinergic projection from CH to thalamus, and more spindles appear when the strength of cholinergic input from CH to thalamocortical neurons increases; (3) the duration of oscillation patterns during NREM sleep including K-complexes, spindles, and slow oscillations is longer when cholinergic input from CH to thalamocortical neurons becomes stronger. Our modeling results provide insights into the mechanisms by which the sleep state is controlled, and provide a theoretical basis for future experimental and clinical studies.

Seizure tracking of epileptic EEGs using a model-driven approach

OBJECTIVE

As a chronic neurological disorder, epilepsy is characterized by recurrent and unprovoked epileptic seizures that can disrupt the normal neuro-biologic, cognitive, psychological conditions of patients. Therefore, it is worthwhile to give a detailed account of how the epileptic EEG evolves during a period of seizure so that an effective control can be guided for epileptic patients in clinics.

APPROACH

Considering the successful application of the neural mass model (NMM) in exploring the insights into brain activities for epilepsy, in this paper, we aim to construct a model-driven approach to track the development of seizures using epileptic EEGs. We first propose a new time-delay Wendling model with sub-populations (TD-W-SP model) with respect to three aspects of improvements. Then we introduce a model-driven seizure tracking approach, where a model training method is designed based on extracted features from epileptic EEGs and a tracking index is defined as a function of the trained model parameters.

MAIN RESULTS

Numerical results on eight patients on CHB-MIT database demonstrate that our proposed method performs well in simulating epileptic-like EEGs as well as tracking the evolution of three stages (that is, from pre-ictal to ictal and from ictal to post-ictal) during a period of epileptic seizure.

SIGNIFICANCE

A useful attempt to track epileptic seizures by combining the NMM with the data analysis.

Detection of mesial temporal lobe epileptiform discharges on intracranial electrodes using deep learning

OBJECTIVE

Develop a high-performing algorithm to detect mesial temporal lobe (mTL) epileptiform discharges on intracranial electrode recordings.

METHODS

An epileptologist annotated 13,959 epileptiform discharges from a dataset of intracranial EEG recordings from 46 epilepsy patients. Using this dataset, we trained a convolutional neural network (CNN) to recognize mTL epileptiform discharges from a single intracranial bipolar channel. The CNN outputs from multiple bipolar channel inputs were averaged to generate the final detector output. Algorithm performance was estimated using a nested 5-fold cross-validation.

RESULTS

On the receiver-operating characteristic curve, our algorithm achieved an area under the curve (AUC) of 0.996 and a partial AUC (for specificity > 0.9) of 0.981. AUC on a precision-recall curve was 0.807. A sensitivity of 84% was attained at a false positive rate of 1 per minute. 35.9% of the false positive detections corresponded to epileptiform discharges that were missed during expert annotation.

CONCLUSIONS

Using deep learning, we developed a high-performing, patient non-specific algorithm for detection of mTL epileptiform discharges on intracranial electrodes.

SIGNIFICANCE

Our algorithm has many potential applications for understanding the impact of mTL epileptiform discharges in epilepsy and on cognition, and for developing therapies to specifically reduce mTL epileptiform activity.

A novel neural computational model of generalized periodic discharges in acute hepatic encephalopathy

Acute hepatic encephalopathy (AHE) due to acute liver failure is a common form of delirium, a state of confusion, impaired attention, and decreased arousal. The electroencephalogram (EEG) in AHE often exhibits a striking abnormal pattern of brain activity, which epileptiform discharges repeat in a regular repeating pattern. This pattern is known as generalized periodic discharges, or triphasic-waves (TPWs). While much is known about the neurophysiological mechanisms underlying AHE, how these mechanisms relate to TPWs is poorly understood. In order to develop hypotheses how TPWs arise, our work builds a computational model of AHE (AHE-CM), based on three modifications of the well-studied Liley model which emulate mechanisms believed central to brain dysfunction in AHE: increased neuronal excitability, impaired synaptic transmission, and enhanced postsynaptic inhibition. To relate our AHE-CM to clinical EEG data from patients with AHE, we design a model parameter optimization method based on particle filtering (PF-POM). Based on results from 7 AHE patients, we find that the proposed AHE-CM not only performs well in reproducing important aspects of the EEG, namely the periodicity of triphasic waves (TPWs), but is also helpful in suggesting mechanisms underlying variation in EEG patterns seen in AHE. In particular, our model helps explain what conditions lead to increased frequency of TPWs. In this way, our model represents a starting point for exploring the underlying mechanisms of brain dynamics in delirium by relating microscopic mechanisms to EEG patterns.

Expert-level sleep scoring with deep neural networks

Objectives

Scoring laboratory polysomnography (PSG) data remains a manual task of visually annotating 3 primary categories: sleep stages, sleep disordered breathing, and limb movements. Attempts to automate this process have been hampered by the complexity of PSG signals and physiological heterogeneity between patients. Deep neural networks, which have recently achieved expert-level performance for other complex medical tasks, are ideally suited to PSG scoring, given sufficient training data.

Methods

We used a combination of deep recurrent and convolutional neural networks (RCNN) for supervised learning of clinical labels designating sleep stages, sleep apnea events, and limb movements. The data for testing and training were derived from 10 000 clinical PSGs and 5804 research PSGs.

Results

When trained on the clinical dataset, the RCNN reproduces PSG diagnostic scoring for sleep staging, sleep apnea, and limb movements with accuracies of 87.6%, 88.2% and 84.7% on held-out test data, a level of performance comparable to human experts. The RCNN model performs equally well when tested on the independent research PSG database. Only small reductions in accuracy were noted when training on limited channels to mimic at-home monitoring devices: frontal leads only for sleep staging, and thoracic belt signals only for the apnea-hypopnea index.

Conclusions

By creating accurate deep learning models for sleep scoring, our work opens the path toward broader and more timely access to sleep diagnostics. Accurate scoring automation can improve the utility and efficiency of in-lab and at-home approaches to sleep diagnostics, potentially extending the reach of sleep expertise beyond specialty clinics.

A Mean Field Model of Acute Hepatic Encephalopathy

Acute hepatic encephalopathy (AHE) is a common form of delirium, a state of confusion, impaired attention, and decreased arousal due to acute liver failure. However, the neurophysiological mechanisms underlying AHE are poorly understood. In order to develop hypotheses for mechanisms of AHE, our work builds on an existing neural mean field model for similar EEG patterns in cerebral anoxia, the bursting Liley model. The model proposes that generalized periodic discharges, similar to the triphasic waves (TPWs) seen in severe AHE, arise through three types of processes a) increased neuronal excitability; b) defective brain energy metabolism leading to impaired synaptic transmission; c) and enhanced postsynaptic inhibition mediated by increased GABA-ergic and glycinergic transmission. We relate the model parameters to human EEG data using a particle-filter based optimization method that matches the TPW inter-event-interval distribution of the model with that observed in patients EEGs. In this way our model relates microscopic mechanisms to EEG patterns. Our model represents a starting point for exploring the underlying mechanisms of brain dynamics in delirium.

Rapid Annotation of Seizures and Interictal-ictal Continuum EEG Patterns

Seizures, status epilepticus, and seizure-like rhythmic or periodic activities are common, pathological, harmful states of brain electrical activity seen in the electroencephalogram (EEG) of patients during critical medical illnesses or acute brain injury. Accumulating evidence shows that these states, when prolonged, cause neurological injury. In this study we developed a valid method to automatically discover a small number of homogeneous pattern clusters, to facilitate efficient interactive labelling by EEG experts. 592 time domain and spectral features were extracted from continuous EEG (cEEG) data of 369 ICU (intensive care unit) patients. For each patient, feature dimensionality was reduced using principal component analysis (PCA), retaining 95% of the variance. K-medoids clustering was applied to learn a local dictionary from each patient, consisting of k=100 exemplars/words. Changepoint detection (CPD) was utilized to break each EEG into segments. A bag-of-words (BoW) representation was computed for each segment, specifically, a normalized histogram of the words found within each segment. Segments were further clustered using the BoW histograms by Affinity Propagation (AP) using a χ² distance to measure similarities between histograms. The resulting 30 50 clusters for each patient were scored by EEG experts through labeling only the cluster medoids. Embedding methods t-SNE (t-distributed stochastic neighbor embedding) and PCA were used to provide a 2D representation for visualization and exploration of the data. Our results illustrate that it takes approximately 3 minutes to annotate 24 hours of cEEG by experts, which is at least 60 times faster than unaided manual review.

EEG CLassification Via Convolutional Neural Network-Based Interictal Epileptiform Event Detection

Diagnosis of epilepsy based on visual inspection of electroencephalogram (EEG) abnormalities is an inefficient, time-consuming, and expert-centered process. Moreover, the diagnosis based on ictal epileptiform events is challenging as the ictal patterns are infrequent. Consequently, the development of an automated, fast, and reliable epileptic EEG diagnostic system is essential. The interictal epileptiform discharges (IEDs) are recurring patterns that are highly suggestive of epilepsy. In this paper, we propose an epileptic EEG classification system based on IED detection. The proposed system comprises of three modules: pre-processing, waveform-level classification, and EEG-level classification. We employ a Convolutional Neural Network (CNN) for waveform-level classification and a Support Vector Machine (SVM) for EEG-level classification. We evaluated the proposed system on a dataset of 156 EEGs recorded at Massachusetts General Hospital (MGH), Boston. The system achieved a mean 4-fold classification accuracy of 83.86% for classifying EEGs with and without IEDs.

Quantitative EEG reactivity and machine learning for prognostication in hypoxic-ischemic brain injury

OBJECTIVE

Electroencephalogram (EEG) reactivity is a robust predictor of neurological recovery after cardiac arrest, however interrater-agreement among electroencephalographers is limited. We sought to evaluate the performance of machine learning methods using EEG reactivity data to predict good long-term outcomes in hypoxic-ischemic brain injury.

METHODS

We retrospectively reviewed clinical and EEG data of comatose cardiac arrest subjects. Electroencephalogram reactivity was tested within 72 h from cardiac arrest using sound and pain stimuli. A Quantitative EEG (QEEG) reactivity method evaluated changes in QEEG features (EEG spectra, entropy, and frequency features) during the 10 s before and after each stimulation. Good outcome was defined as Cerebral Performance Category of 1-2 at six months. Performance of a random forest classifier was compared against a penalized general linear model (GLM) and expert electroencephalographer review.

RESULTS

Fifty subjects were included and sixteen (32%) had good outcome. Both QEEG reactivity methods had comparable performance to expert EEG reactivity assessment for good outcome prediction (mean AUC 0.8 for random forest vs. 0.69 for GLM vs. 0.69 for expert review, respectively; p non-significant).

CONCLUSIONS

Machine-learning models utilizing quantitative EEG reactivity data can predict long-term outcome after cardiac arrest.

SIGNIFICANCE

A quantitative approach to EEG reactivity assessment may support prognostication in cardiac arrest.

Clinical EEG slowing correlates with delirium severity and predicts poor clinical outcomes

OBJECTIVE

To determine which findings on routine clinical EEGs correlate with delirium severity across various presentations and to determine whether EEG findings independently predict important clinical outcomes.

METHODS

We prospectively studied a cohort of nonintubated inpatients undergoing EEG for evaluation of altered mental status. Patients were assessed for delirium within 1 hour of EEG with the 3-Minute Diagnostic Interview for Confusion Assessment Method (3D-CAM) and 3D-CAM severity score. EEGs were interpreted clinically by neurophysiologists, and reports were reviewed to identify features such as theta or delta slowing and triphasic waves. Generalized linear models were used to quantify associations among EEG findings, delirium, and clinical outcomes, including length of stay, Glasgow Outcome Scale scores, and mortality.

RESULTS

We evaluated 200 patients (median age 60 years, IQR 48.5-72 years); 121 (60.5%) met delirium criteria. The EEG finding most strongly associated with delirium presence was a composite of generalized theta or delta slowing (odds ratio 10.3, 95% confidence interval 5.3-20.1). The prevalence of slowing correlated not only with overall delirium severity (R 2 = 0.907) but also with the severity of each feature assessed by CAM-based delirium algorithms. Slowing was common in delirium even with normal arousal. EEG slowing was associated with longer hospitalizations, worse functional outcomes, and increased mortality, even after adjustment for delirium presence or severity.

CONCLUSIONS

Generalized slowing on routine clinical EEG strongly correlates with delirium and may be a valuable biomarker for delirium severity. In addition, generalized EEG slowing should trigger elevated concern for the prognosis of patients with altered mental status.

Randomized trial of lacosamide versus fosphenytoin for nonconvulsive seizures

OBJECTIVE

The optimal treatment of nonconvulsive seizures in critically ill patients is uncertain. We evaluated the comparative effectiveness of the antiseizure drugs lacosamide (LCM) and fosphenytoin (fPHT) in this population.

METHODS

The TRENdS (Treatment of Recurrent Electrographic Nonconvulsive Seizures) study was a noninferiority, prospective, multicenter, randomized treatment trial of patients diagnosed with nonconvulsive seizures (NCSs) by continuous electroencephalography (cEEG). Treatment was randomized to intravenous (IV) LCM 400mg or IV fPHT 20mg phenytoin equivalents/kg. The primary endpoint was absence of electrographic seizures for 24 hours as determined by 1 blinded EEG reviewer. The frequency with which NCS control was achieved in each arm was compared, and the 90% confidence interval (CI) was determined. Noninferiority of LCM to fPHT was to be concluded if the lower bound of the CI for relative risk was >0.8.

RESULTS

Seventy-four subjects were enrolled (37 LCM, 37 fPHT) between August 21, 2012 and December 20, 2013. The mean age was 63.6 years; 38 were women. Seizures were controlled in 19 of 30 (63.3%) subjects in the LCM arm and 16 of 32 (50%) subjects in the fPHT arm. LCM was noninferior to fPHT (p = 0.02), with a risk ratio of 1.27 (90% CI = 0.88-1.83). Treatment emergent adverse events (TEAEs) were similar in both arms, occurring in 9 of 35 (25.7%) LCM and 9 of 37 (24.3%) fPHT subjects (p = 1.0).

INTERPRETATION

LCM was noninferior to fPHT in controlling NCS, and TEAEs were comparable. LCM can be considered an alternative to fPHT in the treatment of NCSs detected on cEEG. Ann Neurol 2018;83:1174-1185.

A fast machine learning approach to facilitate the detection of interictal epileptiform discharges in the scalp electroencephalogram

BACKGROUND

Finding interictal epileptiform discharges (IEDs) in the EEG is a part of diagnosing epilepsy. Automated software for annotating EEGs of patients with suspected epilepsy can therefore help with reaching a diagnosis. A large amount of data is required for training and evaluating an effective IED detection system. IEDs occur infrequently in the most patients' EEG, therefore, interictal EEG recordings contain mostly background waveforms.

NEW METHOD

As the first step to detect IEDs, we propose a machine learning technique eliminating most EEG background data using an ensemble of simple fast classifiers based on several EEG features. This could save computation time for an IED detection method, allowing the remaining waveforms to be classified by more computationally intensive methods. We consider several efficient features and reject background by applying thresholds on them in consecutive steps.

RESULTS

We applied the proposed algorithm on a dataset of 156 EEGs (93 and 63 with and without IEDs, respectively). We were able to eliminate 78% of background waveforms while retaining 97% of IEDs on our cross-validated dataset.

COMPARISON WITH EXISTING METHODS

We applied support vector machine, k-nearest neighbours, and random forest classifiers to detect IEDs with and without initial background rejection. Results show that rejecting background by our proposed method speeds up the overall classification by a factor ranging from 1.8 to 4.7 for the considered classifiers.

CONCLUSIONS

The proposed method successfully reduces computation time of an IED detection system. Therefore, it is beneficial in speeding up IED detection especially when utilizing large EEG datasets.

Drug-Specific Models Improve the Performance of an EEG-based Automated Brain-State Prediction System

Maintaining anesthetic states using automated brain-state prediction systems is expected to reduce drug overdosage and associated side-effects. However, commercially available brain-state monitoring systems perform poorly on drug-class combinations. We assume that current automated brain-state prediction systems perform poorly because they do not account for brain-state dynamics that are unique to drug-class combinations. In this work, we develop a k-nearest neighbors model to test whether improvements to automated brain-state prediction of drug-class combinations are feasible. We utilize electroencephalogram data collected from human subjects who received general anesthesia with sevoflurane and general anesthesia with the drug-class combination of sevoflurane-plus-ketamine. We demonstrate improved performance predicting anesthesia-induced brain-states using drug-specific models.

Comparison of machine learning models for seizure prediction in hospitalized patients

Objective

To compare machine learning methods for predicting inpatient seizures risk and determine the feasibility of 1‐h screening EEG to identify low‐risk patients (<5% seizures risk in 48 h).

Methods

The Critical Care EEG Monitoring Research Consortium (CCEMRC) multicenter database contains 7716 continuous EEGs (cEEG). Neural networks (NN), elastic net logistic regression (EN), and sparse linear integer model (RiskSLIM) were trained to predict seizures. RiskSLIM was used previously to generate 2HELPS2B model of seizure predictions. Data were divided into training (60% for model fitting) and evaluation (40% for model evaluation) cohorts. Performance was measured using area under the receiver operating curve (AUC), mean risk calibration (CAL), and negative predictive value (NPV). A secondary analysis was performed using Monte Carlo simulation (MCS) to normalize all EEG recordings to 48 h and use only the first hour of EEG as a “screening EEG” to generate predictions.

Results

RiskSLIM recreated the 2HELPS2B model. All models had comparable AUC: evaluation cohort (NN: 0.85, EN: 0.84, 2HELPS2B: 0.83) and MCS (NN: 0.82, EN; 0.82, 2HELPS2B: 0.81) and NPV (absence of seizures in the group that the models predicted to be low risk): evaluation cohort (NN: 97%, EN: 97%, 2HELPS2B: 97%) and MCS (NN: 97%, EN: 99%, 2HELPS2B: 97%). 2HELPS2B model was able to identify the largest proportion of low‐risk patients.

Interpretation

For seizure risk stratification of hospitalized patients, the RiskSLIM generated 2HELPS2B model compares favorably to the complex NN and EN generated models. 2HELPS2B is able to accurately and quickly identify low‐risk patients with only a 1‐h screening EEG.

Resident training and interrater agreements using the ACNS critical care EEG terminology

PURPOSE

Electroencephalography (EEG) remains the gold standard for identifying rhythmic and periodic patterns in critically ill patients. Residents have frequent exposures to EEG and critically ill patients during their training. Our study aimed to assess resident competency in the use of the American Clinical Neurophysiology Society (ACNS) critical care EEG terminology.

METHODS

After self-guided reading and a 2-hour session reviewing the ACNS critical care EEG Terminology training slides, 16 adult neurology residents (PGY 2-4) completed the ACNS certification test. Performance scores were reported as average percent agreement (PA%) with a previously established 5-member expert panel. Interrater agreement was calculated to gauge consensus among peers within the resident cohort. Self-reported comfort levels using the terminology were also obtained.

RESULTS

The overall pass rate for our cohort was 50% and the median score was 74%. The terms with the highest PA% were: seizures (86.4%), main term 1 (78%), main term 2 (74%). Interrater agreement scores (kappa values) were almost perfect for seizure, and substantial for main terms 1 and 2.

CONCLUSIONS

Our data suggests that with minimal investment, adult neurology residents at various stages of training can effectively learn the ACNS critical care EEG Terminology.

Lateralized periodic discharges frequency correlates with glucose metabolism

OBJECTIVE

To investigate the correlation between characteristics of lateralized periodic discharges (LPDs) and glucose metabolism measured by ¹⁸F-fluorodeoxyglucose (FDG)-PET.

METHODS

We retrospectively reviewed medical records to identify patients who underwent FDG-PET during EEG monitoring with LPDs present during the FDG uptake period. Two blinded board-certified neurophysiologists independently interpreted EEGs. FDG uptake was measured using standardized uptake value (SUV). Structural images were fused with PET images to aid with localization of SUV. Two PET readers independently measured maximum SUV. Relative SUV values were obtained by normalization of the maximum SUV to the SUV of pons (SUVRpons). LPD frequency was analyzed both as a categorical variable and as a continuous measure. Other secondary variables included duration, amplitude, presence of structural lesion, and "plus" EEG features such as rhythmic or fast sharp activity.

RESULTS

Nine patients were identified and 7 had a structural etiology for LPDs. Analysis using frequency as a categorical variable and continuous variable showed an association between increased LPD frequency and increased ipsilateral SUVRpons (p = 0.02). Metabolism associated with LPDs (0.5 Hz as a baseline) increased by a median of 100% at 1 Hz and for frequencies >1 Hz increased by a median of 309%. There were no statistically significant differences in SUVRpons for other factors including duration (p = 0.10), amplitude (p = 0.80), structural etiology (p = 0.55), or "plus" features such as rhythmic or fast sharp activity (p = 0.84).

CONCLUSIONS

Metabolic activity increases monotonically with LPD frequency. LPD frequency should be a measure of interest when developing neuroprotection strategies in critical neurologic illness.

Predicting drug-resistant epilepsy - A machine learning approach based on administrative claims data

Patients with drug-resistant epilepsy (DRE) are at high risk of morbidity and mortality, yet their referral to specialist care is frequently delayed. The ability to identify patients at high risk of DRE at the time of treatment initiation, and to subsequently steer their treatment pathway toward more personalized interventions, has high clinical utility. Here, we aim to demonstrate the feasibility of developing algorithms for predicting DRE using machine learning methods. Longitudinal, intersected data sourced from US pharmacy, medical, and adjudicated hospital claims from 1,376,756 patients from 2006 to 2015 were analyzed; 292,892 met inclusion criteria for epilepsy, and 38,382 were classified as having DRE using a proxy measure for drug resistance. Patients were characterized using 1270 features reflecting demographics, comorbidities, medications, procedures, epilepsy status, and payer status. Data from 175,735 randomly selected patients were used to train three algorithms and from the remainder to assess the trained models' predictive power. A model with only age and sex was used as a benchmark. The best model, random forest, achieved an area under the receiver operating characteristic curve (95% confidence interval [CI]) of 0.764 (0.759, 0.770), compared with 0.657 (0.651, 0.663) for the benchmark model. Moreover, predicted probabilities for DRE were well-calibrated with the observed frequencies in the data. The model predicted drug resistance approximately 2 years before patients in the test dataset had failed two antiepileptic drugs (AEDs). Machine learning models constructed using claims data predicted which patients are likely to fail ≥3 AEDs and are at risk of developing DRE at the time of the first AED prescription. The use of such models can ensure that patients with predicted DRE receive specialist care with potentially more aggressive therapeutic interventions from diagnosis, to help reduce the serious sequelae of DRE.

Interictal Epileptiform Discharge Detection in EEG in Different Practice Settings

OBJECTIVE

The goal of the study was to measure the performance of academic and private practice (PP) neurologists in detecting interictal epileptiform discharges in routine scalp EEG recordings.

METHODS

Thirty-five EEG scorers (EEGers) participated (19 academic and 16 PP) and marked the location of ETs in 200 30-second EEG segments using a web-based EEG annotation system. All participants provided board certification status, years of Epilepsy Fellowship Training (EFT), and years in practice. The Persyst P13 automated IED detection algorithm was also run on the EEG segments for comparison.

RESULTS

Academic EEGers had an average of 1.66 years of EFT versus 0.50 years of EFT for PP EEGers (P < 0.0001) and had higher rates of board certification. Inter-rater agreement for the 35 EEGers was fair. There was higher performance for EEGers in academics, with at least 1.5 years of EFT, and with American Board of Clinical Neurophysiology and American Board of Psychiatry and Neurology-E specialty board certification. The Persyst P13 algorithm at its default setting (perception value = 0.4) did not perform as well at the EEGers, but at substantially higher perception value settings, the algorithm performed almost as well human experts.

CONCLUSIONS

Inter-rater agreement among EEGers in both academic and PP settings varies considerably. Practice location, years of EFT, and board certification are associated with significantly higher performance for IED detection in routine scalp EEG. Continued medical education of PP neurologists and neurologists without EFT is needed to improve routine scalp EEG interpretation skills. The performance of automated detection algorithms is approaching that of human experts.

Detecting abnormal electroencephalograms using deep convolutional networks

OBJECTIVES

Electroencephalography (EEG) is a central part of the medical evaluation for patients with neurological disorders. Training an algorithm to label the EEG normal vs abnormal seems challenging, because of EEG heterogeneity and dependence of contextual factors, including age and sleep stage. Our objectives were to validate prior work on an independent data set suggesting that deep learning methods can discriminate between normal vs abnormal EEGs, to understand whether age and sleep stage information can improve discrimination, and to understand what factors lead to errors.

METHODS

We train a deep convolutional neural network on a heterogeneous set of 8522 routine EEGs from the Massachusetts General Hospital. We explore several strategies for optimizing model performance, including accounting for age and sleep stage.

RESULTS

The area under the receiver operating characteristic curve (AUC) on an independent test set (n = 851) is 0.917 marginally improved by including age (AUC = 0.924), and both age and sleep stages (AUC = 0.925), though not statistically significant.

CONCLUSIONS

The model architecture generalizes well to an independent dataset. Adding age and sleep stage to the model does not significantly improve performance.

SIGNIFICANCE

Insights learned from misclassified examples, and minimal improvement by adding sleep stage and age suggest fruitful directions for further research.

Variability in pharmacologically-induced coma for treatment of refractory status epilepticus

Objective

To characterize the amount of EEG suppression achieved in refractory status epilepticus (RSE) patients treated with pharmacologically-induced coma (PIC).

Methods

We analyzed EEG recordings from 35 RSE patients between 21–84 years-old who received PIC that target burst suppression and quantified the amount of EEG suppression using the burst suppression probability (BSP). Then we measured the variability of BSPs with respect to a reference level of BSP 0.8 ± 0.15. Finally, we also measured the variability of BSPs with respect to the amount of intravenous anesthetic drugs (IVADs) received by the patients.

Results

Patients remained in the reference BSP range for only 8% (median, interquartile range IQR [0, 29] %) of the total time under treatment. The median time with BSP below the reference range was 84% (IQR [37, 100] %). BSPs in some patients drifted significantly over time despite constant infusion rates of IVADs. Similar weight-normalized infusion rates of IVADs in different patients nearly always resulted in distinct BSPs (probability 0.93 (IQR [0.82, 1.0]).

Conclusion

This study quantitatively identified high variability in the amount of EEG suppression achieved in clinical practice when treating RSE patients. While some of this variability may arise from clinicians purposefully deviating from clinical practice guidelines, our results show that the high variability also arises in part from significant inter- and intra- individual pharmacokinetic/pharmacodynamic variation. Our results indicate that the delicate balance between maintaining sufficient EEG suppression in RSE patients and minimizing IVAD exposure in clinical practice is challenging to achieve. This may affect patient outcomes and confound studies seeking to determine an optimal amount of EEG suppression for treatment of RSE. Therefore, our analysis points to the need for developing an alternative paradigm, such as vigilant anesthetic management as happens in operating rooms, or closed-loop anesthesia delivery, for investigating and providing induced-coma therapy to RSE patients.

Brain age from the electroencephalogram of sleep

The human electroencephalogram (EEG) of sleep undergoes profound changes with age. These changes can be conceptualized as "brain age (BA)," which can be compared to chronological age to reflect the degree of deviation from normal aging. Here, we develop an interpretable machine learning model to predict BA based on 2 large sleep EEG data sets: the Massachusetts General Hospital (MGH) sleep lab data set (N = 2532; ages 18-80); and the Sleep Heart Health Study (SHHS, N = 1974; ages 40-80). The model obtains a mean absolute deviation of 7.6 years between BA and chronological age (CA) in healthy participants in the MGH data set. As validation, a subset of SHHS containing longitudinal EEGs 5.2 years apart shows an average of 5.4 years increase in BA. Participants with significant neurological or psychiatric disease exhibit a mean excess BA, or "brain age index" (BAI = BA-CA) of 4 years relative to healthy controls. Participants with hypertension and diabetes have a mean excess BA of 3.5 years. The findings raise the prospect of using the sleep EEG as a potential biomarker for healthy brain aging.

Antiepileptic drug treatment after an unprovoked first seizure: A decision analysis

OBJECTIVE

To compare the expected quality-adjusted life-years (QALYs) in adult patients undergoing immediate vs deferred antiepileptic drug (AED) treatment after a first unprovoked seizure.

METHODS

We constructed a simulated clinical trial (Markov decision model) to compare immediate vs deferred AED treatment after a first unprovoked seizure in adults. Three base cases were considered, representing patients with varying degrees of seizure recurrence risk and effect of seizures on quality of life (QOL). Cohort simulation was performed to determine which treatment strategy would maximize the patient's expected QALYs. Sensitivity analyses were guided by clinical data to define decision thresholds across plausible measurement ranges, including seizure recurrence rate, effect of seizure recurrence on QOL, and efficacy of AEDs.

RESULTS

For patients with a moderate risk of recurrent seizures (52.0% over 10 years after first seizure), immediate AED treatment maximized QALYs compared to deferred treatment. Sensitivity analyses showed that for the preferred choice to change to deferred AED treatment, key clinical measures needed to reach implausible values were 10-year seizure recurrence rate ≤38.0%, QOL reduction with recurrent seizures ≤0.06, and efficacy of AEDs on lowering seizure recurrence rate ≤16.3%.

CONCLUSION

Our model determined that immediate AED treatment is preferable to deferred treatment in adult first-seizure patients over a wide and clinically relevant range of variables. Furthermore, our analysis suggests that the 10-year seizure recurrence rate that justifies AED treatment (38.0%) is substantially lower than the 60% threshold used in the current definition of epilepsy.

CLASSIFIER CASCADE TO AID IN DETECTION OF EPILEPTIFORM TRANSIENTS IN INTERICTAL EEG

The presence of Epileptiform Transients (ET) in the electroencephalogram (EEG) is a key finding in the medical workup of a patient with suspected epilepsy. Automated ET detection can increase the uniformity and speed of ET detection. Current ET detection methods suffer from insufficient precision and high false positive rates. Since ETs occur infrequently in the EEG of most patients, the majority of recordings comprise background EEG waveforms. In this work we establish a method to exclude as much background data as possible from EEG recordings by applying a classifier cascade. The remaining data can then be classified using other ET detection methods. We compare a single Support Vector Machine (SVM) to a cascade of SVMs for detecting ETs. Our results show that the precision and false positive rate improve significantly by incorporating a classifier cascade before ET detection. Our method can help improve the precision and false positive rate of an ET detection system. At a fixed sensitivity, we were able to improve precision by 6.78%; and at a fixed false positive rate, the sensitivity improved by 2.83%.

Effect of epileptiform abnormality burden on neurologic outcome and antiepileptic drug management after subarachnoid hemorrhage

OBJECTIVE

To quantify the burden of epileptiform abnormalities (EAs) including seizures, periodic and rhythmic activity, and sporadic discharges in patients with aneurysmal subarachnoid hemorrhage (aSAH), and assess the effect of EA burden and treatment on outcomes.

METHODS

Retrospective analysis of 136 high-grade aSAH patients. EAs were defined using the American Clinical Neurophysiology Society nomenclature. Burden was defined as prevalence of <1%, 1-9%, 10-49%, 50-89%, and >90% for each 18-24 hour epoch. Our outcome measure was 3-month Glasgow Outcome Score.

RESULTS

47.8% patients had EAs. After adjusting for clinical covariates EA burden on first day of recording and maximum daily burden were associated with worse outcomes. Patients with higher EA burden were more likely to be treated with anti-epileptic drugs (AEDs) beyond the standard prophylactic protocol. There was no difference in outcomes between patients continued on AEDs beyond standard prophylaxis compared to those who were not.

CONCLUSIONS

Higher burden of EAs in aSAH independently predicts worse outcome. Although nearly half of these patients received treatment, our data suggest current AED management practices may not influence outcome.

SIGNIFICANCE

EA burden predicts worse outcomes and may serve as a target for prospective interventional controlled studies to directly assess the impact of AEDs, and create evidence-based treatment protocols.

You Snooze, You Win: the PhysioNet/Computing in Cardiology Challenge 2018

The PhysioNet/Computing in Cardiology Challenge 2018 focused on the use of various physiological signals (EEG, EOG, EMG, ECG, SaO₂) collected during polysomnographic sleep studies to detect sources of arousal (non-apnea) during sleep. A total of 1,983 polysomnographic recordings were made available to the entrants. The arousal labels for 994 of the recordings were made available in a public training set while 989 labels were retained in a hidden test set. Challengers were asked to develop an algorithm that could label the presence of arousals within the hidden test set. The performance metric used to assess entrants was the area under the precision-recall curve. A total of twenty-two independent teams entered the Challenge, deploying a variety of methods from generalized linear models to deep neural networks.

Bilateral independent periodic discharges are associated with electrographic seizures and poor outcome: A case-control study

OBJECTIVE

To determine the clinical correlates bilateral independent periodic discharges (BIPDs) and their association with electrographic seizures and outcome.

METHODS

Retrospective case-control study of patients with BIPDs compared to patients without periodic discharges ("No PDs") and patients with lateralized periodic discharges ("LPDs"), matched for age, etiology and level of alertness.

RESULTS

We included 85 cases and 85 controls in each group. The most frequent etiologies of BIPDs were stroke, CNS infections, and anoxic brain injury. Acute bilateral cerebral injury was more common in the BIPDs group than in the No PDs and LPDs groups (70% vs. 37% vs. 35%). Electrographic seizures were more common with BIPDs than in the absence of PDs (45% vs. 8%), but not than with LPDs (52%). Mortality was higher in the BIPDs group (36%) than in the No PDs group (18%), with fewer patients with BIPDs achieving good outcome (moderate disability or better; 18% vs. 36%), but not than in the LPDs group (24% mortality, 26% good outcome). In multivariate analyses, BIPDs remained associated with mortality (OR: 3.0 [1.4-6.4]) and poor outcome (OR: 2.9 [1.4-6.2]).

CONCLUSION

BIPDs are caused by bilateral acute brain injury and are associated with a high risk of electrographic seizures and of poor outcome.

SIGNIFICANCE

BIPDs are uncommon but their identification in critically ill patients has potential important implications, both in terms of clinical management and prognostication.

Design, implementation, and evaluation of a physiological closed-loop control device for medically-induced coma

Concerns regarding reliability and safety, as well as uncertainties about what constitutes adequate performance evaluation, have impeded the clinical translation of PCLC devices. We describe an attempt to address these challenges through design, implementation, and evaluation of a PCLC device for delivering medically-induced coma, with the intention to eventually conduct a clinical trial. This device works by automatically adjusting the infusion rate of propofol - a general anesthetic - in response to an electroencephalogram (EEG) pattern called burst suppression. We also designed and implemented a computational patient model which interfaces with hardware and produces realistic EEG signals in response to propofol infusion. The computational patient model is used in hardware-in-the-loop studies to evaluate the behavior of our PCLC device under realistic perturbations. Finally, we have tested the performance of our PCLC device in rodents. Results from these studies suggest that closed-loop control of medically-induced coma in humans is feasible and robust. Consequently, our work produced a PCLC device and relevant pre-clinical evidence in support of a pilot clinical trial.

Continuous electroencephalography predicts delayed cerebral ischemia after subarachnoid hemorrhage: A prospective study of diagnostic accuracy

OBJECTIVE

Delayed cerebral ischemia (DCI) is a common, disabling complication of subarachnoid hemorrhage (SAH). Preventing DCI is a key focus of neurocritical care, but interventions carry risk and cannot be applied indiscriminately. Although retrospective studies have identified continuous electroencephalographic (cEEG) measures associated with DCI, no study has characterized the accuracy of cEEG with sufficient rigor to justify using it to triage patients to interventions or clinical trials. We therefore prospectively assessed the accuracy of cEEG for predicting DCI, following the Standards for Reporting Diagnostic Accuracy Studies.

METHODS

We prospectively performed cEEG in nontraumatic, high-grade SAH patients at a single institution. The index test consisted of clinical neurophysiologists prospectively reporting prespecified EEG alarms: (1) decreasing relative alpha variability, (2) decreasing alpha-delta ratio, (3) worsening focal slowing, or (4) late appearing epileptiform abnormalities. The diagnostic reference standard was DCI determined by blinded, adjudicated review. Primary outcome measures were sensitivity and specificity of cEEG for subsequent DCI, determined by multistate survival analysis, adjusted for baseline risk.

RESULTS

One hundred three of 227 consecutive patients were eligible and underwent cEEG monitoring (7.7-day mean duration). EEG alarms occurred in 96.2% of patients with and 19.6% without subsequent DCI (1.9-day median latency, interquartile range = 0.9-4.1). Among alarm subtypes, late onset epileptiform abnormalities had the highest predictive value. Prespecified EEG findings predicted DCI among patients with low (91% sensitivity, 83% specificity) and high (95% sensitivity, 77% specificity) baseline risk.

INTERPRETATION

cEEG accurately predicts DCI following SAH and may help target therapies to patients at highest risk of secondary brain injury. Ann Neurol 2018;83:958-969.

Minimizing ICU Neurological Dysfunction with Dexmedetomidine-induced Sleep (MINDDS): protocol for a randomised, double-blind, parallel-arm, placebo-controlled trial

INTRODUCTION

Delirium, which is prevalent in postcardiac surgical patients, is an acute brain dysfunction characterised by disturbances in attention, awareness and cognition not explained by a pre-existing neurocognitive disorder. The pathophysiology of delirium remains poorly understood. However, basic science and clinical studies suggest that sleep disturbance may be a modifiable risk factor for the development of delirium. Dexmedetomidine is a α-2A adrenergic receptor agonist medication that patterns the activity of various arousal nuclei similar to sleep. A single night-time loading dose of dexmedetomidine promotes non-rapid eye movement sleep stages N2 and N3 sleep. This trial hypothesises dexmedetomidine-induced sleep as pre-emptive therapy for postoperative delirium.

METHODS AND ANALYSIS

The MINDDS (Minimizing ICU Neurological Dysfunction with Dexmedetomidine-induced Sleep) trial is a 370-patient block-randomised, placebo-controlled, double-blinded, single-site, parallel-arm superiority trial. Patients over 60 years old, undergoing cardiac surgery with planned cardiopulmonary bypass, will be randomised to receive a sleep-inducing dose of dexmedetomidine or placebo. The primary outcome is the incidence of delirium on postoperative day 1, assessed with the Confusion Assessment Method by staff blinded to the treatment assignment. To ensure that the study is appropriately powered for the primary outcome measure, patients will be recruited and randomised into the study until 370 patients receive the study intervention on postoperative day 0. Secondary outcomes will be evaluated by in-person assessments and medical record review for in-hospital end points, and by telephone interview for 30-day, 90-day and 180-day end points. All trial outcomes will be evaluated using an intention-to-treat analysis plan. Hypothesis testing will be performed using a two-sided significance level (type I error) of α=0.05. Sensitivity analyses using the actual treatment received will be performed and compared with the intention-to-treat analysis results. Additional sensitivity analyses will assess the potential impact of missing data due to loss of follow-up.

ETHICS AND DISSEMINATION

The Partners Human Research Committee approved the MINDDS trial. Recruitment began in March 2017. Dissemination plans include presentations at scientific conferences, scientific publications and popular media.

TRIAL REGISTRATION NUMBER

NCT02856594.

Epileptiform activity in traumatic brain injury predicts post-traumatic epilepsy

We hypothesize that epileptiform abnormalities (EAs) in the electroencephalogram (EEG) during the acute period following traumatic brain injury (TBI) independently predict first-year post-traumatic epilepsy (PTE₁ ). We analyze PTE₁ risk factors in two cohorts matched for TBI severity and age (n = 50). EAs independently predict risk for PTE₁ (odds ratio [OR], 3.16 [0.99, 11.68]); subdural hematoma is another independent risk factor (OR, 4.13 [1.18, 39.33]). Differences in EA rates are apparent within 5 days following TBI. Our results suggest that increased EA prevalence identifies patients at increased risk for PTE₁ , and that EAs acutely post-TBI can identify patients most likely to benefit from antiepileptogenesis drug trials. Ann Neurol 2018;83:858-862.

An enhanced cerebral recovery index for coma prognostication following cardiac arrest

Prognostication of coma outcomes following cardiac arrest is both qualitative and poorly understood in current practice. Existing quantitative metrics are powerful, but lack rigorous approaches to classification. This is due, in part, to a lack of available data on the population of interest. In this paper we describe a novel retrospective data set of 167 cardiac arrest patients (spanning three institutions) who received electroencephalography (EEG) monitoring. We utilized a subset of the collected data to generate features that measured the connectivity, complexity and category of EEG activity. A subset of these features was included in a logistic regression model to estimate a dichotomized cerebral performance category score at discharge. We compared the predictive performance of our method against an established EEG-based alternative, the Cerebral Recovery Index (CRI) and show that our approach more reliably classifies patient outcomes, with an average increase in AUC of 0.27.

Electroencephalogram Based Detection of Deep Sedation in ICU Patients Using Atomic Decomposition

OBJECTIVE

This study was performed to evaluate how well states of deep sedation in ICU patients can be detected from the frontal electroencephalogram (EEG) using features based on the method of atomic decomposition (AD).

METHODS

We analyzed a clinical dataset of 20 min of EEG recordings per patient from 44 mechanically ventilated adult patients receiving sedatives in an intensive care unit (ICU) setting. Several features derived from AD of the EEG signal were used to discriminate between awake and sedated states. We trained support vector machine (SVM) classifiers using AD features and compared the classification performance with SVM classifiers trained using standard spectral and entropy features using leave-one-subject-out validation. The potential of each feature to discriminate between awake and sedated states was quantified using area under the receiver operating characteristic curve (AUC).

RESULTS

The sedation level classification system using AD was able to reliably discriminate between sedated and awake states achieving an average AUC of 0.90, which was significantly better () than performance achieved using spectral (AUC = 0.86) and entropy (AUC = 0.81) domain features. A combined feature set consisting of AD, entropy, and spectral features provided better discrimination (AUC = 0.91, ) than any individual feature set.

CONCLUSIONS

Features derived from the atomic decomposition of EEG signals provide useful discriminative information about the depth of sedation in ICU patients.

SIGNIFICANCE

With further refinement and external validation, the proposed system may be able to assist clinical staff with continuous surveillance of sedation levels in mechanically ventilated critically ill ICU patients.

Association of an Electroencephalography-Based Risk Score With Seizure Probability in Hospitalized Patients

Importance

Continuous electroencephalography (EEG) use in critically ill patients is expanding. There is no validated method to combine risk factors and guide clinicians in assessing seizure risk.

Objective

To use seizure risk factors from EEG and clinical history to create a simple scoring system associated with the probability of seizures in patients with acute illness.

Design, Setting, and Participants

We used a prospective multicenter (Emory University Hospital, Brigham and Women's Hospital, and Yale University Hospital) database containing clinical and electrographic variables on 5427 continuous EEG sessions from eligible patients if they had continuous EEG for clinical indications, excluding epilepsy monitoring unit admissions. We created a scoring system model to estimate seizure risk in acutely ill patients undergoing continuous EEG. The model was built using a new machine learning method (RiskSLIM) that is designed to produce accurate, risk-calibrated scoring systems with a limited number of variables and small integer weights. We validated the accuracy and risk calibration of our model using cross-validation and compared its performance with models built with state-of-the-art logistic regression methods. The database was developed by the Critical Care EEG Research Consortium and used data collected over 3 years. The EEG variables were interpreted using standardized terminology by certified reviewers.

Exposures

All patients had more than 6 hours of uninterrupted EEG recordings.

Main Outcomes and Measures

The main outcome was the average risk calibration error.

Results

There were 5427 continuous EEGs performed on 4772 participants (2868 men, 49.9%; median age, 61 years) performed at 3 institutions, without further demographic stratification. Our final model, 2HELPS2B, had an area under the curve of 0.819 and average calibration error of 2.7% (95% CI, 2.0%-3.6%). It included 6 variables with the following point assignments: (1) brief (ictal) rhythmic discharges (B[I]RDs) (2 points); (2) presence of lateralized periodic discharges, lateralized rhythmic delta activity, or bilateral independent periodic discharges (1 point); (3) prior seizure (1 point); (4) sporadic epileptiform discharges (1 point); (5) frequency greater than 2.0 Hz for any periodic or rhythmic pattern (1 point); and (6) presence of "plus" features (superimposed, rhythmic, sharp, or fast activity) (1 point). The probable seizure risk of each score was 5% for a score of 0, 12% for a score of 1, 27% for a score of 2, 50% for a score of 3, 73% for a score of 4, 88% for a score of 5, and greater than 95% for a score of 6 or 7.

Conclusions and Relevance

The 2HELPS2B model is a quick accurate tool to aid clinical judgment of the risk of seizures in critically ill patients.