Ellen R. Grass Lecture: Rapid EEG Analysis for Intensive Care Decisions in Status Epilepticus

Non-convulsive status epilepticus in the immediate postoperative period following spine surgery

Background

Non-convulsive status epilepticus (NCSE), in which continuous epileptiform discharges occur without seizure-like movement, is rare and unfamiliar to anesthesiologists, both of which make this condition overlooked in patients with decreased levels of consciousness following general anesthesia.

Case

We report on an elderly female patient who developed NCSE in the immediate postoperative period after the spine surgery. Initially, delayed emergence from anesthesia was suspected, but the electroencephalogram confirmed NCSE, and anticonvulsant therapy was initiated.

Conclusions

Delayed emergence is commonly attributed to cerebrovascular events or residual anesthetic effects, but NCSE must be included in the differential diagnosis, especially in elderly patients. Anticonvulsant therapy should be initiated as soon as possible for a better prognosis.

Automated spectrographic seizure detection using convolutional neural networks

PURPOSE

Non-convulsive seizures are common in critically ill patients, and delays in diagnosis contribute to increased morbidity and mortality. Many intensive care units employ continuous EEG (cEEG) for seizure monitoring. Although cEEG is continuously recorded, it is often reviewed intermittently, which may delay seizure diagnosis and treatment. This may be mitigated with automated seizure detection. In this study, we develop and evaluate convolutional neural networks (CNN) to automate seizure detection on EEG spectrograms.

METHODS

Adult EEGs (12 patients, 12 EEGs, 33 seizures) from New-York Presbyterian Hospital (NYP) and pediatric EEGs (22 patients, 130 EEGs, 177 seizures) from Children's Hospital Boston (CHB) were converted into spectrograms. To simulate a telemetry display, seizure and non-seizure events on spectrograms were sequentially sampled as images across a detection window (26,380 total images). Four CNN models of increasing complexity (number of layers) were trained, cross-validated, and tested on CHB and NYP spectrographic images. All CNNs were based on the VGG-net architecture, with adjustments to alleviate overfitting.

RESULTS

For spectrographically visible seizures, two CNN models (containing 4 and 7 convolution layers) achieved >90% seizure detection sensitivity and specificity on the CHB test set and >90% sensitivity and 75-80% specificity on the NYP test set. The one CNN model (10 convolution layers) did not converge during training; while another CNN (2 convolution layers) performed poorly (60% sensitivity and 32% specificity) on the NYP test set.

CONCLUSIONS

Seizure detection on EEG spectrograms with CNN models is feasible with sensitivity and specificity potentially suitable for clinical use.

Can a collaborative healthcare network improve the care of people with epilepsy?

New opportunities are now available to improve care in ways not possible previously. Information contained in electronic medical records can now be shared without identifying patients. With network collaboration, large numbers of medical records can be searched to identify patients most like the one whose complex medical situation challenges the physician. The clinical effectiveness of different treatment strategies can be assessed rapidly to help the clinician decide on the best treatment for this patient. Other capabilities from different components of the network can prompt the recognition of what is the best available option and encourage the sharing of information about programs and electronic tools. Difficulties related to privacy, harmonization, integration, and costs are expected, but these are currently being addressed successfully by groups of organizations led by those who recognize the benefits.

StatNet Electroencephalogram: A Fast and Reliable Option to Diagnose Nonconvulsive Status Epilepticus in Emergency Setting

BACKGROUND

The StatNet electrode set is a system that can be applied by a non-electroencephalogram (EEG) technologist after minimal training. The primary objectives of this study are to assess the quality and reliability of the StatNet recordings in comparison to the conventional EEG.

METHODS

Over 10 months, 19 patients with suspected nonconvulsive status epilepticus were included from university hospital emergency settings. Each patient received a StatNet EEG by a trained epilepsy fellow and a conventional EEG by registered technologists. We compared the studies in a blinded fashion, for the timeframe from EEG order to the setup time, start of acquisition, amount of artifact, and detection of abnormalities. The nonparametric Mann-Whitney two-sample t test was used for comparisons. The kappa score was used to assess reliability.

RESULTS

Mean age of patients was 61±16.3 (25-93) years. The inter-observer agreement for detection of abnormal findings was 0.83 for StatNet and 0.75 for conventional EEG. Nonconvulsive status epilepticus was detected in 10% (2/19) in both studies. The delay from the time of EEG requisition to acquisition was shorter in the StatNet (22.4±2.5 minutes) than the conventional EEG (217.7±44.6 minutes; p<0.0001). The setup time was also shorter in the StatNet (9.9±0.8 minutes) compared with the conventional EEG (17.8±0.8 minutes; p<0.0001). There was no difference in the percentage of artifact duration between the two studies (p=0.89).

CONCLUSION

This study demonstrates that StatNet EEG is a practical and reliable tool in the emergency setting, which reduces the delay of testing compared with conventional EEG, without significant compromise of study quality.

Continuous EEG monitoring in adults in the intensive care unit (ICU)

Continuous EEG monitoring in the ICU is different from planned EEG due to the rather urgent nature of the indications, explaining the fact that recording is started in certain cases by the clinical team in charge of the patient's care. Close collaboration between neurophysiology teams and intensive care teams is essential. Continuous EEG monitoring can be facilitated by quantified analysis systems. This kind of analysis is based on certain signal characteristics, such as amplitude or frequency content, but raw EEG data should always be interpreted if possible, since artefacts can sometimes impair quantified EEG analysis. It is preferable to work within a tele-EEG network, so that the neurophysiologist has the possibility to give an interpretation on call. Continuous EEG monitoring is thus useful in the diagnosis of non-convulsive epileptic seizures or purely electrical discharges and in the monitoring of status epilepticus when consciousness disorders persist after initial treatment. A number of other indications are currently under evaluation.

[French guidelines on electroencephalogram]

Electroencephalography allows the functional analysis of electrical brain cortical activity and is the gold standard for analyzing electrophysiological processes involved in epilepsy but also in several other dysfunctions of the central nervous system. Morphological imaging yields complementary data, yet it cannot replace the essential functional analysis tool that is EEG. Furthermore, EEG has the great advantage of being non-invasive, easy to perform and allows control tests when follow-up is necessary, even at the patient's bedside. Faced with the advances in knowledge, techniques and indications, the Société de Neurophysiologie Clinique de Langue Française (SNCLF) and the Ligue Française Contre l'Épilepsie (LFCE) found it necessary to provide an update on EEG recommendations. This article will review the methodology applied to this work, refine the various topics detailed in the following chapters. It will go over the summary of recommendations for each of these chapters and underline proposals for writing an EEG report. Some questions could not be answered by the review of the literature; in those cases, an expert advice was given by the working and reading groups in addition to the guidelines.

Electroencephalographic monitoring in the pediatric intensive care unit

Continuous electroencephalographic (CEEG) monitoring is used with increasing frequency in critically ill children to provide insight into brain function and to identify electrographic seizures. CEEG monitoring use often impacts clinical management, most often by identifying electrographic seizures and status epilepticus. Most electrographic seizures have no clinical correlate, and thus would not be identified without CEEG monitoring. There are increasing data showing that electrographic seizures and electrographic status epilepticus are associated with worse outcome. Seizure identification efficiency may be improved by further development of quantitative electroencephalography trends. This review describes the clinical impact of CEEG data, the epidemiology of electrographic seizures and status epilepticus, the impact of electrographic seizures on outcome, the utility of quantitative electroencephalographic trends for seizure identification, and practical considerations regarding CEEG monitoring.

Seizure detection using digital trend analysis: Factors affecting utility

BACKGROUND

EEG monitoring is important for the early detection of seizures during the course of critical illness. However, the logistics of real time EEG interpretation is challenging for the neurophysiology and critical care medicine teams. This study evaluated factors affecting the utility of digital trend analysis (DTA) for rapid seizure identification in children.

METHODS

digital EEG files of seizures in critically ill children were retrieved for DTA. The envelop trend (ET) and compressed spectral array (CSA) were applied to the raw EEG data and presented to an experienced and inexperienced user for interpretation who were blinded to conventional EEG findings. The EEG findings with and without presence of seizures and features of seizures were analyzed.

RESULTS

we found that a number of factors affected accurate seizure detection including factors related to interpreter's experiences, display size and type of DTA methods used for analysis in addition to baseline EEG findings. ET was more dependent on user experience, furthermore, display size and multimodal DTA application (CSA and ET combined) increased the sensitivity of seizure detection for the experienced user compared to inexperience users. The artifacts were reported as seizures regardless of experience without presence of conventional EEG recording. The maximum spike amplitude, seizure duration, and seizure frequency were other important determinants for accuracy. Electrographic seizures with shorter duration were better detected by ET, and the maximum spike amplitude was important for both the ET and CSA. Repetitive seizures are readily detected by both digital trending methods. Artifacts may be reported as seizures regardless of experience if conventional EEG recording is not available for the interpretation.

CONCLUSION

DTA applied to the raw EEG data does produce a graphic display that facilitates identification of seizures. The actual characteristics of the electrographic seizure may predict which DTA method is better and the overall accuracy of seizure detection may increase when multimodal trending is used simultaneously. Application of DTA alone with display of conventional EEG is beneficial for rapid interpretation of EEG findings regardless of experience.

Anticonvulsant medications in the pediatric emergency room and intensive care unit

Seizures are common in pediatric emergency care units, either as the main medical issue or in association with an additional neurological problem. Rapid treatment prolonged and repetitive seizures or status epilepticus is important. Multiple anti-convulsant medications are useful in this setting, and each has various indications and potential adverse effects that must be considered in regard to individual patients. This review discusses new data regarding anticonvulsants that are useful in these settings, including fosphenytoin, valproic acid, levetiracetam, and topiramate. A status epilepticus treatment algorithm is suggested, incorporating changes from traditional algorithms based on these new data. Treatment issues specific to complex medical patients, including patients with brain tumors, renal dysfunction, hepatic dysfunction, transplant, congenital heart disease, and anticoagulation, are also discussed.