Neonatal Seizure Detection Research Articles

Exploring temporal information in neonatal seizures using a dynamic time warping based SVM kernel

Seizure events in newborns change in frequency, morphology, and propagation. This contextual information is explored at the classifier level in the proposed patient-independent neonatal seizure detection system. The system is based on the combination of a static and a sequential SVM classifier. A Gaussian dynamic time warping based kernel is used in the sequential classifier. The system is validated on a large dataset of EEG recordings from 17 neonates. The obtained results show an increase in the detection rate at very low false detections per hour, particularly achieving a 12% improvement in the detection of short seizure events over the static RBF kernel based system.

Toward a Personalized Real-Time Diagnosis in Neonatal Seizure Detection.

The problem of creating a personalized seizure detection algorithm for newborns is tackled in this paper. A probabilistic framework for semi-supervised adaptation of a generic patient-independent neonatal seizure detector is proposed. A system that is based on a combination of patient-adaptive (generative) and patient-independent (discriminative) classifiers is designed and evaluated on a large database of unedited continuous multichannel neonatal EEG recordings of over 800 h in duration. It is shown that an improvement in the detection of neonatal seizures over the course of long EEG recordings is achievable with on-the-fly incorporation of patient-specific EEG characteristics. In the clinical setting, the employment of the developed system will maintain a seizure detection rate at 70% while halving the number of false detections per hour, from 0.4 to 0.2 FD/h. This is the first study to propose the use of online adaptation without clinical labels, to build a personalized diagnostic system for the detection of neonatal seizures.

Detecting Neonatal Seizures With Computer Algorithms.

It is now generally accepted that EEG is the only reliable way to accurately detect newborn seizures and, as such, prolonged EEG monitoring is increasingly being adopted in neonatal intensive care units. Long EEG recordings may last from several hours to a few days. With neurophysiologists not always available to review the EEG during unsociable hours, there is a pressing need to develop a reliable and robust automatic seizure detection method-a computer algorithm that can take the EEG signal, process it, and output information that supports clinical decision making. In this study, we review existing algorithms based on how the relevant seizure information is exploited. We start with commonly used methods to extract signatures from seizure signals that range from those that mimic the clinical neurophysiologist to those that exploit mathematical models of neonatal EEG generation. Commonly used classification methods are reviewed that are based on a set of rules and thresholds that are either heuristically tuned or automatically derived from the data. These are followed by techniques to use information about spatiotemporal seizure context. The usual errors in system design and validation are discussed. Current clinical decision support tools that have met regulatory requirements and are available to detect neonatal seizures are reviewed with progress and the outstanding challenges are outlined. This review discusses the current state of the art regarding automatic detection of neonatal seizures.

The Impact of Continuous Digital Video Electroencephalogram (CEEG) Introduction on the Use of Anti-Epileptic Drugs (AEDS) in Neonates with Hypoxic Ischemic Encephalopathy (HIE)

Abstract BACKGROUND: Use of AEDs in neonates may disrupt neuronal development by synaptic re-organization and altering brain plasticity. Introduction of cEEG as part of neonatal neurocritical care long term monitoring (LTM) aims to improve seizure detection and control in neo-nates with HIE. OBJECTIVES: To assess the impact of cEEG on seizure detection and the use of AEDs in neonates with HIE DESIGN/METHODS: A retrospective cohort study of neonates admitted with HIE to a level III NICU two years before and after introduction of LTM in August 2013. The cEEGs were read remotely by a pediatric epileptologist. Before LTM, a 60 minute routine EEG (rEEG) was the standard practice. Primary outcome was the number of AEDs used. Secondary outcomes were AEDs used as maintenance and on discharge, and AEDs burden (total mg/kg used during hospital stay). HIE staging was based on modified Sarnat scoring system. Drug burden for each AED was calculated. Categorical variables were analysed using Chi-square test and continuous variables using student's t test. RESULTS: n=157 (total number of HIE patients); median gestation 40 weeks (IQR 38-40); 71 infants(45%) had LTM monitoring;103 infants(66%) were cooled. Baseline and clinical characteristics including disease severity and cooling were similar. In the pre LTM era, 12 of 52 (23%) infants with clinical seizures, only 12 had seizures by rEEG . Using cEEG, 17 out of 33 (52%) clinical seizures were confirmed by cEEG; 5/36(1%) had subclinical seizures compared to none before LTM. Detection of EEG seizures increased significantly with LTM (p 0.02). Discharge on any AEDs was less in the LTM group and the mean AEDs burden was reduced from 176 mg/kg in the pre LTM to 91 mg/kg in the post LTM group. CONCLUSION: Introduction of cEEG improved seizure detection and control; but decreased the overall use and burden of AEDs. p value <0.05 is denoted as *

Optimal training dataset composition for SVM-based, age-independent, automated epileptic seizure detection.

Automated seizure detection is a valuable asset to health professionals, which makes adequate treatment possible in order to minimize brain damage. Most research focuses on two separate aspects of automated seizure detection: EEG feature computation and classification methods. Little research has been published regarding optimal training dataset composition for patient-independent seizure detection. This paper evaluates the performance of classifiers trained on different datasets in order to determine the optimal dataset for use in classifier training for automated, age-independent, seizure detection. Three datasets are used to train a support vector machine (SVM) classifier: (1) EEG from neonatal patients, (2) EEG from adult patients and (3) EEG from both neonates and adults. To correct for baseline EEG feature differences among patients feature, normalization is essential. Usually dedicated detection systems are developed for either neonatal or adult patients. Normalization might allow for the development of a single seizure detection system for patients irrespective of their age. Two classifier versions are trained on all three datasets: one with feature normalization and one without. This gives us six different classifiers to evaluate using both the neonatal and adults test sets. As a performance measure, the area under the receiver operating characteristics curve (AUC) is used. With application of FBC, it resulted in performance values of 0.90 and 0.93 for neonatal and adult seizure detection, respectively. For neonatal seizure detection, the classifier trained on EEG from adult patients performed significantly worse compared to both the classifier trained on EEG data from neonatal patients and the classier trained on both neonatal and adult EEG data. For adult seizure detection, optimal performance was achieved by either the classifier trained on adult EEG data or the classifier trained on both neonatal and adult EEG data. Our results show that age-independent seizure detection is possible by training one classifier on EEG data from both neonatal and adult patients. Furthermore, our results indicate that for accurate age-independent seizure detection, it is important that EEG data from each age category are used for classifier training. This is particularly important for neonatal seizure detection. Our results underline the under-appreciated importance of training dataset composition with respect to accurate age-independent seizure detection.Electronic supplementary materialThe online version of this article (doi:10.1007/s11517-016-1468-y) contains supplementary material, which is available to authorized users.

In-depth performance analysis of an EEG based neonatal seizure detection algorithm.

To describe a novel neurophysiology based performance analysis of automated seizure detection algorithms for neonatal EEG to characterize features of detected and non-detected seizures and causes of false detections to identify areas for algorithmic improvement. EEGs of 20 term neonates were recorded (10 seizure, 10 non-seizure). Seizures were annotated by an expert and characterized using a novel set of 10 criteria. ANSeR seizure detection algorithm (SDA) seizure annotations were compared to the expert to derive detected and non-detected seizures at three SDA sensitivity thresholds. Differences in seizure characteristics between groups were compared using univariate and multivariate analysis. False detections were characterized. The expert detected 421 seizures. The SDA at thresholds 0.4, 0.5, 0.6 detected 60%, 54% and 45% of seizures. At all thresholds, multivariate analyses demonstrated that the odds of detecting seizure increased with 4 criteria: seizure amplitude, duration, rhythmicity and number of EEG channels involved at seizure peak. Major causes of false detections included respiration and sweat artefacts or a highly rhythmic background, often during intermediate sleep. This rigorous analysis allows estimation of how key seizure features are exploited by SDAs. This study resulted in a beta version of ANSeR with significantly improved performance.

Interobserver agreement for neonatal seizure detection using multichannel EEG.

To determine the interobserver agreement (IOA) of neonatal seizure detection using the gold standard of conventional, multichannel EEG. A cohort of full-term neonates at risk of acute encephalopathy was included in this prospective study. The EEG recordings of these neonates were independently reviewed for seizures by three international experts. The IOA was estimated using statistical measures including Fleiss' kappa and percentage agreement assessed over seizure events (event basis) and seizure duration (temporal basis). A total of 4066h of EEG recordings from 70 neonates were reviewed with an average of 2555 seizures detected. The IOA was high with temporal assessment resulting in a kappa of 0.827 (95% CI: 0.769-0.865; n=70). The median agreement was 83.0% (interquartile range [IQR]: 76.6-89.5%; n=33) for seizure and 99.7% (IQR: 98.9-99.8%; n=70) for nonseizure EEG. Analysis of events showed a median agreement of 83.0% (IQR: 72.9-86.6%; n=33) for seizures with 0.018 disagreements per hour (IQR: 0.000-0.090 per hour; n=70). Observers were more likely to disagree when a seizure was less than 30sec. Overall, 33 neonates were diagnosed with seizures and 28 neonates were not, by all three observers. Of the remaining nine neonates with contradictory EEG detections, seven presented with low total seizure burden. The IOA is high among experts for the detection of neonatal seizures using conventional, multichannel EEG. Agreement is reduced when seizures are rare or have short duration. These findings support EEG-based decision making in the neonatal intensive care unit, inform EEG interpretation guidelines, and provide benchmarks for seizure detection algorithms.

Improvement of an automated neonatal seizure detector using a post-processing technique.

Visual recognition of neonatal seizures during continuous EEG monitoring in neonatal intensive care units (NICUs) is labor-intensive, has low inter-rater agreement and requires special expertise that is not available around the clock. Development of an accurate automated seizure detection system with a low false alarm rate will support clinical decision making and alleviate significantly the workload. However, this is an ongoing difficult challenge for engineers as the neonatal EEG signal is non-stationary and often includes complex patterns of seizures and artifacts. In this study, we show an improvement of our previously developed neonatal seizure detector (developed using heuristic if-then rules). In order to improve the detection accuracy, mean phase coherence as a new feature is used to characterize artifacts and also support vector machine is applied to perform the post-processing step to remove false detections. As a result, the false alarm rate drops 42% (from 2.6 h(-1) to 1.5 h(-1)), whereas the good detection rate reduces only by 4%.

Using trend templates in a neonatal seizure algorithm improves detection of short seizures in a foetal ovine model

Seizures below one minute in duration are difficult to assess correctly using seizure detection algorithms. We aimed to improve neonatal detection algorithm performance for short seizures through the use of trend templates for seizure onset and end. Bipolar EEG were recorded within a transiently asphyxiated ovine model at 0.7 gestational age, a common experimental model for studying brain development in humans of 30–34 weeks of gestation. Transient asphyxia led to electrographic seizures within 6–8 h. A total of 3159 seizures, 2386 shorter than one minute, were annotated in 1976 h-long EEG recordings from 17 foetal lambs. To capture EEG characteristics, five features, sensitive to seizures, were calculated and used to derive trend information. Feature values and trend information were used as input for support vector machine classification and subsequently post-processed. Performance metrics, calculated after post-processing, were compared between analyses with and without employing trend information. Detector performance was assessed after five-fold cross-validation conducted ten times with random splits. The use of trend templates for seizure onset and end in a neonatal seizure detection algorithm significantly improves the correct detection of short seizures using two-channel EEG recordings from 54.3% (52.6–56.1) to 59.5% (58.5–59.9) at FDR 2.0 (median (range); p < 0.001, Wilcoxon signed rank test). Using trend templates might therefore aid in detection of short seizures by EEG monitoring at the NICU.

Clinical implementation of a neonatal seizure detection algorithm

Technologies for automated detection of neonatal seizures are gradually moving towards cot-side implementation. The aim of this paper is to present different ways to visualize the output of a neonatal seizure detection system and analyse their influence on performance in a clinical environment. Three different ways to visualize the detector output are considered: a binary output, a probabilistic trace, and a spatio-temporal colormap of seizure observability. As an alternative to visual aids, audified neonatal EEG is also considered. Additionally, a survey on the usefulness and accuracy of the presented methods has been performed among clinical personnel. The main advantages and disadvantages of the presented methods are discussed. The connection between information visualization and different methods to compute conventional metrics is established. The results of the visualization methods along with the system validation results indicate that the developed neonatal seizure detector with its current level of performance would unambiguously be of benefit to clinicians as a decision support system. The results of the survey suggest that a suitable way to visualize the output of neonatal seizure detection systems in a clinical environment is a combination of a binary output and a probabilistic trace. The main healthcare benefits of the tool are outlined. The decision support system with the chosen visualization interface is currently undergoing pre-market European multi-centre clinical investigation to support its regulatory approval and clinical adoption.

Neonatal seizure detection using atomic decomposition with a novel dictionary.

Atomic decomposition (AD) can be used to efficiently decompose an arbitrary signal. In this paper, we present a method to detect neonatal electroencephalogram (EEG) seizure based on AD via orthogonal matching pursuit using a novel, application-specific, dictionary. The dictionary consists of pseudoperiodic Duffing oscillator atoms which are designed to be coherent with the seizure epochs. The relative structural complexity (a measure of the rate of convergence of AD) is used as the sole feature for seizure detection. The proposed feature was tested on a large clinical dataset of 826 h of EEG data from 18 full-term newborns with 1389 seizures. The seizure detection system using the proposed dictionary was able to achieve a median receiver operator characteristic area of 0.91 (IQR 0.87-0.95) across 18 neonates.

EEG feature pre-processing for neonatal epileptic seizure detection.

Aim of our project is to further optimize neonatal seizure detection using support vector machine (SVM). First, a Kalman filter (KF) was used to filter both feature and classifier output time series in order to increase temporal precision. Second, EEG baseline feature correction (FBC) was introduced to reduce inter patient variability in feature distributions. The performance of the detection methods is evaluated on 54 multi channel routine EEG recordings from 39 both term and pre-term newborns. The area under the receiver operating characteristics curve (AUC) as well as sensitivity and specificity are used to evaluate the performance of the classification method. SVM without KF and FBC achieves an AUC of 0.767 (sensitivity 0.679, specificity 0.707). The highest AUC of 0.902 (sensitivity 0.801, specificity 0.831) is achieved on baseline corrected features with a Kalman smoother used for training data pre-processing and a KF used to filter the classifier output. Both FBC and KF significantly improve neonatal epileptic seizure detection. This paper introduces significant improvements for the state of the art SVM based neonatal epileptic seizure detection.

Robustness of time frequency distribution based features for automated neonatal EEG seizure detection.

In this paper we examined the robustness of a feature-set based on time-frequency distributions (TFDs) for neonatal EEG seizure detection. This feature-set was originally proposed in literature for neonatal seizure detection using a support vector machine (SVM). We tested the performance of this feature-set with a smoothed Wigner-Ville distribution and modified B distribution as the underlying TFDs. The seizure detection system using time-frequency signal and image processing features from the TFD of the EEG signal using modified B distribution was able to achieve a median receiver operator characteristic area of 0.96 (IQR 0.91-0.98) tested on a large clinical dataset of 826 h of EEG data from 18 full-term newborns with 1389 seizures. The mean AUC was 0.93.

Neonatal EEG audification for seizure detection.

Technologies for automated detection of neonatal seizures are gradually moving towards cot-side implementation. The aim of this paper is to present an alternative way to visualize the output of a neonatal seizure detection algorithm. For this purpose audified neonatal EEG is considered. The EEG is audified with the aid of the neonatal seizure detection algorithm which selects the representative channels for stereo audio image and controls the signal gain. A survey on the usefulness and accuracy of the presented audification method has been performed. The results of the audification method compare favourably to that of using amplitude integrated EEG for detection of neonatal seizures.

Advances in management of neonatal seizures.

Seizures are more common in the neonatal period than any other time in the human lifespan. A high index of suspicion for seizures should be maintained for infants who present with encephalopathy soon after birth, have had a stroke, central nervous system (CNS) infection or intracranial hemorrhage or have a genetic or metabolic condition associated with CNS malformations. Complicating the matter, most neonatal seizures lack a clinical correlate with only subtle autonomic changes and often no clinical indication at all. Over the last three decades, several tools have been developed to enhance the detection and treatment of neonatal seizures. The use of electroencephalography (EEG) and the later development of amplitude-integrated EEG (aEEG), allows for Neurologists and non-Neurologists alike, to significantly increase the sensitivity of seizure detection. When applied to the appropriate clinical setting, time to diagnosis and start of therapy is greatly reduced. Phenobarbital maintains the status of first-line therapy in worldwide use. However, newer anti-epileptic agents such as, levetiracetam, bumetanide, and topiramate are increasingly being applied to the neonatal population, offering the potential for seizure treatment with a significantly better side-effect profile. Seizures in premature infants, continue to confound clinicians and researchers alike. Though the apparent seizure burden is significant and there is an association between seizures and adverse outcomes, the two are not cleanly correlated. Compounding the issue, GABA-ergic anti-epileptic drugs are not only less effective in this age group due to reversed neuronal ion gradients but may cause harm. Selecting an appropriate treatment group remains a challenge.

Effective implementation of time–frequency matched filter with adapted pre and postprocessing for data-dependent detection of newborn seizures

Neonatal EEG seizures often manifest as nonstationary and multicomponent signals, necessitating analysis in the time–frequency (TF) domain. This paper presents a novel neonatal seizure detector based on effective implementation of the TF matched filter. In the detection process, the TF signatures of EEG seizure are extracted to construct the TF templates used by the matched filter. Matching pursuit (MP) decomposition and narrowband filtering are proposed for the reduction of artifacts prior to seizure detection. Geometrical correlation is used to consolidate the multichannel detections and to reduce the number of false detections due to remnant artifacts. A data-dependent threshold is defined for the classification of EEG. Using 30 newborn EEG records with seizures, the classification process yielded an overall detection accuracy of 92.4% with good detection rate (GDR) of 84.8% and false detection rate of 0.36FD/h. Better detection performance (accuracy >95%) was recorded for relatively long EEG records with short seizure events.

Robust neonatal EEG seizure detection through adaptive background modeling.

Adaptive probabilistic modeling of the EEG background is proposed for seizure detection in neonates with hypoxic ischemic encephalopathy. The decision is made based on the temporal derivative of the seizure probability with respect to the adaptively modeled level of background activity. The robustness of the system to long duration "seizure-like" artifacts, in particular those due to respiration, is improved. The system was developed using statistical leave-one-patient-out performance assessment, on a large clinical dataset, comprising 38 patients of 1479 h total duration. The developed technique was then validated by a single test on a separate totally unseen randomized prospective dataset of 51 neonates totaling 2540 h of duration. By exploiting the proposed adaptation, the ROC area is increased from 93.4% to 96.1% (41% relative improvement). The number of false detections per hour is decreased from 0.42 to 0.24, while maintaining the correct detection of seizure burden at 70%. These results on the unseen data were predicted from the rigorous leave-one-patient-out validation and confirm the validity of our algorithm development process.

Acute neonatal encephalopathy and seizures recurrence: A combined aEEG/EEG study

PurposeTo evaluate amplitude-integrated EEG (aEEG) in comparison with conventional (cEEG) for the identification of electrographic seizures in neonates with acute neonatal encephalopathies. MethodsThirty-one conventional cEEG/aEEG long-term recordings from twenty-eight newborns were reviewed in order to assess the electrographic seizure detection rate and recurrence in newborns. Two paediatric neurologists and one neonatologist, blinded to the raw full array cEEG, were asked to mark any events suspected to be an electrographic seizures on aEEG. They were asked to decide if the displayed aEEG trace showed the pattern of a single seizure (SS), repetitive seizures (RS) or status epilepticus (SE). Their ability to recognize electrographic seizures on aEEG was compared to seizures identified on full array cEEG. Results25 of the 31 long-term cEEGs recordings showed electrographic seizures. The two paediatric neurologists and the neonatologist identified SE in 100% of the reviewed traces using aEEG alone while they identified 49.4% and 37.5% of electrographic seizures using aEEG alone. Overall, the correct identification ranged from 23.5% to 30.7% for SS and 66% for RS. The inter-observer agreement (k) for the identification of SE for the two paediatric neurologists and the neonatologist was 1.0. Overall the inter-observer agreement (k) for the detection of SS, RS and SE of the two paediatric neurologists was 0.91. ConclusionsIn our study the observers identified SE in 100% of the reviewed traces using raw aEEG alone, thus aEEG might represent a useful tool to detect SE in the setting of NICU. SS may not be reliably identified using aEEG alone. Simultaneous recording of the raw cEEG/aEEG provides a good level of sensitivity for the detection of neonatal electrographic seizures.

Discriminative and Generative Classification Techniques Applied to Automated Neonatal Seizure Detection

A number of automated neonatal seizure detectors have been proposed in recent years. However, there exists a large variability in the morphology of seizure and background patterns, both across patients and over time. This has resulted in relatively poor performance from systems which have been tested over large datasets. Here, the benefits of employing a pattern recognition approach are discussed. Such a system may use numerous features paired with nonlinear classifiers. In particular, two types of nonlinear classifiers are contrasted for the task. Additionally, it is shown that the proposed architecture allows for efficient classifier combination which improves the performance of the algorithm. The resulting automated detector is shown to achieve field leading performance. A particular strength of the proposed algorithm is the performance of the algorithm when very low false detections are required, at 0.25 false detections per hour, the system is able to detect 75.4% of the seizure events.

Neonatal seizure identification on reduced channel EEG

Using a standard digital EEG system, we conducted simulations to determine the optimal locations and numbers of electrodes for seizure detection in neonates with reduced-channel EEG monitoring. The results showed...