Automatic Seizure Detection Algorithm Research Articles

S127. Computer aided evaluation of intracerebral depth electrode recordings by automatic seizure detection

Introduction Preoperative evaluation of patients with medically intractable epilepsies requires accurate localization of the epileptogenic zone. Intracerebral depth electrodes are used to asses regions of interest and to enclose the resection zone during presurgical planning. Often a high number of needles is needed to accurately define the resection volume. Manual evaluation of these prolonged recordings including a large number of channels which is highly time consuming. To raise efficiency we developed a computer algorithm for automatic detection of epileptic seizures in depth electrode recordings. Primary goal was to detect seizures with high sensitivity without the need to set patient specific parameters. Methods The automatic seizure detection algorithm for depth electrode data was based on an existing seizure detection algorithm for surface EEG. Evaluation of the frequency and amplitude range was extended to allow detection of ictal activity of up to 25 Hz and amplitudes of up to 1 mV. To reduce the number of false detections a method to recognize loose contacts was implemented. For clinical validation, recordings of 11 patients that underwent depth electrode investigation in the Academic Centre of Epileptology, Kempenhaeghe were utilized. Recordings were evaluated manually by clinical neurophysiologists and seizures were annotated. For 10 patients the first 24 h of their depth electrode recording were analyzed, yielding in total 23 seizures detected for five of the patients studied. For one patient the complete recording was analyzed with a duration of 138 h, yielding 36 seizures. Manual seizure annotations were compared to computer detections for assessment of sensitivity and false detection rate. Results The automatic seizure detection algorithm found 84% of all seizures on average. All or more than 80% of the seizures were detected in the first 24 h of 10 patients. Analysis of the complete patient recording showed suppressed seizure onset activity of less than 2 s duration followed by paroxysmal fast activity. In this patient 14 out of the 36 seizures that evolved into a rhythmic ictal pattern were detected. The average false detection rate was 15 false detections in 24 h. Validation showed that most of the false positives either point to interictal activity or to background alpha activity. Conclusion We proposed a computer aided workflow for evaluation of depth electrode recordings. Our automatic seizure detection algorithm was able to detect either all or more than 7 seizures of a patient which allows deduction of important medical findings. The low false detection rate facilitates fast review of data. Our proposed approach showed that automatic evaluation of seizure activity in depth electrode recordings is feasible and will raise the overall efficiency of diagnostic.

Mouse epileptic seizure detection with multiple EEG features and simple thresholding technique

Objective. Epilepsy is a common neurological disorder characterized by recurrent, unprovoked seizures. The search for new treatments for seizures and epilepsy relies upon studies in animal models of epilepsy. To capture data on seizures, many applications require prolonged electroencephalography (EEG) with recordings that generate voluminous data. The desire for efficient evaluation of these recordings motivates the development of automated seizure detection algorithms. Approach. A new seizure detection method is proposed, based on multiple features and a simple thresholding technique. The features are derived from chaos theory, information theory and the power spectrum of EEG recordings and optimally exploit both linear and nonlinear characteristics of EEG data. Main result. The proposed method was tested with real EEG data from an experimental mouse model of epilepsy and distinguished seizures from other patterns with high sensitivity and specificity. Significance. The proposed approach introduces two new features: negative logarithm of adaptive correlation integral and power spectral coherence ratio. The combination of these new features with two previously described features, entropy and phase coherence, improved seizure detection accuracy significantly. Negative logarithm of adaptive correlation integral can also be used to compute the duration of automatically detected seizures.

Improved neonatal seizure detection using adaptive learning.

In neonatal intensive care units performing continuous EEG monitoring, there is an unmet need for around-the-clock interpretation of EEG, especially for recognizing seizures. In recent years, a few automated seizure detection algorithms have been proposed. However, these are suboptimal in detecting brief-duration seizures (<; 30s), which frequently occur in neonates with severe neurological problems. Recently, a multi-stage neonatal seizure detector, composed of a heuristic and a data-driven classifier was proposed by our group and showed improved detection of brief seizures. In the present work, we propose to add a third stage to the detector in order to use feedback of the Clinical Neurophysiologist and adaptively retune a threshold of the second stage to improve the performance of detection of brief seizures. As a result, the false alarm rate (FAR) of the brief seizure detections decreased by 50% and the positive predictive value (PPV) increased by 18%. At the same time, for all detections, the FAR decreased by 35% and PPV increased by 5% while the good detection rate remained unchanged.

Automatic multimodal detection for long-term seizure documentation in epilepsy.

This study investigated sensitivity and false detection rate of a multimodal automatic seizure detection algorithm and the applicability to reduced electrode montages for long-term seizure documentation in epilepsy patients. An automatic seizure detection algorithm based on EEG, EMG, and ECG signals was developed. EEG/ECG recordings of 92 patients from two epilepsy monitoring units including 494 seizures were used to assess detection performance. EMG data were extracted by bandpass filtering of EEG signals. Sensitivity and false detection rate were evaluated for each signal modality and for reduced electrode montages. All focal seizures evolving to bilateral tonic-clonic (BTCS, n=50) and 89% of focal seizures (FS, n=139) were detected. Average sensitivity in temporal lobe epilepsy (TLE) patients was 94% and 74% in extratemporal lobe epilepsy (XTLE) patients. Overall detection sensitivity was 86%. Average false detection rate was 12.8 false detections in 24h (FD/24h) for TLE and 22 FD/24h in XTLE patients. Utilization of 8 frontal and temporal electrodes reduced average sensitivity from 86% to 81%. Our automatic multimodal seizure detection algorithm shows high sensitivity with full and reduced electrode montages. Evaluation of different signal modalities and electrode montages paces the way for semi-automatic seizure documentation systems.

Crowdsourcing seizure detection: algorithm development and validation on human implanted device recordings.

There exist significant clinical and basic research needs for accurate, automated seizure detection algorithms. These algorithms have translational potential in responsive neurostimulation devices and in automatic parsing of continuous intracranial electroencephalography data. An important barrier to developing accurate, validated algorithms for seizure detection is limited access to high-quality, expertly annotated seizure data from prolonged recordings. To overcome this, we hosted a kaggle.com competition to crowdsource the development of seizure detection algorithms using intracranial electroencephalography from canines and humans with epilepsy. The top three performing algorithms from the contest were then validated on out-of-sample patient data including standard clinical data and continuous ambulatory human data obtained over several years using the implantable NeuroVista seizure advisory system. Two hundred teams of data scientists from all over the world participated in the kaggle.com competition. The top performing teams submitted highly accurate algorithms with consistent performance in the out-of-sample validation study. The performance of these seizure detection algorithms, achieved using freely available code and data, sets a new reproducible benchmark for personalized seizure detection. We have also shared a 'plug and play' pipeline to allow other researchers to easily use these algorithms on their own datasets. The success of this competition demonstrates how sharing code and high quality data results in the creation of powerful translational tools with significant potential to impact patient care.

Detection of Paroxysms in Long-Term, Single-Channel EEG-Monitoring of Patients with Typical Absence Seizures.

Absence seizures are associated with generalized 2.5–5 Hz spike-wave discharges in the electroencephalogram (EEG). Rarely are patients, parents, or physicians aware of the duration or incidence of seizures. Six patients were monitored with a portable EEG-device over four times 24 h to evaluate how easily outpatients are monitored and how well an automatic seizure detection algorithm can identify the absences. Based on patient-specific modeling, we achieved a sensitivity of 98.4% with only 0.23 false detections per hour. This yields a clinically satisfying performance with a positive predictive value of 87.1%. Portable EEG-recorders identifying paroxystic events in epilepsy outpatients are a promising tool for patients and physicians dealing with absence epilepsy. Albeit the small size of the EEG-device, some children still complained about the obtrusive nature of the device. We aim at developing less obtrusive though still very efficient devices, e.g., hidden in the ear canal or below the skin.

Hardware-friendly seizure detection with a boosted ensemble of shallow decision trees.

Efficient on-chip learning is becoming an essential element of implantable biomedical devices. Despite a substantial literature on automated seizure detection algorithms, hardware-friendly implementation of such techniques is not sufficiently addressed. In this paper, we propose to employ a gradientboosted ensemble of decision trees to achieve a reasonable trade-off between detection accuracy and implementation cost. Combined with the proposed feature extraction model, we show that these classifiers quickly become competitive with more complex learning models previously proposed for hardware implementation, with only a small number of low-depth (d <; 4) "shallow" trees. The results are verified on more than 3460 hours of intracranial EEG data including 430 seizures from 27 patients with epilepsy.

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.

Validation of an automated seizure detection algorithm for term neonates

ObjectiveThe objective of this study was to validate the performance of a seizure detection algorithm (SDA) developed by our group, on previously unseen, prolonged, unedited EEG recordings from 70 babies from 2 centres. MethodsEEGs of 70 babies (35 seizure, 35 non-seizure) were annotated for seizures by experts as the gold standard. The SDA was tested on the EEGs at a range of sensitivity settings. Annotations from the expert and SDA were compared using event and epoch based metrics. The effect of seizure duration on SDA performance was also analysed. ResultsBetween sensitivity settings of 0.5 and 0.3, the algorithm achieved seizure detection rates of 52.6–75.0%, with false detection (FD) rates of 0.04–0.36FD/h for event based analysis, which was deemed to be acceptable in a clinical environment. Time based comparison of expert and SDA annotations using Cohen’s Kappa Index revealed a best performing SDA threshold of 0.4 (Kappa 0.630). The SDA showed improved detection performance with longer seizures. ConclusionThe SDA achieved promising performance and warrants further testing in a live clinical evaluation. SignificanceThe SDA has the potential to improve seizure detection and provide a robust tool for comparing treatment regimens.

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

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

Additional value of two-channel amplitude integrated EEG recording in full-term infants with unilateral brain injury

BackgroundAmplitude integrated electroencephalography (aEEG) is a valuable tool for evaluating neonatal encephalopathy and identifying electrographic seizures.ObjectiveTo compare seizure activity and background pattern (BGP) between one-channel and two-channel aEEG recordings in...

Evaluation of automated seizure detection algorithms in neonates

Evaluation of automated seizure detection algorithms in neonates