Cardiac-based detection of seizures in children with epilepsy
Introduction
Mortality risk within the population with epilepsy is 2–3 times higher than in a healthy population [1]. The ability to detect a seizure and alert a caregiver would enable timely intervention for mitigating adverse events including traumatic accidents and sudden unexpected death in epilepsy (SUDEP) which together account for over 60% of epilepsy-related deaths [1]. Seizure detection also enables accurate logging of seizures to support drug trials and patient medication management.
Electroencephalogram (EEG) monitoring is the gold standard method for detecting seizures in the clinical setting, but is not convenient for daily monitoring due to the requirement for electrodes on the head and a large volume of data to analyze. Non-EEG-based research devices and commercially available systems aimed at detecting seizures and warning caregivers have taken many different approaches [2], [3], [4]. Systems that identify abnormal movements related to convulsive seizures using sensors that are worn or placed on a mattress are the most prevalent. Movement-based systems can achieve high sensitivity (∼90%) for detecting generalized tonic-clonic seizures (GTCS), but are prone to false detections (1–10/day) [2], [5]. Improved specificity has been achieved using electrodermal activity (EDA), which reflects sympathetic regulation of sweat gland activity, in combination with movement-based detection. Detection with over 90% sensitivity for GTCS and a low false-positive rate (0.2–1/day) has been reported with this approach [6], [7] which is commercially available in the Empatica Embrace device.
Not all seizures involve shaking movements, but other physiological changes may manifest which may be leveraged for detection including autonomic and cardiovascular functions that can be monitored with unobtrusive sensors [3], [8], [9]. Systems leveraging these physiological changes may also be beneficial for SUDEP prevention because underlying arrhythmias and autonomic nervous system (ANS) dysfunction have been proposed as possible contributors to SUDEP [10]. A variety of methods have been reported to characterize cardiac changes during seizures. Algorithms using heartrate (HR) alone can achieve over 90% sensitivity, but tend to be error prone (i.e., over 1FP/hour) [3]. Algorithms using heart rate variability (HRV) measures can also achieve high sensitivity, but again false-positive rates are high (0.49–5.4/h) [3], [11]. Developing generalizable algorithms using cardiac data is challenging because cardiac profiles during both seizure and non-seizure periods can vary considerably between individuals and within-individuals for different types of seizures [12], [13], [14]. Adding patient specificity to algorithms improves performance [11], [15], [16], but can be difficult to implement as existing seizure data for each specific patient are required. A possible solution is the use of an initial patient-independent algorithm that is automatically adapted [17].
Another challenge of a real-time seizure detection based on cardiac activity is developing a fully automated algorithm that is robust to variations in signal quality. Approaches demonstrated on manually corrected HR data or evaluated only on short segments around seizures [13], [16] may not translate well to realistic monitoring applications. Additionally, some methods for quantifying ANS function are resource intensive, making real-time implementation on a mobile platform difficult. Consequently, few groups have demonstrated integrated systems [18], [19], [20], and these have only been validated with small numbers of users.
In this study, we present a real-time seizure algorithm that can be used with a variety of commercial sensors that provide accurate interbeat interval (IBI) data. The algorithm was developed with data from a single-lead, chest-worn electrocardiogram (ECG) sensor with an integrated accelerometer. Seizures were detected using a combination of simple, patient-independent HR and HRV indicators of elevated autonomic activity. Motion-related parameters were also explored. We evaluated algorithm performance across a range of seizure types and achieved high sensitivity with a low false-positive rate. We implemented the algorithm on a custom standalone unit that received data streamed from the chest-worn ECG sensor in real-time. This prototype unit could be replaced by a smartphone in future applications.
Section snippets
Data collection
Children aged 2–17 years with active epilepsy with all seizure types and undergoing video-EEG evaluation at Children’s National Hospital (CNH) as part of clinical care were recruited into the study after written informed consent, and assent when applicable, was obtained. Sensor devices were attached to the chest using disposable ECG electrodes, and the time on the devices and the EEG computer were aligned. The entire EEG recordings were evaluated by the epileptologist at least every 24 h as
Results
Data were collected on 62 children at CNH admitted for long term video-EEG monitoring. There were 38 males and 25 females ranging in age from 2 to 17 with a mean age of 10.2 years. Eight patients were excluded from analysis because of improper device usage resulting in missing data or poor signal quality. Signal quality issues were attributed to degraded electrical connections in some of the Zephyr Biopatch sensors and were eliminated with the switch to the Faros 180 devices in the latter part
Discussion
Our multi-parametric approach shows promise for accurate, real-time detection of different types of seizures in a varied population. The dataset used for algorithm development and evaluation included examples of GTCS, tonic, clonic, and focal seizures without motor activity from 18 pediatric patients and 5 adult patients. Similar to the findings described in [15], [18], we noted significant differences in the time course cardiac changes, and these variations may be attributed to large
Conclusions
The data presented here suggest that cardiac monitoring may detect a wider range of seizure types than motion-based approaches. While high detection sensitivity for seizures with generalized motor features (not myoclonic) can be accomplished with a patient-independent algorithm, detection of focal seizures may be improved with individualized settings.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This work was supported by the National Institute of Biomedical Imaging and Bioengineering of the National Institutes of Health under Award Number R01EB014742.
References (26)
- et al.
Non-EEG seizure-detection systems and potential SUDEP prevention: State of the art
Seizure
(2013) - et al.
Non-EEG seizure detection systems and potential SUDEP prevention: State of the art: review and update
Seizure
(2016) - et al.
Multimodal wrist-worn devices for seizure detection and advancing research: focus on the Empatica wristbands
Epilepsy Res
(2019) - et al.
Incidence and mechanisms of cardiorespiratory arrests in epilepsy monitoring units (MORTEMUS): a retrospective study
Lancet Neurol
(2013) - et al.
Can ECG monitoring identify seizures?
J Electrocardiol
(2015) - et al.
A model of heart rate changes to detect seizures in severe epilepsy
Seizure
(2006) - et al.
of epileptic seizures with a modified heart rate variability algorithm based on Lorenz plot
Seizure
(2015) - et al.
Adaptive nocturnal seizure detection using heart rate and low-complexity novelty detection
Seizure
(2018) - et al.
A prospective, multicenter study of cardiac-based seizure detection to activate vagus nerve stimulation
Seizure
(2015) - et al.
Statistical strategies to quantify respiratory sinus arrhythmia: are commonly used metrics equivalent?
Biol Psychol
(2012)
Multimodal seizure detection: A review
Epilepsia
Ictal autonomic changes as a tool for seizure detection: a systematic review
Clin Auton Res
Movement-based seizure detection
Epilepsia
Cited by (5)
Non-electroencephalogram-based seizure detection devices: State of the art and future perspectives
2023, Epilepsy and BehaviorModular Open-Core System for Collection and Near Real-Time Processing of High-Resolution Data from Wearable Sensors
2023, Applied System InnovationSeizure Detection Mechanism in Children
2023, International Conference on Applied Intelligence and Sustainable Computing, ICAISC 2023