Long-term home monitoring of hypermotor seizures by patient-worn accelerometers
Highlights
► First to focus on nocturnal hypermotor seizures in pediatric patients. ► Ambulant movement detection compared to video/EEG-based detection. ► Sensitivity of 95.71% and PPV of 57.84%.
Introduction
Epilepsy is a serious neurological condition affecting almost one percent of the world population of which 30% suffers from refractory or intractable (uncontrollable) seizures [1], [2]. In order to both assist such patients and reassure parents or caregivers, a seizure detection system is needed that can be used in a home situation. This will entail the use of long-term monitoring and creating a system that is both easy to install and as unobtrusive as possible. It would permit actions such as putting the patient in a correct position (e.g., prone to supine to prevent suffocation), stimulating the patient by e.g., rolling him or her over (passive muscle movement), cardiopulmonary resuscitation and defibrillation, medication administration, clearing of the airway, and seeking help if needed. Furthermore, it would allow the parent or caregiver to reassure the patient and to clean the bed in case of vomiting or incontinence.
The gold standard of video/electro-encephalogram (EEG) monitoring does not fulfill the above mentioned criteria. It is a short-term measurement. It is uncomfortable (electrodes attached to scalp) and professional installation and supervision are required. In addition, a system that allows automated seizure detection with sufficient sensitivity (seizure detection) and positive predictive value (PPV, false positive rejection) based on this intracerebral signal has yet to be developed.
Given that extracerebral body signals are also altered by seizures, they provide the opportunity for the development of detection systems that fulfill the home monitoring requirements and can be available sooner. The authors have experienced that the availability of commercial extracerebral seizure detection devices in Belgium is limited. Those that are available are aimed at detecting tonic-clonic seizures or ‘violent movement’ only. Aside from this limitation, the authors partake in the general criticism [3] that under mattress- or bed-attached systems (the most widespread) evoke many false positive alerts.
The authors consist of a multidisciplinary team composed of engineers and caregivers from different settings in Flanders, Belgium: a university, university hospitals, a university college and a rehabilitation center. The team focuses on accelerometer (ACM) [4], [5], [6] and video [7], [8], [9], [10] recordings since they are commonly used (in other fields) for movement detection and since many epileptic seizures have a motor component, often early in the course of the event [11]. Three-axial ACM measures the acceleration in three orthogonal directions fixed to the sensor.
The authors focused on hypermotor seizures (in pediatric patients) as described below given the availability of data, their occurrence during sleep, their often refractory nature and the possible serious consequences such as dizziness, headache and confusion, or injuries because of the uncontrolled movements, or even death. Furthermore, hypermotor seizures are manifestations of nocturnal frontal lobe seizures, known to have subtle or absent (the latter in about 50%) ictal patterns on scalp EEG [12]. See Fig. 1 for different ictal patterns occurring in the seven patients that were monitored.
Hypermotor or hyperkinetic seizures are convulsive seizures with a frontal lobe onset and consist of natural movements in abnormal circumstances or with an abnormal violence or repetition, mainly including bilateral (but often not coordinated) proximal limbs or axial muscles (opposed to distal limbs and oral muscles in temporal automotor seizures), e.g., pedaling, flailing, pelvic thrusting, thrashing (swim kicking), and boxing [13], [14], [15]. The onset is often from non-REM sleep, mainly during stages 1 and 2 (at a significantly higher frequency than and opposed to onset from later stages for arousal disorders) and stereotyped in one patient (opposed to being quite variable in arousal disorders) throughout the night, and can last from a couple of seconds to some minutes. Post-ictal confusion is often absent [16], [17] which means that the patients often recall a ‘strange feeling’ and need comforting.
Monitoring was conducted during patient's sleep as hypermotor and other (mainly frontal lobe) [18] seizures occur mainly at night. It is also advantageous as there is less noise from voluntary movement and because seizures occur more or less in a controlled reproducible manner. Development of a nocturnal detection system is also important because of reduced or no supervision at night.
The goal of this study was to examine the added value of ACM for hypermotor seizure detection in seven pediatric patients.
Section snippets
Database
The authors have, thus far, monitored 65 pediatric patients, obtaining a total of 1020 motor seizures of different types for a period of 192 nights (typically from 20:00 in the evening until 8:00 in the morning) by EEG (electro-encephalography), ECG (electrocardiography), EMG (electromyography, of upper arms), video, audio and ACM (accelerometry, of wrists and ankles) (Fig. 2). Approval by the Medical Ethical Commission of the Antwerp University Hospital, Belgium and signed informed consent
Results
Positive predictive value (PPV) is calculated next to specificity, as it is more sensitive to false positives and because the data sets contain a much higher frequency of true negative than true positive events. For seizure detection systems, it is important that there are no false negatives (sensitivity 100%), and that the amount of incorrectly detected seizures is as low as possible (high PPV).
The results of the testing on 2 seizures and 33 normal movements per patient are given in Table 1.
Discussion
There is an obvious need for ambulant alarm systems that allow for long-term and home monitoring of seizures. The authors' research contributes by focusing on hypermotor seizure detection in pediatric patients using an accelerometer (ACM) attached to wrists and ankles. The seizure detection algorithm was tested on 7 patients and 51 seizures labeled based on video and electro-encephalography (EEG) and resulted in a mean sensitivity of 95.71% and a mean positive predictive value (PPV) of 57.84%.
Acknowledgment
Anouk Van de Vel is funded by TBM grant 070713 of the Agency for Innovation by Science and Technology in Flanders (IWT-Vlaanderen).
The authors would like to thank Dr. Kristien Verhaert, Ria Krols, Annick Fonteyne, Lut Gijsemans, Jill Van de Ven, Nathalie Baert, Jan Vervisch and Katrien Lemmens for their help with the measurements in the Epilepsy Centre for Children and Youth, Pulderbos.
References (24)
- et al.
Quantitative movement analysis differentiates focal seizures characterized by automatisms
Epilepsy Behav
(2011) - et al.
Movement disorders in sleep: guidelines for differentiating epileptic from non-epileptic motor phenomena arising from sleep
Sleep Med Rev
(2007) - et al.
Measurement and quantification of generalized tonic-clonic seizures in epilepsy patients by means of accelerometry—an explorative study
Epilepsy Res
(2011) - et al.
Detection of seizure-like movements using a wrist accelerometer
Epilepsy Behav
(2011) - et al.
Miniaturized wireless ECG-monitor for real-time detection of epileptic seizures
Sudden unexpected death in epilepsy: continuing the global conversation
- et al.
A magnetometer-based approach for studying human movements
IEEE Trans Biomed Eng
(2007) - et al.
Detection of nocturnal frontal lobe seizures in pediatric patients by means of accelerometers: preliminary results
Conf Proc IEEE Eng Med Biol Soc
(2007) - et al.
Detection of nocturnal frontal lobe seizures in pediatric patients by means of accelerometers: a first study
Conf Proc IEEE Eng Med Biol Soc
(2009) - et al.
Accelerometers used for the detection of nocturnal frontal lobe seizures in pediatric patients