Towards tonic seizure detection based on multimodal detection methods using the EpiSense sensor

J. van Sluis

    Research output: ThesisEngD Thesis

    150 Downloads (Pure)


    Tonic seizures in epilepsy are characterized by a severe continuous contraction of multiple muscle groups in the body, including the respiratory muscles. This type of seizure may lead to unconsciousness, cardiorespiratory depression, and in severe cases even to Sudden Unexpected Death in EPilepsy (SUDEP). Current detection methods are based on measurement of heartrate and movement. However, a tonic seizure is not always associated with a rise in heartrate. And little movement (or a lack thereof) makes current detection methods suboptimal for tonic seizure detection. Especially during night-time, when supervision is minimal, a real-time method for detecting tonic seizures is necessary.

    Design project
    The overall goal of this project is to design an algorithm for the detection of tonic seizures using the EpiSense system in order to apply it in patients’ home environments.
    Families and friends fear a tonic seizure, with its possible detrimental outcomes, to pass unnoticed during night-time. The use of a sensor for the detection of tonic seizures, could decrease anxiety amongst the patient’s loved ones. Because then, if a seizure does occur, direct medical intervention (which can be of vital importance) can be provided immediately if necessary.
    The EpiSense is a system developed by Kempenhaeghe in collaboration with Holst centre/Imec, and Eindhoven University of Technology (TU/e). This system consists of a multimodal sensor (measuring electromyography (EMG), electrocardiography (ECG), and accelerometry (ACM)) and a control tablet (with Android application software).

    Setup and implementation
    The complexity of this design project mainly lies in the designing, setting up, and performing of clinical trials with the newly developed sensor for the design of an adequate tonic seizure detection algorithm. Furthermore, prior to setting up and initiating the trials, the sensory system should first comply to specific requirements for safety and to be able to collect high-quality data. Subsequently, clinical trial protocols must be written (and adjusted) for approval from ethical committees to execute the clinical trials (phase 1: intramural, phase 2: extramural). For setting up and performing the clinical trials a close collaboration between the investigator, neurologists and other healthcare providers is necessary. Ultimately, after sufficient measurement data have been collected, a detection algorithm should be designed (considering all three measurement modalities of the EpiSense sensor) and validated.

    During this project, first the EpiSense system has been tested and re-designed. Next it has been used in the first clinical trials, both intramural (in Kempenhaeghe) and extramural (at the long-stay facilities of Kempenhaeghe). Unfortunately, the development of an adequate detection algorithm has been hindered by the relatively small amount of measured tonic seizures captured during these clinical trials. In the trials, only 11 seizures with a tonic component were collected in 11 patients over approximately 800 hours of measurements. For the moment, it can be concluded that solely based on EMG measurements, nocturnal tonic seizures can be detected with a sensitivity (SEN) of 72.7% and a positive predictive value (PPV) of 80%.
    In this study, additional ACM and ECG measurements (synchronously measured with the corresponding EMG-data) have not been of extra value for the detection of nocturnal tonic seizures.

    Based on this first preliminary analysis and limited data it may be concluded that EMG measurements could be valuable in detecting nocturnal tonic seizures.

    The loss of the ECG signal in most recordings as well as the loss of the EMG signal in some recordings during these clinical trials were caused by letting loose of the skin electrodes.
    Furthermore, the conclusions in this design project have been drawn on a very small number of measured tonic seizures.

    Follow-up studies in which more patients can be measured during longer periods of time should validate this new detection algorithm and prove whether the sensor can effectively be used in the patients’ home environments.
    To be able to include heartrate data in future analysis, skin electrodes should be replaced by photo plethysmography (PPG) based heart rate measurement to avoid loss of the signal through electrodes letting loose (as was the case during this project). A suggestion for expansion of the detection algorithm to be able to include ACM measurements in the detection algorithm as well, is to get rid of the ever-changing offset in the ACM signal.
    Original languageEnglish
    • van Dijk, J.P. (Hans), Supervisor
    • Lammerts, Ivonne M.M., Supervisor
    Award date31 Jan 2018
    Place of PublicationEindhoven
    Publication statusPublished - 31 Jan 2018

    Bibliographical note

    PDEng thesis.


    Dive into the research topics of 'Towards tonic seizure detection based on multimodal detection methods using the EpiSense sensor'. Together they form a unique fingerprint.

    Cite this