Computer Vision to Detect Epileptic Seizures in Video

  • Research type

    Research Study

  • Full title

    Computer Vision and Machine Learning to Detect Epileptic Seizures in Video

  • IRAS ID

    270705

  • Contact name

    Parisa Patel

  • Contact email

    parisa.patel@nhs.net

  • Sponsor organisation

    Leeds Teaching Hospitals NHS Trust

  • Duration of Study in the UK

    3 years, 0 months, 1 days

  • Research summary

    People who suffer from epilepsy have an increased risk of mortality and there are approximately 400 preventable deaths each year from epilepsy in the UK. Improved detection of seizures could improve patient care and reduce mortality/morbidity from seizures by allowing faster management of seizures and detecting patterns of seizures (allowing medication plans to be better tailored to patients). \n\nCurrently there are many different seizure alarm systems to help patients and healthcare professionals detect seizures, but they have a low rate of accuracy (many ’false alarms’ and many seizures undetected) and/or they require specialised equipment making them expensive. EEG recording (in which surface electrodes applied to the scalp detect electrical discharges occurring in the brain) combined with continuous video recording is a test known as ‘video-EEG’. Epileptic patients often come into hospital for a video-EEG recording for several days, with the aim of recording a seizure (on EEG and video), either to better locate the seizure (for planning surgery) or to help to diagnose seizures from other types of attack. Our study aims to use this video-EEG data and artificial intelligence (’machine learning’) techniques so that a computer programme can learn to detect the movements on video that occur during a seizure (which is reliably detected by the EEG electrodes). If these movements can be learnt, it could provide a new way to detect seizures (’seizure alarm’), simply using a camera, with no other specialised equipment.\n

  • REC name

    Yorkshire & The Humber - Bradford Leeds Research Ethics Committee

  • REC reference

    19/YH/0393

  • Date of REC Opinion

    13 Nov 2019

  • REC opinion

    Favourable Opinion