The incidence of sudden unexpected death in epilepsy (SUDEP) in young adults is many times higher than the general population (standardised mortality ratio of 23.7), with a peak between the ages of 20 and 40 years and an incidence quoted between 2-5.9/1000 person-years in patients attending specialist epilepsy clinics. A number of factors are associated with increased risk of SUDEP, including sleeping alone and nocturnal seizures.
Alarm systems have been developed to address the risk of nocturnal seizures. These are usually “bed alarm” systems, actigraphy (wrist worn accelerometers), EEG, or heart, respiratory or temperature monitors. However, these are not specific for seizure and may identify normal, nocturnal movements as seizure to cause false alarms. If the sensitivity is changed to limit these alarms, the ability to detect an epileptic seizure may be impaired.
We aim to develop and validate a computer vision system using machine learning algorithms to detect nocturnal seizures, differentiating them from normal sleep behaviours.
Using patient videos from NHS Tayside, we intend to define precise performance criteria and targets in order to enable quantitative assessment of the software developed. This will require annotations of videos, e.g. segments in which the patient is having a seizure and concise descriptions as necessary. These videos will be split into disjoint development and test sets. The development set will be used to train and tune the prototype system. The test set will be used to validate performance. We expect to access video and EEG data from at least 50 patients who have undergone approximately 5 days and 4 nights of study. Based on our experience, this seems a feasible quantity for a pilot aimed to develop a machine learning system over the projected timeframe of 3 years.
The student will :
• receive training in image processing, machine learning and AI in the CVIP group (Computing), and in clinical aspects relevant for the
project from Morrison;
• join the CVIP research group and its activities, including journal club, bi-annual technical group workshops, and participation in
international medical imaging challenges;
• access training opportunities in SSEN and SoM, and at university level;
• undertake the usual regular progress assessment iter in Computing or Medicine (TBD).
Yuanfa Wang, Zunchao Li, Lichen Feng, Chuang Zheng, and Wenhao Zhang, “Automatic Detection of Epilepsy and Seizure Using Multiclass Sparse Extreme Learning Machine Classification,” Computational and Mathematical Methods in Medicine, vol. 2017, Article ID 6849360, 10 pages, 2017. https://doi.org/10.1155/2017/6849360.
A Karbouch, Automatic detection of epileptic seizure onset and termination using intracranial EEG, [email protected]
, 2012. http://hdl.handle.net/1721.1/75638
A Ulate-Campos, F Coughlin, M Gainza-Lein et al.: Automated seizure detection systems and their effectiveness for each type of seizure. Seizure, Vol 40, August 2016, 88-101. https://doi.org/10.1016/j.seizure.2016.06.008 .