Safe and Sound? A systematic literature review of seizure detection methods for personal use
Introduction from Dr Markus Reuber, editor-in-chief of Seizure
One of the most troublesome aspects of epilepsy is the unpredictability of seizures. If people living with epilepsy knew when their seizures were going to happen they could try pharmacological or other intervention techniques, make sure they were safe and consider calling for help. Unfortunately, the early disruption of self-control or awareness is a feature of many types of seizures, and seizure warnings may be too short or non-specific to allow people to take any useful action. Even those individuals who do notice the onset of their seizures may not be able recall this (or any other part of their seizure) subsequently because seizure-related changes have stopped their brain from laying down a secure memory trace of their seizure. In fact, over one half of seizures objectively recorded during video-EEG recordings were not reported by hospital in-patients likely to be highly focussed on noticing their seizures (1).
However, the fact that epileptic seizures are associated with a host of observable alterations (such as unusual movements, changes in breathing, posture or skin colour) as well as neurobiological changes (such as in electroencephalographic or electrocardiographic signals) should make the detection of seizures by technological devices straightforward and reliable. So, with the use of technology, all should be well.
The review article that is my editor’s choice in this issue of Seizure by Jory et al. demonstrates that things are not quite that straightforward – at least not yet (2). For a start there are many different types of seizures – some, for instance, involve marked changes in motor activity and some not. To complicate matters, one person may have different seizure types. Even seizures in the same individual, looking superficially similar, may sometimes be associated with biological changes (such as an increase in heart rate) and sometimes not. This means that simple technological solutions using only one mode of observation, such as movement sensors will have limited sensitivity. If the sensitivity is increased by changing device settings to interpret less significant changes as seizures, the specificity is reduced – in other words, the number of false alarms increases.
It is likely that devices will need to use multimodal monitoring to capture seizures with sufficient accuracy to be useful for general clinical use. An ideal seizure detection device would need to be adaptable to the specific seizures of a particular person and should probably be able to learn as well. The challenges for developers will not stop once the device has been invented. As the paper by Jory et al. demonstrates, there is also plenty of scope for the improvement of future utility and acceptability studies.
(1) Hoppe C, Poepel A, Elger CE, et al. Epilepsy: accuracy of patient seizure counts. Arch Neurol 2007;64:1595-1599.
(2) Jory C, Shankar R, Coker D, McLean B, Hanna J, Newman C. Safe and Sound? A systematic literature review of seizure detection methods for personal use. Seizure 2016;36:4-15.