Epilepsy is characterised by seizures which are caused by excessive electrical activity in the brain and can often result in loss of consciousness for a short duration of time. This is referred to as a Transient Loss of Consciousness (TLOC) and often involves abnormal activity in the nervous system and memory loss around the unconscious period. It is one of the most common reasons why people access emergency care services. However, a TLOC can also be caused by cardiovascular or psychological factors. It is very important to make the correct diagnosis, because the correct treatment depends on knowing what caused TLOC. Tests often don’t help much with the diagnosis; about 20% of individuals receive an incorrect diagnosis and the correct diagnosis may be delayed because the wrong tests are carried out initially.
TLOC may be associated with different symptoms, depending on whether it was caused by epileptic activity, fainting or a psychological process. The most important tool for the differentiation between the common causes of TLOC is the history from patients. Our research is investigating whether it is possible to reliably predict what caused a TLOC by analysing what patients can tell us about these symptoms using artificial intelligence methods.
As part of the project, we have created a website hosted by the University of Sheffield where potential participants can learn about our research and express an interest in participating. The study activities are all completed online, the first of which is to complete a short questionnaire about their symptoms.
Computer modelling suggests that the questionnaire (the ‘iPEP’) will correctly predict the underlying cause of TLOC in 86% of cases. We want to investigate whether this level of accuracy can be demonstrated in a real life setting among patients first referred to seizure or syncope (TLOC) clinics.
After completing this questionnaire, participants then have a consultation with a “digital doctor” – a talking head that asks questions about the TLOC and records the patient’s verbal descriptions of what happened. Previous qualitative research has identified differences in how patients with epilepsy and non-epileptic seizures talk about their seizure. Automated language analysis methods will be used to examine whether differences in the spoken answers of our research participants can improve the diagnostic accuracy of the iPEP.
Our study has turned out to be an ideal project for our “socially distanced” times. Whereas many clinical research have had to stop completely, and laboratory research is impeded by infection control measures, we have been able to stick to our research plan. What is more, our training of an automatic speech recognition algorithm translating TLOC talk into text and the programming work we are doing on our diagnostic classifier, is research that will be extremely useful if social distancing is here to stay for the foreseeable future.
Find out more about this cutting edge research here.