Epilepsy detection

About the project

Improving diagnostic procedures for epilepsy through automated recording and analysis of patients.

This PhD project combines computing, quantitative and qualitative clinical research to develop a fully-automated diagnostic stratification method for patients first presenting with TLOC in emergency or primary care settings.

First, the student will adapt a previously developed computer-presented talking head ('digital doctor') to ask patients presenting with transient loss of consciousness (TLOC) open and closed questions about their experience and interpret their answers using automatic speech recognition and classification. The questions will include 36 symptom-based questions, which classified 86% of patients correctly to one of the common causes of TLOC in previous modelling.

In the second stage of the project the digital doctor will be trained and tested in a clinical study involving 200 consecutive new referrals to specialist syncope or seizure clinics. In addition to seeking to validate the 36-item stratification tool for clinical use, this study will explore whether the automatic analysis of interactional, linguistic or phonetic features extracted from patients' answers to open questions will enhance the diagnostic prediction performance of the closed symptom-based enquiries.

Additional qualitative work with research users will ensure that the digital doctor will be acceptable to patients.