Evaluation of wearable epileptic seizure monitors

Abstract

The research presented in this thesis contributes to the evaluation and future evolution of wearable epilepsy seizure monitoring devices with a systematic literature review and three research studies that include two device evaluation studies, and a survey of stakeholder opinions and experiences of wearable epilepsy monitoring devices.

The thesis comprises background literature relevant to epilepsy, wearable technology, seizure monitoring and device evaluation. This review is followed by chapters for the systematic literature review and the three research studies.

The systematic review is focused on evaluations of wearable epilepsy seizure monitors in the academic literature. It demonstrates that although there are over 3000 works in the literature proposing and evaluating novel and incremental approaches to epilepsy seizure detection, there are very few that report evaluations of available devices and, amongst studies that do report evaluations, there is a lack data for important metrics such as false alarm rates as well as other details that would support reproducibility. The first device evaluation study contributes an assessment of the ‘photoplethysmography’ optical heart rate performance of the medical-grade Empatica E4 data streaming wrist-worn wearable that is based on the Empatica Embrace epilepsy monitor. Heart rates were acquired from the E4 and a reference electrocardiogram (ECG) chest strap monitor for four participants during treadmill walking and 12 hours of free living. Mean Absolute Percentage Errors (MAPEs) and correlations are reported and demonstrate variable performance that includes negative correlation with the reference. This finding contributes insights into the poor seizure detection performance of studies that have relied on wrist-worn heart rate sensing during motor seizures.

The second study reports device evaluation results for the Empatica Embrace wrist-worn seizure monitor. No other studies in the literature have evaluated the interfaces of wearable seizure monitors. Eight of the Embrace display indications were assessed for ‘guessabilty’ by fourteen computer science participants who also performed a heuristic evaluation of the interface. The guessability results demonstrate confusion between different interface indications. The heuristic evaluation identified i) concerns about accessibility and reliance on recall and ii) satisfaction in terms of the minimal aesthetic of a simple light pattern interface.

The third and final study reports opinions and experiences of wearable epilepsy monitors reported by 61 respondents comprising 36 individuals with epilepsy, 14 carers, and 11 healthcare professionals. Overall, survey responses indicate that stakeholders have mixed opinions of wearable epilepsy seizure monitors and a degree of concern, particularly in terms of false alarms, missed seizures and other aspects of device reliability, as well as concerns about costs.