Epilepsy is characterised by unpredictable seizures that significantly impact the quality of life. Although today’s wearable technology is able to continuously monitor physiological data such as heart rate, respiratory rate, motion, and sleep, these data alone often lack the contextual information required to understand triggers related to an individual’s epilepsy. There is an opportunity to record subjective experiences and accurately align these events with their corresponding physiological data to analyse epileptic triggers and generate preventive warnings related to an individual’s historical physiological conditions.

This graduation project addresses this opportunity by designing and evaluating a journal application as a method to integrate physiological data from a wearable with subjective contextual information through a mobile application and Apple Watch. The goal is to create a time-synchronized data set that can be used to train personalised machine learning models capable of predicting seizures and generating preventive warnings.

The proposed design enables epilepsy patients to remotely log scenarios in real time while a wearable continuously measures physiological data types. These two data streams are synchronized through precise timestamps. Once the data is labeled, it is suitable for future machine learning purposes.

The research study of this project focuses on the nudging strategy of the developed mobile application. A usability study was performed with 7 participants to evaluate which type of nudging strategy yields the best user engagement and journaling compliance through the mobile application.
In a within-subjects study (N=7, three conditions for a total of 9 days), no significant differences were found between nudging strategies for the total amount of logged labels, total time covered with labels, or latency. However, a small positive trend has been observed suggesting that a personalised approach yields the best journaling behaviour, indicating that in future research, with a larger sample size, the significant difference between nudge strategies could be proven.

In conclusion, this thesis describes the iterative development of a mobile application able to align subjective contextual information to continuously measured physiological wearable data. Classification of the physiological data with subjective contextual information enables future personalised machine learning applications that support the identification, detection, and forecasting of epileptic-related triggers and seizures. The design balances an epileptic safe interface, user engagement, data processing, and interpretable data visualisation.