Seminar - Julio Vega - 11am Thursday 27th September 2018 0.3 MVB
27th Sep 2018 11:00am
Title: Personalised Monitoring of Parkinson’s Disease Using Smartphones and Human Behaviour
Abstract: We explore a novel longitudinal, in-the-wild, combinatorial, unobtrusive, and personalised methodology to monitor Parkinson’s symptom fluctuations using smartphone data, clinical knowledge, and human behavioural inferences.
Parkinson’s clinical assessment is inaccurate because symptoms vary weekly or even daily, but patients are examined in short six-monthly sessions. Previous technology-based research has relied on specialist sensors that are more precise than traditional methods. Nevertheless, most efforts were tested over short periods, under controlled conditions, were uncomfortable to use, focused on motor symptoms, or disrupted people’s routines with periodic assessment tasks.
We hypothesise we can measure Parkinson’s symptoms using digital biomarkers as a proxy. A digital biomarker is a metric representing human activity inferred from smartphone sensors (location, interaction, etc.). Thus, we monitored ten people with mild Parkinson’s, 24/7, for one year collecting data from up to 22 smartphone sensors to extract 33 digital biomarkers. To validate our algorithms, we assessed participants every six weeks with motor and non-motor clinical tests besides daily paper diaries for their top three symptoms. This is the longest and richest dataset to date for in-the-wild, tech-based Parkinson’s monitoring.
We used Indicator Waves, Anomaly Detection, and Personal Predictions to analyse digital biomarkers 7 and 30 days before each clinical assessment to uncover patterns relevant to Parkinson’s monitoring. Preliminary analysis suggests that the maximum travelled distance from home is associated with daily self-reported fatigue and that we can use this digital biomarker to predict fatigue unobtrusively.
We introduce digital biomarkers that researchers can leverage to further explore Parkinson’s monitoring. Our methodology holds promise for providing clinicians and patients with a more accurate and fine-grained picture of Parkinson’s symptoms which would support health consultations, allow personalised care, and improve the efficiency of health services.
< Back to all events