Aims:
1. To explore how physical activity data of patients with neurological conditions can support clinical decision-making by the use of interpretable machine learning.
2. To assess how physical activity measures compete with or complement other patient assessment methods, namely with imaging and patient reported outcome questionnaires.
Scientific Rationale:
As both wearable technology and artificial intelligence permeates society, the intersection of both offers great potential in patient health assessments. In particular, patients with neurological diseases must often undergo assessments that are time-consuming, physically exhausting and expensive. Furthermore, patients in developing countries often do not have access to necessary equipment. Physical activity data from wearables, analysed and explained by machine learning techniques, can provide an objective, accurate, longitudinal and relatively inexpensive solution.
Current research showcases the potential for physical activity data as an assessment tool of patients with neurological diseases. Often, as disease progresses, patients show reduced physical functioning. However, this alone is insufficient for deciding how best to utilise this data and current research fails to fill this gap. Most importantly, we must focus on how best to communicate the analysis of physical activity to clinicians and patients.
Our research aims to empower clinicians to make better decisions based on predicted patient outcomes using physical activity data with a combination of other measures, including imaging and patient reported outcomes where possible. Predictions are explained using argumentative reasoning which has been shown to be an effective technique for communicating comprehensive decision outcomes, particularly in situations with conflicting information or when new information requires updating existing conclusions, as will often be the case with long-term monitoring of patients.
Project Duration:
Three Years.
Public Health Impact:
We aim to build on evidence of physical activity in healthcare assessments. We plan on creating concrete examples with techniques publicly available and published in peer-reviewed journals and at respected conferences. We will showcase how this data can be best used and communicated to both clinicians and patients.
