An Automated Online Recommender System for Stroke Risk Assessment
Keywords:Telehealth, Explainable Artificial Intelligence (XAI), Online Recommender, Stroke Risk, Smartwatch
A stroke is a complicated emergency event that leads to major neurological impairments and patient disability. It is imperative to have an automated smart recommender system that can help with early stroke prediction or detection and hence assist clinicians in stroke risk management. This study proposes the use of an automated online recommender system that can predict stroke risk levels based on the given patient-specific clinically identified stroke risk factors, such as systolic and diastolic blood pressure, age, gender, smoking habit, and cholesterol level. We integrate this model in an interactive Django-based web framework built upon software engineering best practices that can assist clinicians in monitoring key stroke risk factors dynamically in real-time using the Smartwatch device. We use machine learning (ML) techniques to predict stroke risk levels, and also employ an Explainable Artificial Intelligence (XAI) technique to rank the risk factors to provide meaningful insights for stroke risk management. We also have built a real-time patient monitoring system that can monitor patient vital signals in real-time using a Smartwatch and transmit the data to the web application where the data is concurrently processed by the ML model and can be displayed in an interactive dashboard. The results from the study show that this automated online recommender system can predict stroke risk levels with an average Area Under the Curve (AUC) of 0.98 while providing meaningful insights on the stroke risk factors that can assist clinicians in better managing stroke risk whilst being cost-effective and feasible based on popular Smartwatches and Smartphones.