Rehabilitation through exercise prescription for cardiac patients using an artificial intelligence-based programme
File(s)
Author(s)
Witharana, Pasan
Type
Thesis
Abstract
Cardiac rehabilitation (CR) has been demonstrated to reduce morbidity and mortality following acute coronary syndrome and cardiac surgery. However, participation rates remain low due to barriers including transport difficulties, scheduling difficulties with work commitments, and fear of infection. Additionally, hospitals may also struggle to accommodate patients within the recommended time frame when short-staffed. An artificial intelligence (AI)-enabled mobile application for home-based CR may address these challenges.
This research aimed to determine the feasibility of conducting a large-scale definitive randomised controlled trial (RCT) to assess the effectiveness of an AI-powered home-based CR programme. The primary objective was to develop and evaluate a mobile application capable of automatically prescribing personalised physical activity goals using a preliminary machine learning model prior to conducting the feasibility RCT.
I followed an evidence-based approach to develop a mobile application, web portal, and machine learning algorithm using real-life data from CR patients. The intervention underwent alpha testing with hospital staff, followed by usability testing in cardiac patients (n=15), and concluded in a feasibility RCT (n=51) employing a mixed-method approach. Predefined feasibility criteria were established for patient recruitment (≥70%), randomisation (≥80%), drop-out (≤20%) and adherence (≥75% weekly application usage). Additionally, a qualitative assessment of trial delivery was conducted.
The results demonstrated that all feasibility targets were exceeded: 85% recruitment rate, 100% randomisation rate, 86% retention rate, and 92% weekly application adherence. The mobile application achieved excellent usability scores (82.5/100) across participants with varying technological abilities. No adverse events were reported, and AI-prescribed goals required no manual adjustment by healthcare professionals. Qualitative data confirmed strong feasibility for trial delivery.
The intervention successfully automated exercise prescription while addressing key barriers to traditional CR participation, providing a foundation for future large-scale effectiveness trials and for the development of a definitive machine learning model.
This research aimed to determine the feasibility of conducting a large-scale definitive randomised controlled trial (RCT) to assess the effectiveness of an AI-powered home-based CR programme. The primary objective was to develop and evaluate a mobile application capable of automatically prescribing personalised physical activity goals using a preliminary machine learning model prior to conducting the feasibility RCT.
I followed an evidence-based approach to develop a mobile application, web portal, and machine learning algorithm using real-life data from CR patients. The intervention underwent alpha testing with hospital staff, followed by usability testing in cardiac patients (n=15), and concluded in a feasibility RCT (n=51) employing a mixed-method approach. Predefined feasibility criteria were established for patient recruitment (≥70%), randomisation (≥80%), drop-out (≤20%) and adherence (≥75% weekly application usage). Additionally, a qualitative assessment of trial delivery was conducted.
The results demonstrated that all feasibility targets were exceeded: 85% recruitment rate, 100% randomisation rate, 86% retention rate, and 92% weekly application adherence. The mobile application achieved excellent usability scores (82.5/100) across participants with varying technological abilities. No adverse events were reported, and AI-prescribed goals required no manual adjustment by healthcare professionals. Qualitative data confirmed strong feasibility for trial delivery.
The intervention successfully automated exercise prescription while addressing key barriers to traditional CR participation, providing a foundation for future large-scale effectiveness trials and for the development of a definitive machine learning model.
Version
Open Access
Date Issued
2025-06-30
Date Awarded
2026-05-01
Copyright Statement
Attribution-NonCommercial 4.0 International Licence (CC BY-NC)
License URL
Advisor
Akowuah, Enoch
Athanasiou, Thanos
Qin, Chen
Sponsor
South Tees Hospitals NHS Foundation Trust
Publisher Department
Department of Surgery & Cancer
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)