Bayesian inference and machine learning methods for the advanced analysis of electrocardiographic signals

Thesis

The realisation of precision cardiology requires novel techniques for the non-invasive characterisation of individual patients’ cardiac function to inform therapeutic and diagnostic decision-making. Both electrocardiography and imaging are used for the clinical diagnosis of cardiac disease. The integration of multimodal datasets through advanced computational methods could enable the development of the cardiac ‘digital twin’. A cardiac digital twin is a comprehensive virtual tool that mechanistically reveals a patient’s heart condition from clinical data and simulates treatment outcomes. The adoption of cardiac digital twins requires the non-invasive efficient personalisation of the electrophysiological properties in cardiac models.

This thesis presents novel computational techniques to exploit the synergy between Artificial Intelligence and multimodal datasets towards efficiently producing cardiac digital twins from non-invasive routinely acquired clinical data modalities, namely, ECG and cardiac magnetic resonance (CMR) imaging. I present and evaluate

-> an efficient Bayesian-based inference method for estimating the ventricular activation properties from electrocardiographic data through CMR-based modelling and simulation,

-> and a machine learning-based ECG delineation methodology capable of generalising to new databases to enable the translation of my inference method into real-world clinical applications.

These two methodologies, combined with CMR-based cardiac anatomical reconstruction, compose the foundations for an end-to-end cardiac digital twin generation pipeline from non-invasive standard clinical data modalities.

Authors: Camps, J

• DOI: 10.5287/ora-dannwo2yj
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: Cardiac research; Electrocardiogram; Modelling and simulation; Cardiac digital twin; Conduction speeds; Machine learning; Eikonal; Deep learning; Earliest activation sites; Magnetic resonance imaging; Bayesian inference
• Subjects: Human-based cardiac research; Machine learning; Modelling and simulation