Predicting atrial fbrillation treatment outcomes through mathematical modelling, signal processing and machine learning

Catheter ablation and antiarrhythmic drug therapy approaches for treatment of atrial fibrillation are sub-optimal. This is in part because it is challenging to predict long-term response to therapy from short-term measurements, which makes it difficult to select optimal patient-specific treatment approaches. Clinical trials identify patient demographics that provide prediction of long-term response to standard treatments across populations. Patient-specific biophysical models can be used to assess novel treatment approaches but are typically applied in small cohorts to investigate the acute response to therapies. Our overall aim is to use machine learning approaches together with patient-specific biophysical simulations to predict long-term atrial fibrillation recurrence after ablation or drug therapy in large populations.

In this talk I will present our methodology for constructing personalised atrial models from patient imaging and electrical data; present results from biophysical simulations of ablation treatment; and finally explain how we are combining these methodologies with machine learning techniques for predicting long-term treatment outcomes.