Back to Projects

AI-enabled Imaging, Emerging Imaging

Application of deep learning to predict optimal ablation therapy for atrial fibrillation from imaging data

Project ID: 2019_044

1st supervisor: Oleg Aslanidi, King’s College London
2nd supervisor: Alistair Young, King’s College London

Atrial fibrillation (AF) the most common sustained cardiac arrhythmia that affects about 33 million people worldwide. The disease is associated with increased levels of morbidity and mortality, high risks of developing heart failure and stroke, and therefore very high rates of patient hospitalizations. The overall economic burden of AF amounts to 1% of total healthcare costs in the UK alone. Even advanced first-line therapies, such as catheter ablation (CA), are highly empirical and have poor long-term outcomes, with about half of AF patients returning for the repeated procedures, which further contributes to the healthcare burden. This warrant the development of novel approaches that can improve the efficacy of CA therapy and clinical outcomes in a large patient population.

The mechanisms underlying AF are still unclear, but it has been recognized that ectopic electrical beats from the pulmonary veins (PV) can trigger AF, and that re-entrant drivers (RDs) generated by breakdown of such ectopic waves can provide self-sustained drivers for AF. In addition, areas of fibrotic atrial tissue have been linked with slow conduction of electrical waves, providing anchoring points for the RDs, and thus arrhythmogenic locations in the atria. CA therapy involves controlled destruction of arrhythmogenic locations via delivery of localised energy to atrial tissue through a catheter. The only clinically proven empirical strategy is the pulmonary vein isolation (PVI), which generates circumferential lesions around the PVs. However, even PVI has low success rate in patient with chronic forms of AF. Promising novel strategies include RD- and fibrosis- driven CA. The former targets the core of RDs, while the latter targets fibrosis by applying box isolation of fibrotic areas (BIFA) or linear lesions across fibrotic tissue. The heterogeneous results obtained by different clinical centres suggest that a single ablation strategy is unlikely to be successful for all patients, and the improvement of CA therapy can come from personalised approaches to each patient.

Thus, image-guided CA procedures are increasingly used to move away from empirical therapy and improve the patient outcomes. However, even advanced imaging systems do not provide crucial functional information about the origins of arrhythmogenesis, and the success of image-based patient stratification and CA guidance remains suboptimal. Image-based computational modelling can provide such information by predictive simulations of the 3D atrial function in a given patient, particularly by linking atrial structural featured obtained from imaging with AF arrhythmogenesis. Downsides of this approach include (i) huge computational power needed to simulate multiple AF scenarios in detailed 3D atrial models and (ii) the need to rerun the models each time novel data is integrated into them, which both make the application of models in a clinical setting impractical.

In this project, we will create image-based models of 3D atria for a cohort of 80 AF patients, simulate multiple scenarios of AF and customised CA strategies for each patient-specific model, utilise the model simulation outputs to label the images and then use the latter to train deep neural networks. Once the network is trained, it will provide a fast and flexible tool to identify optimal CA therapy for patients outside of the training cohort based on image only, without the need to run simulations.

Back to Projects