Multimodal, Multicentre Registry of Clinical and Imaging Data to Develop Predictive Models Based on Artificial Intelligence to Support the Diagnostic and Therapeutic Process for Patients with Atrial Fibrillation Undergoing Catheter Ablation and Cardioversion.

1. Introduction

Atrial fibrillation (AF) is the most common cardiac arrythmia, that affects 37% Europeans over 55 years old, representing a major burden for the society and health systems. One of the most devastating consequences of AF is ischemic stroke, often caused by embolization of cerebral arteries with a left atrial appendage thrombus (LAT), which can form in left atrial appendage in the presence of AF. As it has little contractility on its own, emptying and filling of left atrial appendage relies on the function of the left ventricle and the left atrium.

AF as well as atrial flutter promote the formation of LAT and are therefore associated with an increased risk of thromboembolic events. The mechanism of stroke is complex, but embolization with LAT accounts for the majority of events and it can be triggered by sinus rhythm restoration through procedures commonly performed in patients with AF - cardioversion or catheter ablation. Therefore, cardioversion and catheter ablation in the presence of LAT are contraindicated.

Transoesophageal echocardiography (TEE) is considered the modality of choice to detect LAT with high sensitivity and specificity and is currently recommended before cardioversion or catheter ablation as an alternative to a 3-week course of oral anticoagulation. Current guidelines, however, leave room for an individual decision to either perform or not perform TEE before catheter ablation or cardioversion in a chronically anticoagulated patient but suggest no tools to assess the risk of LAT. To address this unmet clinical need, we have previously shown that by combining clinical data and measurements from transthoracic echocardiography in a machine-learning model it is possible to predict the risk of LAT formation. Importantly, our results indicated, that TTE-derived measurements played the greatest role in predicting LAT.

Despite significant advancements in interventional technologies, the selection of appropriate therapeutic methods and perioperative procedures remains a significant challenge. Patients eligible for interventional procedures related to AF often undergo numerous diagnostic tests, including computed tomography of the heart, transoesophageal and transthoracic echocardiography, electrocardiography (EKG), and electroanatomic mapping during the ablation procedure. Non-invasive multimodal imaging plays central role in the peri-ablation period, informing important therapeutic decisions and facilitating the procedure. The essence of this project is the prospective collection of high-quality imaging data from these multiple modalities and clinical data, which will allow for a better understanding of the relationships between anatomy, hemodynamic function, and clinical observation in patients with AF in the future. Ultimately the data might serve for the development of tools which could efficiently predict the risk of LAT by means of less invasive and cheaper approaches than the ones which are currently used. The resulting multimodal database will also form a foundation for future studies targeted at informing therapeutic decisions in patients undergoing catheter ablation of AF. 2. Study Objectives