Ventricular tachycardia (VT) is an important cause of mortality and morbidity in patients with heart diseases. The majority of life-threatening VT episodes are caused by an electrical short circuit? that travels through narrow strands of surviving tissue inside myocardial scar. Catheter ablation treats scar-related VT by ?blocking? the surviving channel that forms the circuit, commonly at the site the circuit exits from the scar. To localize a VT exit, however, remains a significant challenge. A common approach, known as pace-mapping, utilizes the principle that the VT exit serves as the origin of ventricular activation and determines the QRS morphology on 12-lead electrocardiograms (ECGs). It thus involves repetitive electrical simulation at various sites of the heart, until locating the site that reproduces the QRS of the VT on all 12 ECG leads. While the principle behind pace- mapping is time tested, the current practice is of a trial-and-error nature and requires rapid qualitative interpretation of the ECG by clinicians, which can be time-consuming and inaccurate. This research proposes to leverage modern machine learning techniques to reform the way the principle behind pace-mapping is used.
It aims to learn the relationship between the origin of ventricular activation and ECG morphology, and then use it to directly predict the exit of a VT from its ECG data. To this end, this project will include the following activities: 1) to develop a population-based model to provide pre-procedural initial localizations of VT exits using standard 12-lead ECG; 2) to integrate the population-based model with a patient-specific model in clinically-usable software to provide intra-procedural real-time guidance for localizing the exit site of a clinical VT; and 3) to assess the ability of the proposed software to improve the efficiency and accuracy of pace- mapping in a prospective clinical study. This project will be carried out by a multidisciplinary team of computational and clinical scientists with a fruitful record of collaboration. The software delivered by this project will provide real-time assistance to clinicians for narrowing down a VT exit with a minimum amount of time and localization errors. This will substantially reduce the workload for ablating multiple VTs, potentially allowing clinicians to ablate more or even all VTs seen in a procedure. This may reduce the duration of an ablation procedure while improving its outcome. The development and deployment of the software also adds minimal cost or distractions to routine workflow. With a low barrier to clinical implementation, it will have a real potential to challenge and improve the standard practice of catheter ablation.
Catheter ablation treats ventricular arrhythmia by destroying the culprit tissue responsible for the arrhythmia. To localize the culprit tissue, however, remains a tedious process with limited success in current practice. This project will develop inexpensive software to guide clinicians towards the culprit tissue in real time during the procedure, helping narrow down the ablation target with a minimum amount of time and localization errors.
