Non-Invasive Machine Learned Device to Personalize Arrhythmia Therapy

项目总结

Project Summary Cardiac arrhythmias are a very common cause of symptoms, days off work, hospitalization, procedures and healthcare costs. ECG monitoring devices have emerged to help management, including wearables and smart phones. However, while these ECG devices detect arrhythmias, they give limited information to inform treatment decisions between drug and invasive ablation therapy. Notably, current devices omit critical information on spatial patterns of arrhythmias and whether they arise in left or right heart that, if available, could be used to personalize management decisions for each patient. The project develops a non-invasive AI-based torso mapping device that extends any available ambulatory monitor by fully characterizing arrhythmias in terms of rate, spatial pattern and location including left or right atrium. The tool will be a wearable device that provides first-in-class arrhythmia ‘movies’ in the heart, yet is simple enough to be applied by patients at home without the need for in hospital computed tomography (CT) or magnetic resonance (MR) imaging. Computations are performed in the cloud and transmitted to caregivers, enabling them to decide whether to refer a patient directly for invasive ablation or start a medication. This approach has the potential to greatly improve clinical care. The project builds on novel torso mapping technology and machine learning methods published by the PI and Co-Is to map arrhythmias without CT or MR imaging using 57 body surface leads, smaller than existing technologies. Aim 1 will develop machine learning and vectorially-based approaches to identify arrhythmia location from the torso, and compare its accuracy to machine learning and expert analysis of traditional ECGs. Aim 2 will identify the smallest torso lead configuration and site to localize and characterize arrhythmias. This forms the basis for our planned phase II application to build a wearable patch as part of a machine-based novel ambulatory management system. This study delivers impact at multiple levels. Scientifically, we build novel vectorial and machine learning strategies to characterize simple (non-fibrillatory) arrhythmias on a non-invasive platform. Future projects will extend to other arrhythmias. Clinically, the personalization of arrhythmia therapy by a fully remote wearable device can disrupt current sequential care and resource utilization, and improve outcomes for patients in remote and under-served areas. From a business perspective, this approach can be readily monetized to healthcare organizations, physicians, strategic partners and patients. Our team is experienced in the science, clinical, regulatory and business aspects of this proposal.