HRX 2024 AbstracX Library-Digital Pipeline for Extraction and Processing of Implantable Cardioverter Defibrillator Electrograms for Machine Learning Analysis

Digital Pipeline for Extraction and Processing of Implantable Cardioverter Defibrillator Electrograms for Machine Learning Analysis

Back to course

Pdf Summary

The document discusses the development of a digital pipeline for extracting and processing implantable cardioverter defibrillator (ICD) electrograms for machine learning (ML) analysis. Dr. Neal Bhatia emphasizes the necessity of leveraging ICD data due to its high volume and real-time recordings from diverse populations. ICDs provide valuable diagnostics, such as heart sounds, heart rate, respiration rate, activity, and ventricular pacing. However, challenges arise from the proprietary nature and siloed storage of this data.<br /><br />The paper outlines a study involving 530 patients equipped with subcutaneous ICDs (S-ICDs), documenting a total of 8,545 recorded rhythm episodes. Among these, 401 episodes were related to ventricular arrhythmias, with 243 confirmed as true ventricular events. Additionally, the data included episodes of atrial fibrillation (AF) and specific challenges like T wave oversensing and supraventricular tachycardia (SVT).<br /><br />The creation of a digital pipeline aims to facilitate the use of intracardiac electrograms, which are crucial for advancing ML research in cardiology. These electrograms have potential but are difficult to obtain due to logistical and legal barriers. Unlike electrocardiograms (ECGs), these require professional interpretation by a physician. The pipeline seeks to overcome these hurdles, accelerating ML research by providing high-quality, annotated data.<br /><br />Overall, Bhatia's work underscores the potential of a digital infrastructure to harness the vast data generated by ICDs, driving advancements in machine learning applications for better cardiac care.

Keywords

digital pipeline

implantable cardioverter defibrillator

ICD data

machine learning

electrograms

ventricular arrhythmias

atrial fibrillation

cardiology research

data annotation

cardiac care