HRX 2024 AbstracX Library-Performance of a Novel Machine Learning Software: Automated QTc Measurement

Performance of a Novel Machine Learning Software: Automated QTc Measurement

Pdf Summary

The study titled "Performance of a Novel Machine Learning Software: Automated QTc Measurement and AAD Dosing in a Prospective Cohort" examines the effectiveness of new machine learning software designed to automate the measurement of QTc intervals and assist in antiarrhythmic drug (AAD) dosing for atrial fibrillation (AF) management. Atrial fibrillation affects 43 million people globally, with a predicted 12.1 million cases in the U.S. by 2030. Despite the high prevalence, only 4% of patients undergo ablation, highlighting a gap in AF care, especially in the streamlined titration and continued monitoring of AADs.<br /><br />The study, supported by the National Heart, Lung, and Blood Institute (NHLBI), compares the software's QTc measurement performance against electrophysiologist (EP) physicians. The software's endpoints include QTc accuracy, EP agreement with dose recommendations, and app usability. It was found that 95.6% of software-recommended doses were approved by physicians without modifications. The software's QTc measurements using 6-lead ECGs were within 18.6 ms of manual 12-lead QTc measurements, aligning closely with the 20 ms interobserver variability typically seen. Usability scored nearly perfect on validated medical software surveys, with ratings between 6.8 and 6.9 out of 7.<br /><br />The study concludes that the machine learning software is highly accurate in measuring QTc intervals from reduced-lead ECGs and in recommending AAD doses, aligning closely with EP decisions. Ongoing validation in diverse clinical settings is planned to further support its efficacy and potential widespread adoption in AF management. The research is partially financially backed by equity holders in SafeBeat Rx Inc., indicating a potential financial interest.

Keywords

machine learning

QTc measurement

antiarrhythmic drug dosing

atrial fibrillation

software performance

electrophysiologist comparison

ECG measurements

app usability

SafeBeat Rx Inc.

clinical validation