AliveCor QT Service

The AliveCor QT Service is a cloud-based software utilizing deep neural networks to analyze previously recorded 6-lead ECGs from adult patients with normal sinus rhythm, providing QT and QTc interval measurements for healthcare professionals. Its methodology involves sophisticated algorithms trained and validated on diverse datasets to ensure accurate cardiac interval assessment based on Bazett's and Fridericia's correction formulas.

The AliveCor QT Service is a cloud-based software as a medical device (SaMD) designed for use by qualified healthcare professionals in medical facilities. Its purpose is to measure the QT interval and heart rate corrected QT (QTc) intervals from 30-second electrocardiograms (ECGs) previously recorded from adult patients (older than 22 years) using AliveCor 6-lead ambulatory ECG devices. A prerequisite for analysis is that the ECG must have been classified as Normal Sinus Rhythm by KardiaAI. The service, accessed via a REST-based API, provides QT interval measurements and QTc values using both Bazett's and Fridericia's formulas. The technology relies on internal algorithms, including deep neural networks (DNNs), to analyze ECGs and compute the QT and RR intervals. These algorithms were developed using extensive datasets: training involved over 25,000 ECGs from more than 600 patients, and validation was performed on two distinct datasets, one with over 2,000 ECGs and another with approximately 100 ECGs, ensuring representation of diverse patient demographics. Non-clinical bench performance testing aimed to verify that the estimated QT and QTc intervals achieve a mean interval difference of ≤ 5 ms and a standard deviation of interval difference of ≤ 15 ms when compared to expert-annotated reference intervals from 12-lead ECGs. The results of this testing confirmed that the device performs according to its specifications.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. What is the specific performance of the deep neural network in identifying T-wave offset across various morphologies, especially in borderline cases or when U-waves are present?; How does the algorithm's accuracy for QT/QTc measurement vary with different heart rates within the normal sinus rhythm classification, particularly at the extremes of this range?; What are the detailed demographic breakdowns (age, sex, ethnicity, comorbidities) of the >600 patients in the training dataset and the >2000 ECGs in the primary validation dataset, and how do these reflect the intended patient population?; Beyond Bazett's and Fridericia's formulas, were other QTc correction formulas considered or tested, and what was the rationale for selecting these two specific formulas?; What are the system's error handling mechanisms and reported failure rates when encountering ECGs that, despite initial classification as Normal Sinus Rhythm, contain significant noise or artifact that could impede accurate QT measurement?