Core Concepts
Machine learning surpasses expert manual measurement in detecting long QT syndrome on resting ECGs.
Abstract
TOPLINE:
Machine learning model more effective than manual measurement by experts in detecting long QT syndrome on resting ECGs.
METHODOLOGY:
Developed a convolutional neural network to detect long QT syndrome on baseline ECGs.
Tested the model on patients from the Canadian National Hearts in Rhythm Organization Registry.
Study included a population-based group with milder forms of long QT syndrome.
TAKEAWAY:
Model showed high diagnostic capacity for long QT syndrome detection and genotype differentiation.
Outperformed expert-measured QTc intervals in sensitivity and accuracy.
IN PRACTICE:
Long QT syndrome associated with serious cardiac issues, but correct management can lead to excellent outcomes.
Model can detect long QT syndrome even in patients with normal or borderline QTc intervals.
SOURCE:
Study led by River Jiang, MD, University of British Columbia, published in JAMA Cardiology.
LIMITATIONS:
Control cohort for training the model was at low risk for long QT syndrome.
Model showed reduced discriminatory ability in a higher-risk patient cohort.
DISCLOSURES:
National Hearts in Rhythm Organization funded by Canadian Institutes of Health Research.
Stats
AUC for long QT syndrome detection: 0.93
AUC for distinguishing genotypes: 0.91
F1 score for model vs expert QTc intervals: 0.84 vs 0.22
Sensitivity for model vs expert QTc intervals: 0.90 vs 0.36
Quotes
"Excellent outcomes can be achieved with correct management."
"Model surpassed expert-measured QTc intervals in detecting long QT syndrome."