Skip to content

imi-ms/qtcnet

Repository files navigation

QTcNet

QTcNet is a deep-learning model for precise measurement of the heart-rate–corrected QT-interval (QTc) from 12-lead ECGs. The model uses a regression variant of the InceptionTime architecture:

  • Pre-Training: 120 300 algorithm-labelled ECGs
  • 100 HZ sampling rate
  • Vendor QTc bias (+15 ms) removed before training
  • Fine-Tuning: 445 cardiologist-annotated ECGs (PTB-Diagnostic Database)

Performance (expert-labelled cohorts)

Dataset PTB QTcMS ECGRDVQ Average
MAE (ms) 18.84 13.88 7.42 13.38
RMSE (ms) 29.61 24.85 11.78 22.08

Without fine-tuning, QTcNet already cuts mean absolute error from 23.4 ms to 13.4 ms and nearly halves large (> 50 ms) outliers.

Why use QTcNet?

  • State-of-the-art accuracy - ~50 % less error than typical commercial algorithms
  • Rapidly adaptable - a few hundred expert labels are enough to match local clinical standards
  • Openn Source - code & pretrained weights plus an online demo at https://qtcnet.uni-muenster.de
  • Explainable - Integrated-Gradient maps show focus on QRS onset & T-wave offset

Lead order expected by the model

I, II, III, aVR, aVF, aVL, V1, V2, V3, V4, V5, V6

Legal disclaimer

This software tool is intended solely for research and informational purposes. It is NOT a medical device and has NOT been approved or certified by any regulatory agency. The tool must NOT be used for clinical decision-making or patient management. Users are solely responsible for compliance with applicable laws and regulations. The authors and contributors accept no responsibility or liability for any damages or consequences arising from the use of this software.

Reference

If you use this repository, please cite:

@article{10.1093/europace/euaf274,
    author = {Plagwitz, Lucas and Doldi, Florian and Magerfleisch, Jannes and Zotov, Maxim and Bickmann, Lucas and Heider, Dominik and Varghese, Julian and Eckardt, Lars and Büscher, Antonius},
    title = {QTcNet: A Deep Learning Model for Direct Heart Rate Corrected QT Interval Estimation},
    journal = {EP Europace},
    pages = {euaf274},
    year = {2025},
    month = {10},
    issn = {1099-5129},
    doi = {10.1093/europace/euaf274},
    url = {https://doi.org/10.1093/europace/euaf274},
    eprint = {https://academic.oup.com/europace/advance-article-pdf/doi/10.1093/europace/euaf274/64953062/euaf274.pdf},
}

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors