Automated digitization of paper ECG records using convolutional networks: a Faster R-CNN and U-Net approach

Authors

Haoliang Shang, Clemens Hutter, and Yani Zhang

Reference

CinC 2024, Oct. 2024, to appear.

Abstract

As part of the George B. Moody PhysioNet Challenge 2024, we developed a deep learning model based on detection and segmentation to recover electrocardiogram (ECG) time series from ECG record printouts. Our team, mins-eth, designed a hybrid pipeline of convolutional neural networks (CNNs) that leverages the strength of Faster Region-based Convolutional Neural Network (Faster R-CNN) for precise detection of the signals and that of U-Net for pixel-level accurate segmentation. Our model can handle a variety of distortions present in scanned ECG records, including rotation, cropping, creases, as well as text artifacts, and efficiently identifies and extracts ECG waveforms. For the digitization task, our model received an SNR of 0.893 (ranked 6/16) on the hidden test set.

Download this document:

This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author's copyright. In most cases, these works may not be reposted without the explicit permission of the copyright holder.