DeRC_ECG: Combined UNet-ResNet Framework for Automated Denoising and Classification of Noisy Paper-Based ECGs

Weijie Sun and Siqi Cao and Sunil Kalmady and Md Saiful Islam and Tayyib Ul Hassan and Abram Hindle and Russ Greiner and Padma Kaul

2025/08/28

DeRC_ECG: Combined UNet-ResNet Framework for Automated Denoising and Classification of Noisy Paper-Based ECGs

Authors

Weijie Sun and Siqi Cao and Sunil Kalmady and Md Saiful Islam and Tayyib Ul Hassan and Abram Hindle and Russ Greiner and Padma Kaul

Venue

Abstract

As part of the George B. Moody PhysioNet Challenge 2024, we developed a computational approach, leverag- ing a combination of UNet and ResNet models, to analyze and classify electrocardiogram (ECG) images. Our team, DeRC ECG, introduced an innovative method that applies UNet for denoising ECG images, followed by the use of ResNet18 Networks and other CNN-based models to ac- curately classify ECGs into various diagnostic categories. For the classification task, our approach achieved a macro F-measure of 0.359 on the official test set.

Bibtex

@inproceedings{sun2025CINC-dercecg,
 abstract = {As part of the George B. Moody PhysioNet Challenge 2024, we developed a computational approach, leverag- ing a combination of UNet and ResNet models, to analyze and classify electrocardiogram (ECG) images. Our team, DeRC ECG, introduced an innovative method that applies UNet for denoising ECG images, followed by the use of ResNet18 Networks and other CNN-based models to ac- curately classify ECGs into various diagnostic categories. For the classification task, our approach achieved a macro F-measure of 0.359 on the official test set.},
 accepted = {2025-08-28},
 author = {Weijie Sun and Siqi Cao and Sunil Kalmady and Md Saiful Islam and Tayyib Ul Hassan and Abram Hindle and Russ Greiner and Padma Kaul},
 authors = {Weijie Sun and Siqi Cao and Sunil Kalmady and Md Saiful Islam and Tayyib Ul Hassan and Abram Hindle and Russ Greiner and Padma Kaul},
 booktitle = {51st International Computing in Cardiology Conference},
 code = {sun2025CINC-dercecg},
 date = {2025-08-28},
 funding = {NSERC Discovery},
 location = {Edmonton, Canada},
 pagerange = {1--4},
 pages = {1--4},
 rate = {Unknown},
 role = {Co-Author},
 title = {DeRC_ECG: Combined UNet-ResNet Framework for Automated Denoising and Classification of Noisy Paper-Based ECGs},
 type = {inproceedings},
 url = {http://softwareprocess.ca/pubs/sun2025CINC-dercecg.pdf},
 venue = {51st International Computing in Cardiology Conference},
 year = {2025}
}