Crafting Deep Learning Models for Classifying ECG Paper Printouts
As part of the George B. Moody PhysioNet Challenge 2024, we (the GIRAFFE team) built an approach based on InceptionV3 to classify the electrocardiogram (ECG) images.To deal with the class imbalance, we use the Generalized Extreme Value activation function and loss weighting.For the classification task, our best model received a macro F -measure of 0.652 over the hidden test data.Because we had not submitted any unofficial phase entry, we were not included in the official rankings.
Authors
- Jacek Kawa (ORCID: https://orcid.org/0000-0002-2804-7885)
- Damian Kucharski (ORCID: https://orcid.org/0000-0002-5852-4136)
- Gregory Y.H. Lip (ORCID: https://orcid.org/0000-0002-7566-1626)
- Jakub Nalepa (ORCID: https://orcid.org/0000-0002-4026-1569)
- Agata M. Wijata (ORCID: https://orcid.org/0000-0001-6180-9979)
- Yalin Zheng (ORCID: https://orcid.org/0000-0002-7873-0922)
- Arkadiusz Paweł Czerwiński
Publication Details
- Journal
- Computing in cardiology
- Published
- 2024-12-01
- DOI
- https://doi.org/10.22489/cinc.2024.210
- Primary Topic
- Industrial Vision Systems and Defect Detection
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- Infrastruktura PL-Grid
- Academic Computer Centre Cyfronet, AGH University of Science and Technology