Abstract
Introduction: We describe the creation of a ensemble deep neural network architecture to classify cardiac abnormality from 12 lead ECGs. The model was created by the team between a ROC and a heart place for the PhysioNet/Computing in Cardiology Challenge 2020.
Methods: ECGs were downsampled to 257 Hz and then set to a consistent duration by randomly clipping or zeropadding the signal to 4096 samples. To learn effective features, we created a modified ResNet with larger kernel sizes that models long-term dependencies. We embedded
a Squeeze-And-Excitation layer into the modified ResNet to learn the importance of each lead, adaptively. A simple constrained grid-search method was applied to deal with class imbalance.
Results: Using the bespoke weighted accuracy metric, We achieved a 5-fold cross-validation score of 0.684, sensitivity and specificity of 0.758 and 0.969, respectively. The corresponding result for the hidden test set was 0.672.
Conclusion: The proposed prediction model performed well on the validation and hidden test data. Such models
may be potentially used for ECG screening or diagnosis.
Methods: ECGs were downsampled to 257 Hz and then set to a consistent duration by randomly clipping or zeropadding the signal to 4096 samples. To learn effective features, we created a modified ResNet with larger kernel sizes that models long-term dependencies. We embedded
a Squeeze-And-Excitation layer into the modified ResNet to learn the importance of each lead, adaptively. A simple constrained grid-search method was applied to deal with class imbalance.
Results: Using the bespoke weighted accuracy metric, We achieved a 5-fold cross-validation score of 0.684, sensitivity and specificity of 0.758 and 0.969, respectively. The corresponding result for the hidden test set was 0.672.
Conclusion: The proposed prediction model performed well on the validation and hidden test data. Such models
may be potentially used for ECG screening or diagnosis.
Original language | English |
---|---|
Title of host publication | Computing in Cardiology |
Publication status | Accepted/In press - 1 Sept 2020 |
Publication series
Name | Computing in Cardiology |
---|---|
Publisher | IEEE Computer Society |
ISSN (Print) | 2325-8861 |