What is the clinical evidence from this study?

Study design: Other. Population: Cardiac abnormalities (n=37749). Intervention: Adaptive Lead Weighted ResNet. Primary outcome: snormalized (weighted accuracy metric) on 5-fold cross-validation.

December 30, 2020Open Access

Adaptive lead weighted ResNet trained with different duration signals for classifying 12-lead ECGs

Key Result

The Adaptive Lead Weighted ResNet model achieved a 5-fold cross-validation weighted accuracy score of 0.684 and a hidden test score of 0.520 for classifying 24 cardiac abnormalities from 12-lead ECGs.

Structured PICO

Population

37,749 12-lead ECG recordings of varying frequency (257 Hz - 1000 Hz) and duration (6 s - 60 s) from four different sites, with a separate dataset from a fifth site withheld for testing.

Intervention

Adaptive lead weighted ResNet (modified ResNet with Squeeze-And-Excitation layer) trained with different duration signals

Outcome

Classification of 12-lead ECG recordings into 24 clinical diagnoses, evaluated by sensitivity, specificity, and a bespoke snormalized metricsurrogate

An adaptive lead-weighted ResNet model demonstrated strong performance in classifying 24 cardiac abnormalities from 12-lead ECGs, though generalizability to completely unseen datasets remains a challenge.

Limitations

Results are not easily explainable in terms of specific ECG morphology.
Upper bound on accuracy is potentially limited by noisy training data labels.
The model failed to migrate from multiple training data sets to a completely different test data set effectively.

Abstract

Introduction: We describe the creation of a 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 Phys-ioNet/Computing in Cardiology Challenge 2020.

Bookmark

View Full Paper

Cite This Study

Zhao et al. (Wed,) conducted a other in Cardiac abnormalities (n=37,749). Adaptive Lead Weighted ResNet was evaluated on snormalized (weighted accuracy metric) on 5-fold cross-validation. The Adaptive Lead Weighted ResNet model achieved a 5-fold cross-validation weighted accuracy score of 0.684 and a hidden test score of 0.520 for classifying 24 cardiac abnormalities from 12-lead ECGs.

synapsesocial.com/papers/6a160c29354578fa93b56546 https://doi.org/https://doi.org/10.22489/cinc.2020.112

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

Bookmark

View Full Paper