Key result
A dataset of 827 12-lead ECG tracings was compiled and annotated by multiple clinicians to serve as a test set for evaluating a deep neural network for automatic ECG diagnosis.
Why the study?
Does a Deep Neural Network accurately diagnose 6 common abnormalities on short-duration 12-lead ECGs compared to human annotators?
Population
827 ECG tracings from different patients
Comparison
Deep neural network predictions vs annotations by cardiologists, residents, and medical students
Authors
Discussion
Loading...
Member takes
Implication
Facilitates AI ECG model evaluation; leaves open clinical adoption without prospective validation.
Structured PICO
Does a Deep Neural Network accurately diagnose 6 common abnormalities on short-duration 12-lead ECGs compared to human annotators?
P
Population
827 ECG tracings from different patients annotated for 6 different ECG abnormalities by cardiologists, residents, and medical students.
I
Intervention
Deep Neural Network for automatic diagnosis of short-duration 12-lead ECG
C
Comparator
Annotations by cardiologists, residents, and medical students (gold standard derived from cardiologist consensus)
O
Outcome
Detection of 6 ECG abnormalities (1st degree AV block, right bundle branch block, left bundle branch block, sinus bradycardia, atrial fibrillation, and sinus tachycardia)surrogate
This dataset provides a gold-standard annotated set of 827 12-lead ECGs for evaluating automated diagnostic algorithms against human clinicians.
Cite This Study
Ribeiro et al. (2020) studied ECG abnormalities (n=827). Deep Neural Network vs. Gold standard annotations by cardiologists was evaluated on Detection of 6 ECG abnormalities (1dAVb, RBBB, LBBB, SB, AF, ST). A dataset of 827 12-lead ECG tracings was compiled and annotated by multiple clinicians to serve as a test set for evaluating a deep neural network for automatic ECG diagnosis.