ECG Interpretations: Accuracy, Alternative, and Advance
In a systematic review and meta-analysis of 78 studies that assessed the accuracy of
physicians’ or medical students’ ECG interpretations in a test setting, the accuracy
varied widely, ranging from 4% to 95%.
ECG Interpretations Accuracy for Medical Students and Residents.
The median accuracy across all training levels was relatively low (54%),
and scores increased as expected with progressive training and specialization; as follow:
- Cardiologists and Cardiology Fellows: 75%
- Practicing Physicians: 69%
- Residents: 56%
- Medical Students: 42%
ECG Interpretations Accuracy for Practicing Physicians and Cardiologists.
Automatic diagnosis of the 12-lead ECG using a deep neural network.
Deep Neural Networks (DNNs) are models composed of stacked transformations that learn tasks by examples.
In this trial, the investigators trained DNN model in a dataset of more than 2 millions ECGs
The DNN model was then tested employing 827 ECG tracings.
The DNN outperform cardiology resident in recognizing
6 types of abnormalities in 12-lead ECG recordings, with F1 scores above 80%
and specificity over 99%.
Artificial Intelligence-Enabled ECG Algorithm to Identify Patients With Left Ventricular Systolic Dysfunction.
In this trial, an AI-enabled ECG algorithm applied retrospectively to
a sample of 1,606 patients evaluated in an acute care setting for dyspnea.
It effectively identifies LVSD and outperforms NT-proBNP.
Receiver operative characteristic (ROC) curve for identification of left ventricular ejection fraction (LVEF) ≤35%
The artificial intelligence-enabled ECG algorithm identified LVSD with
an AUC of 0.89 (95% CI, 0.86–0.91) and accuracy of 85.9%.
NT-proBNP alone at a cutoff value >800 identified new LVSD (EF≤35%) with
an AUC of 0.80 (95% CI, 0.76–0.84).