What is the clinical evidence from this study?

Study design: Observational. Population: Hypokalemia and Hyperkalemia (n=40180). Intervention: ECG12Net vs. Practicing clinicians. Primary outcome: Balance accuracy for hypokalemia detection (95% CI 75.7-84.9).

March 5, 2020Open Access

A Deep-Learning Algorithm (ECG12Net) for Detecting Hypokalemia and Hyperkalemia by Electrocardiography: Algorithm Development

Key Result

ECG12Net, a deep-learning algorithm, detected hypokalemia and hyperkalemia with areas under the curve of 0.926 and 0.958, respectively, significantly outperforming practicing clinicians.

Study Design

Type

Observational (n=40,180)

Multicenter

Structured PICO

Does ECG12Net improve the detection of hypokalemia and hyperkalemia compared to clinicians in patients admitted to the emergency department?

Population

40,180 patients admitted to the emergency department with 66,321 ECG records and corresponding serum potassium (K+) concentrations within 1 hour.

Intervention

ECG12Net, an 82-layer convolutional neural network deep-learning model based on 12-lead ECG.

Comparator

Six clinicians (three emergency physicians and three cardiologists) in a human-machine competition.

Outcome

Area under the curve (AUC), sensitivity, and specificity for detecting hypokalemia and hyperkalemia.

A deep-learning algorithm (ECG12Net) can accurately detect severe hypokalemia and hyperkalemia from standard 12-lead ECGs, significantly outperforming human clinicians.

Main Result

Absolute Event Rate: 80.4% vs 66.7%

Limitations

Patients were enrolled from only one academic medical center
Only six clinicians participated in the human-machine competition
Selection bias due to enrolling only emergency department patients with both an ECG and a serum K+ test
The reason why particular ECG segments were highlighted by the model remains unclear (black box nature of AI)
Decreased sensitivity in detecting mild-to-moderate hypokalemia

Abstract

Background The detection of dyskalemias—hypokalemia and hyperkalemia—currently depends on laboratory tests. Since cardiac tissue is very sensitive to dyskalemia, electrocardiography (ECG) may be able to uncover clinically important dyskalemias before laboratory results. Objective Our study aimed to develop a deep-learning model, ECG12Net, to detect dyskalemias based on ECG presentations and to evaluate the logic and performance of this model. Methods Spanning from May 2011 to December 2016, 66,321 ECG records with corresponding serum potassium (K+) concentrations were obtained from 40,180 patients admitted to the emergency department. ECG12Net is an 82-layer convolutional neural network that estimates serum K+ concentration. Six clinicians—three emergency physicians and three cardiologists—participated in human-machine competition. Sensitivity, specificity, and balance accuracy were used to evaluate the performance of ECG12Net with that of these physicians. Results In a human-machine competition including 300 ECGs of different serum K+ concentrations, the area under the curve for detecting hypokalemia and hyperkalemia with ECG12Net was 0.926 and 0.958, respectively, which was significantly better than that of our best clinicians. Moreover, in detecting hypokalemia and hyperkalemia, the sensitivities were 96.7% and 83.3%, respectively, and the specificities were 93.3% and 97.8%, respectively. In a test set including 13,222 ECGs, ECG12Net had a similar performance in terms of sensitivity for severe hypokalemia (95.6%) and severe hyperkalemia (84.5%), with a mean absolute error of 0.531. The specificities for detecting hypokalemia and hyperkalemia were 81.6% and 96.0%, respectively. Conclusions A deep-learning model based on a 12-lead ECG may help physicians promptly recognize severe dyskalemias and thereby potentially reduce cardiac events.

Ask AI

Helpful

Bookmark

View Full Paper