What does this research mean for the field?

A residual network-based deep learning algorithm accurately detects acute myocardial infarction and identifies infarct location from 12-lead electrocardiograms. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

June 28, 2021Frontiers in Cardiovascular MedicineOpen Access

Acute Myocardial Infarction Detection Using Deep Learning-Enabled Electrocardiograms

Q: What is the clinical evidence from this study?

Study design: Other. Population: Acute Myocardial Infarction (n=22042). Intervention: Residual network-based deep learning algorithm vs. Standard clinical diagnosis (ground truth labels). Primary outcome: Area under the curve (AUC) for AMI diagnosis in the testing set (AUC 0.977, 95% CI 0.961-0.991).

View Full Paper

Ask AI

Bookmark

Key result

Deep learning model detects AMI from 12-lead ECGs with an AUC of ~0.98.

AUC 0.977
95% CI 0.961-0.991
n=22,042

Why the study?

Acute myocardial infarction carries a poor prognosis, making accurate diagnosis and early intervention of the culprit lesion extremely important.

Can a deep learning residual network accurately diagnose acute myocardial infarction from 12-lead ECGs?

Design

Algorithm development and validation study

Authors

XCXiehui ChenLonggang Central Hospital WGWenqin GuoChinese Academy of Medical Sciences & Peking Union Medical College LZLingyue ZhaoNingxia Medical University

Discussion

Member takes

Implication

May aid ECG-based AMI detection in research; leaves open prospective validation before clinical use.

Structured PICO

Can a deep learning residual network accurately diagnose acute myocardial infarction from 12-lead ECGs?

Population

22,042 clinical 12-lead ECGs were used to train, validate, and test a deep learning algorithm for the automatic diagnosis of acute myocardial infarction.

Intervention

Deep learning model (residual network) for automatic diagnosis of acute myocardial infarction and location from 12-lead ECGs

Outcome

Model performance measured by area under the curve (AUC), precision, sensitivity, specificity, and F1 scoresurrogate

Main Result

Effect estimate: AUC 0.977 (95% CI 0.961-0.991)

A residual network-based deep learning algorithm can accurately diagnose acute myocardial infarction and its location from 12-lead ECGs, demonstrating high sensitivity and specificity.

Limitations

Based on open-access ECG databases lacking other clinical features like target vessel and clinical comorbidities.
Results need to be transformed into clinical applications, such as being implanted in ECG equipment.
Cannot identify patients with multivessel disease.
Algorithm is not suitable for non-ST-elevation acute coronary syndrome (NSTE-ACS) diagnosis.
Lacks a performance comparison between the deep learning model and traditional paradigms.
Despite great performance in diagnosing MI location, it still has low accuracy and sensitivity in diagnosing AMI.
AMI suspicion is mainly based on symptoms and medical history, which cannot be covered by ECG interpretation alone.

Cite This Study

Chen et al. (2021) studied Acute Myocardial Infarction (n=22,042). Residual network-based deep learning algorithm vs. Standard clinical diagnosis (ground truth labels) was evaluated on Area under the curve (AUC) for AMI diagnosis in the testing set (AUC 0.977, 95% CI 0.961-0.991). A residual network-based deep learning algorithm effectively diagnosed acute myocardial infarction from 12-lead ECGs, achieving an area under the curve of 0.977 in the independent testing set.

synapsesocial.com/papers/6aaa3a0d889ad45d867e5cdf https://doi.org/10.3389/fcvm.2021.654515

Topics

Artificial intelligence in cardiology

View Full Paper

Ask AI

Bookmark

Acute Myocardial Infarction Detection Using Deep Learning-Enabled Electrocardiograms

Discussion

Member takes

Implication

Structured PICO

Main Result

Limitations

Cite This Study

Topics

Acute Myocardial Infarction Detection Using Deep Learning-Enabled Electrocardiograms

Discussion

Member takes

Implication

Structured PICO

Main Result

Limitations

Cite This Study

Also Consider

Topics

Also Consider