Diagnostic Accuracy of the Deep Learning Model for the Detection of ST Elevation Myocardial Infarction on...

Area of Science:

• Cardiology
• Artificial Intelligence in Medicine
• Medical Diagnostics

Background:

• ST-elevation myocardial infarction (STEMI) is a critical cardiac emergency.
• Accurate and timely diagnosis of STEMI is crucial for patient outcomes.
• Electrocardiograms (ECGs) are a primary tool for STEMI diagnosis.

Purpose of the Study:

• To evaluate the diagnostic accuracy of a deep learning model (DLM) for STEMI detection using 12-lead ECGs.
• To assess DLM performance across different culprit artery locations in STEMI patients.
• To identify factors influencing DLM accuracy in STEMI diagnosis.

Main Methods:

• A deep learning model (DLM) was trained and validated on a large dataset of 60,157 ECGs (117 STEMI, 60,040 normal sinus rhythm).
• Diagnostic accuracy was measured using area under the receiver operating characteristic curve (AUROC), sensitivity (SEN), and specificity (SPE).
• Performance was analyzed based on three distinct culprit artery classifications.

Main Results:

• The DLM demonstrated high overall diagnostic accuracy for STEMI (AUROC 0.998, SEN 97.4%, SPE 99.2%).
• No significant differences in diagnostic accuracy were observed across the three culprit artery groups.
• Baseline wanders in ECG signals were identified as a factor affecting false positive interpretations (83.7%).

Conclusions:

• The DLM exhibits robust diagnostic performance for STEMI detection irrespective of the culprit artery.
• DLM accuracy for STEMI diagnosis is high, but susceptible to interference from ECG baseline wanders.
• Further refinement of DLMs may be needed to mitigate the impact of signal artifacts on interpretation.