Smartphone AI vs. Medical Experts: A Comparative Study in Prehospital STEMI Diagnosis

Area of Science:

• Cardiology
• Medical Technology
• Artificial Intelligence

Background:

• Prehospital telecardiology aids early ST-elevation myocardial infarction (STEMI) detection but faces implementation challenges.
• Utilizing phone cameras to extract digital STEMI biomarkers from printed ECGs presents an affordable and scalable alternative.

Purpose of the Study:

• To assess the feasibility of extracting a digital STEMI biomarker from prehospital ECGs using phone cameras.
• To compare the performance of this digital biomarker against human expert consensus.

Main Methods:

• A deep learning analyzer (QCG™ analyzer) extracted a STEMI biomarker (qSTEMI) from prehospital ECGs of patients suspected of STEMI.
• The qSTEMI biomarker was compared to the consensus score of emergency physicians and interventional cardiologists.
• Non-inferiority was tested using a 0.100 margin for sensitivity and specificity.

Main Results:

• The qSTEMI biomarker achieved an area under the ROC curve of 0.815, compared to 0.736 for expert consensus (p=0.081).
• qSTEMI demonstrated sensitivity of 0.750 and specificity of 0.862, with comparable positive and negative predictive values to expert consensus.
• Non-inferiority of qSTEMI to expert consensus was confirmed for both sensitivity and specificity.

Conclusions:

• A digital STEMI biomarker derived from printed prehospital ECGs shows non-inferior performance compared to expert consensus.
• This approach holds promise for improving the scalability and affordability of prehospital telecardiology for STEMI detection.