Related Experiment Videos
Machine learning for diagnosing non-ST-segment elevation myocardial infarction: a derivation and validation study
Arnaud Champetier1, Pedro Lopez-Ayala1, Christoph Reich2
1Cardiovascular Research Institute Basel (CRIB) and Department of Cardiology, University Hospital Basel, University of Basel, Basel, Switzerland.
Eclinicalmedicine
|July 20, 2026
Summary
New open-access machine learning models for non-ST-segment elevation myocardial infarction (NSTEMI) show excellent accuracy and improve patient triage compared to current guidelines. These FAST-NSTEMI models offer accessible tools for faster diagnosis and management.
Area of Science:
- Cardiology
- Artificial Intelligence in Medicine
- Diagnostic Accuracy
Background:
- Existing machine learning (ML) decision support tools for suspected non-ST-segment elevation myocardial infarction (NSTEMI) are proprietary, hindering accessibility and clinical use.
- Open-access ML models are needed to improve the diagnosis and management of NSTEMI.
Purpose of the Study:
- To develop and validate open-access ML-based models for the diagnosis of NSTEMI.
- To compare the diagnostic performance and triage efficacy of these models against the European Society of Cardiology (ESC) hs-cTn-0/1 h-algorithm.
Main Methods:
- Two open-access ML models (single high-sensitivity cardiac troponin [hs-cTn] and serial-hs-cTn) were derived and internally validated using data from 8763 patients across multiple international sites.
- External validation was performed on 4882 patients from a large European study.
- Model performance was assessed using discrimination (area under the receiver-operating-characteristic curve) and calibration, and compared to the ESC 0/1 h-algorithm for safety and triage efficacy.
Main Results:
- Both single-hs-cTn and serial-hs-cTn models demonstrated excellent discrimination in internal and external validation (AUCs ranging from 0.91 to 0.96).
- The ML models showed good calibration and comparable safety metrics to the ESC algorithm.
- FAST-NSTEMI models significantly improved triage efficacy, enabling more efficient patient rule-out or rule-in compared to the ESC 0/1 h-algorithm.
Conclusions:
- The open-access FAST-NSTEMI ML models provide excellent diagnostic performance, good calibration, and high safety.
- These models offer improved triage efficacy over the current ESC 0/1 h-algorithm, representing a valuable advancement in NSTEMI diagnosis.
- Further validation and prospective implementation are recommended to confirm generalizability and clinical utility.