Search
Close this search box.

Blog

Machine learning for diagnosis of myocardial infarction

08:07 09 September in Blog
0 Comments

High-sensitivity cardiac troponin assays have been included in the diagnostic approaches for patients with symptoms suggesting acute myocardial infarction 1,2. To improve diagnosis and overcome limitations associated with high-sensitivity cardiac troponin assays, a machine learning model was developed to integrate cardiac troponin concentrations at presentation or on serial testing with clinical features and provide a more individualized approach to assess probability and improve the diagnosis of myocardial infarction 3.

The diagnostic performance of guideline-recommended cardiac troponin thresholds was evaluated and analyzed based on the High-STEACS trial data 4. Then, a clinical decision support system called the Collaboration for the Diagnosis and Evaluation of Acute Coronary Syndrome (CoDE-ACS) was developed 3. This system uses machine learning models to calculate the score (0–100) that corresponds to an individual’s probability of myocardial infarction. CoDE-ACS models combine cardiac troponin as a continuous measure with age, sex, time from symptom onset, the presence of chest pain, known ischemic heart disease, hyperlipidemia, heart rate, systolic blood pressure, Killip class, myocardial ischemia on the electrocardiogram, renal function, and hemoglobin. The diagnostic performance of the CoDE-ACS system was externally validated to demonstrate how it could be used in clinical practice. The external validation cohort consisted of 10,286 patients with possible myocardial infarction pooled from seven prospective cohort studies enrolling patients across six countries 3.

A CoDE-ACS score of less than three or 61 or more corresponds to a low or high probability of myocardial infarction, respectively. Compared to standard cardiac troponin thresholds and risk scores, this system was able to identify more patients as having low probability of myocardial infarction at presentation with a similar negative predictive value and good sensitivity. On the other hand, the system identified fewer patients as having high probability of myocardial infarction at presentation but with improved positive predictive value and good specificity 3.

The suggested advantages of the CoDE-ACS clinical decision support system include 3:

  1. reducing time spent in emergency departments,
  2. preventing unnecessary hospital admission in patients unlikely to have myocardial infarction and at low risk of cardiac death,
  3. improving the recognition and treatment of those with myocardial infarction rather than myocardial injury,
  4. minimizing diagnostic performance heterogeneity,
  5. the ability to incorporate factors including different symptoms, comorbidities and other risk factors,
  6. incorporating time from symptom onset into the CoDE-ACS models which enables early presenters to be ruled out using a single cardiac troponin test,
  7. reducing the number of patients requiring further observation as incorporating information on the time of testing permits a second measurement to be incorporated at a flexible time point.

Overall, machine learning was used to develop a CoDE-ACS clinical decision support system which was trained to estimate the probability of myocardial infarction upon admission. Being a more flexible system with good predictive values, CoDE-ACS may improve clinical decision-making with advantages for both healthcare providers and patients.

References

1.         Than M, Cullen L, Reid CM, et al. A 2-h diagnostic protocol to assess patients with chest pain symptoms in the Asia-Pacific region (ASPECT): a prospective observational validation study. Lancet Lond Engl. 2011;377(9771):1077-1084. doi:10.1016/S0140-6736(11)60310-3

2.         Body R, Carley S, McDowell G, et al. Rapid exclusion of acute myocardial infarction in patients with undetectable troponin using a high-sensitivity assay. J Am Coll Cardiol. 2011;58(13):1332-1339. doi:10.1016/j.jacc.2011.06.026

3.         Doudesis D, Lee KK, Boeddinghaus J, et al. Machine learning for diagnosis of myocardial infarction using cardiac troponin concentrations. Nat Med. 2023;29(5):1201-1210. doi:10.1038/s41591-023-02325-4

4.         Shah ASV, Anand A, Strachan FE, et al. High-sensitivity troponin in the evaluation of patients with suspected acute coronary syndrome: a stepped-wedge, cluster-randomised controlled trial. Lancet Lond Engl. 2018;392(10151):919-928. doi:10.1016/S0140-6736(18)31923-8

No Comments

Post a Comment