Machine Learning to Predict MI – MI3 Index

Machine Learning to Predict MI – MI3 Index


Myocardial Ischemic Injury Index (MI3 Index) is a Machine Learning Algorithm which has been developed and tested to predict the likelihood of developing acute myocardial infarction [1].

It used simple easy to collect variables: age, sex, and paired high-sensitivity cardiac troponin I concentrations. Initial troponin estimation was at presentation and second one between 1-3 hours later.

Their training set had data from 3013 patients and it was tested on data from 7998 patients with suspected myocardial infarction. A technique called gradient boosting was used to compute a value between 0 and 100 to reflect the likelihood of the person to have a diagnosis of type 1 myocardial infarction. It also estimates the sensitivity, negative predictive value, specificity and positive predictive value for the individual.

This was an international collaborative study involving 11011 patients from 9 countries.

Reference

  1. Than MP, Pickering JW, Sandoval Y, Shah ASV, Tsanas A, Apple FS, Blankenberg S, Cullen L, Mueller C, Neumann JT, Twerenbold R, Westermann D, Beshiri A, Mills NL; MI3 collaborative. Machine Learning to Predict the Likelihood of Acute Myocardial Infarction. Circulation. 2019;140:899–909.