Researchers at Uppsala University in Sweden and the University of São Paulo in Brazil have created an artificial intelligence tool that analyses the 12-lead electrocardiogram to detect complete blockage of a coronary artery, the deadliest type of heart attack, and to indicate which artery is affected. The study was published in Nature Communications.
The model was trained on 540,372 electrocardiograms from 225,884 patients seen in Swedish emergency departments between 2005 and 2016, matched with angiography results. It was then validated on 401 patients at the Heart Institute (InCor-USP), whose diagnosis was confirmed by catheterisation, and on other databases.
According to the authors, the tool was correct in 96% of cases with classic ST-segment elevation and correctly identified 82% of cases in which those changes are not evident, which are the hardest to recognise.
The tool is still experimental and is meant to support doctors, not replace them. The next steps are to test it in real clinical conditions at InCor and, for hospital use, to obtain approval from Anvisa, Brazil’s health regulator.
