Artificial intelligence: cardiovascular accidents detected five years in advance

I’atherosclerosisnarrowing or obstruction of arteriesis a serious disease that can cause, among other things, a myocardial infarction. Doctors can visualize the atheroma plaques in order to assess its severity using CT angiography. However, they cannot regularly use this technique because the interpretation must be carried out by a specialist and takes up to half an hour.

Researchers from Cedars-Sinai Medical Centerin the United States, have just developed a artificial intelligence to analyze the results of a CT angiography in five to six seconds. The tool, described in the log The Lancet Digital Healthwould be able to predict the heart attack risk up to five years in advance.

A strong correlation with human experts

The researchers trained their AI to recognize the plates ofatheroma with a database containing CT angiography images of 921 patients. They then tested their tool on images from several hundred patients, and obtained a very strong correlation (r=0.922, p<0.0001) with human experts on the total quantity of plaques measured.

The researchers then established a threshold of atherosclerotic plaques to separate patients into two groups, those with a low or high risk of heart attack within five years. They validated their results on 1,611 different patients who participated in a study called Scot-Heart and were able to correctly predict the risk of heart attack. The researchers said that more studies are needed, but that with their tool and a simple CT angiography, they could predict whether a person is at risk heart attack and when she would be likely to do so.