Survive a heart attack. Artificial intelligence will identify risks in heart patients | Healthy Life | Health

Scientists from Vladivostok trained artificial intelligence to predict the risks of in-hospital mortality in patients with a heart attack. The developed machine learning model will improve methods for diagnosing and treating coronary heart disease. The press service of the Ministry of Education and Science told aif.ru about this.

Reduce risk by 5 times

As scientists note, the high risk of death from heart disease depends on a combination of factors. Moreover, for each patient this is a different set – his physiological characteristics, other past illnesses, and the general condition of the body. Traditional parameter scales used by doctors show the average condition of the patient and do not always allow one to establish a personalized cause for increased risks. Artificial intelligence helps the doctor determine which risk factors are most dangerous for a particular patient and choose the correct course of further action: preparation for surgery, drug or rehabilitation therapy, etc.

Thus, coronary bypass surgery, which is performed when the vessels that supply the heart are narrowed, is in itself unsafe for the patient. The risks of mortality increase if the patient has certain clinical and functional features, for example, a recent myocardial infarction against the background of a large end-diastolic volume of the left ventricle. If the system notifies a specialist about this, the planned operation can be postponed slightly, which will significantly reduce the risk of a fatal outcome. The doctor’s task is to make an informed decision, taking into account the prognosis and explanations of the system.

“A patient’s risk of death may increase if their blood is thick and their kidneys are not functioning well. Individually, these factors are not as dangerous as in combination. Once the system indicates this, the doctor can take action in the preoperative period. This will help reduce the risk of an unfavorable outcome by 5-7 times,” says project manager, director of the Institute of Information Technologies of Vladivostok State University (VVSU), Doctor of Technical Sciences Karina Shakhgeldyan.

AI will find hidden reasons

In the process, scientists analyzed different data sets from the field of cardiac surgery. For example, one of them is data on 4677 patients who suffered myocardial infarction with ST segment elevation (damage to the heart muscle throughout the thickness – editor’s note) and were operated on using percutaneous coronary intervention.

“We are studying not only those predictors (prognostic parameters) that were previously proposed by other authors, but also looking for new ones, developing prognostic models, and determining threshold values ​​that allow us to form cause-and-effect relationships between risk factors and adverse events,” Shahgeldyan clarified.

Artificial intelligence can point out previously unexplored dependencies in data, which will allow doctors to treat patients more efficiently and better.

The researchers also note that it is important not only to train artificial intelligence, but also to provide an explanation for doctors why such a forecast was given, what the choice made by artificial intelligence was based on. Then doctors will have more confidence in such models, and they will be introduced into practice faster.

Developments are being carried out jointly with doctors from the Primorsky Regional Clinical Hospital. From the medical side, he manages the project Honored Scientist of the Russian Federation, Doctor of Medical Sciences, Professor Boris Geltser.