THE ESSENTIAL
- Arrhythmia is a disturbance of the normal heart rhythm, in its frequency and in its regularity.
- 7 times out of 10, cardiac arrest occurs in front of a witness, but less than 20% of these people perform first aid.
“Up to 20% of deaths worldwide are due to sudden cardiac arrest linked to arrhythmia, yet we do not understand why this happens or how to determine the patients most at risk”, explains Natalia Trayanova, one of the authors of a study that has just been published in the journal Nature Cardiovascular Research. In this work, the researchers precisely wanted to remove these unknowns… To do this, they developed an artificial intelligence algorithm capable of predicting much more precisely than a doctor if and when a patient might die of cardiac arrest.
50,000 premature deaths from heart attack
And the stakes are high in terms of public health because in France, 50,000 people die prematurely from cardiac arrest each year, according to the French Federation of Cardiology. This can be defined as the destruction of a more or less extensive area of the heart muscle, caused by the obstruction of an artery of the heart which prevents the blood from circulating. Generally, it manifests itself in the following way: the person loses consciousness, falls, does not breathe and does not react when stimulated.
Helping physicians with decision making
“Some patients at low risk of fatal cardiac arrest are given defibrillators they may not need, while some are at high risk but don’t get the treatment they need and might die in the prime of life. age, continues Natalia Trayanova. The algorithm will be able to determine which patient is at risk of cardiac arrest and when this will happen, allowing doctors to decide exactly what needs to be done”.
Analyze heart scars
To develop this new tool, the research team used hundreds of images of heart scars from real patients. These were usually due to heart disease with a very high risk of mortality. Thanks to these pictures, the algorithm was trained to recognize which scars – characteristics, size, positioning, etc. – were the most at risk. Until now, analysis of clinical cardiac images only considered simple scar characteristics, such as volume and mass.
Better predictions than doctors
“Images contain very important information that doctors do not have access tosays Dan Popescu, one of the authors. Scarring can be distributed in different ways and it says a lot regarding a patient’s chances of survival”. At the same time, the scientists also trained their artificial intelligence to take into account 22 other risk factors such as age, weight, medication taken, etc. Results: Algorithms’ predictions of a patient’s risk of cardiac arrest were more accurate than physicians’ predictions. The artificial intelligence tool was indeed able to specify when the cardiac arrest would occur!
Ultimately, if this new artificial intelligence tool validates all phases of clinical trials, it might therefore help doctors in decision-making for patients with heart problems… And thus help prevent many fatal cardiac arrests .
