AI Predicts Your Risk of Dying from Cardiac Arrest [Study]

More precisely than any doctor, a new ai – powered technique can predict whether or not a patient will have a cardiac arrest. One of medicine’s worst and most confusing illnesses, abrupt and deadly cardiac arrhythmias, may be prevented by a new system based on raw pictures of patients’ sick hearts and patient histories.

Neural networks are being used to provide a tailored survival estimate for each heart disease patient. These risk factors may accurately predict the likelihood of an unexpected cardiac death occurring in the next 10 years, as well as the time period in which it is most probable.

It’s called the Survival Analysis of Risk for Cardiac Arrhythmias (SSCAR). The term is a reference to cardiac scarring, which is a common cause of life-threatening arrhythmias. It also refers to the algorithm’s ability to make accurate forecasts.

To train an algorithm to recognize patterns and associations that are not evident to the human eye, the researchers analyzed contrast-enhanced heart pictures from hundreds of actual individuals at Johns Hopkins Hospital with heart scarring. A vital data set has been revealed in this study to be underutilized in current clinical cardiac imaging processing, which mainly extracts basic scar properties like volume and mass.

Another neural network was trained using 10 years of conventional clinical patient data, including 22 variables such as age, body mass index (BMI), racial origin, and prescription medication usage. This platform has been affirmed in assessments with a diverse patient group from 60 medical centers throughout the US, with various cardiovascular records and various imaging information, indicating that it can be used anywhere. The algorithms’ predictions were also way more accurate than doctors’ on every metric.

Other heart disorders are also being detected using algorithms developed by the researchers. Using the deep-learning idea, Trayanova believes it might be used in other sectors of medicine that depend on visual diagnosis.

The study was published in Nature Cardiovascular Research.