Hopkins researchers develop AI technique to predict cardiac arrests

A new artificial-intelligence-based technique developed Johns Hopkins University researchers could reportedly predict if and when a patient could die of cardiac arrest. According to the university, the technology trained on raw images of diseased patients’ hearts and patient backgrounds improves on doctor’s predictions and could potentially revolutionise clinical decision making and increase the chances of survival for patients with sudden and lethal cardiac arrhythmias.
The researchers documented their findings in a paper titled “Arrhythmic sudden death survival prediction using deep learning analysis of scarring in the heart” published in Nature Cardiovascular Research.
“Sudden cardiac death caused arrhythmia accounts for as many as 20% of all deaths worldwide and we know little about why it’s happening or how to tell who’s at risk,” said Natalia Trayanova, senior author of the paper, in a university press statement. Trayanoya is a professor of biomedical engineering and medicine at the university.
“There are patients who may be at low risk of sudden cardiac death getting defibrillators that they might not need and then there are high-risk patients that aren’t getting the treatment they need and could die in the prime of their life. What our algorithm can do is determine who is at risk for cardiac death and when it will occur, allowing doctors to decide exactly what needs to be done,” explained Trayanoya.

The research team used neural networks to build a survival assessment personalised to each patient with heart disease. These assessments can reportedly predict the chance of sudden cardiac death over 10 years, and even when it’s most likely to happen.
The deep learning technology was named Survival Study of Cardiac Arrhythmia Risk (SSCAR) as an allusion to cardiac scarring caused heart disease, which often result in arrhythmias. The researchers used contrast-enhanced cardiac images from hundreds of real patients (with cardiac scarring) at Johns Hopkins Hospital to train the algorithm to detect patterns and relationships not visible to the naked eye.
According to the university, current clinical cardiac image analysis extracts only simple scar features like volume and mass. This means that they severely underutilise what the new algorithm has demonstrated to be critical data.
The researchers also trained a second neural network to learn from 10 years of standard data that included 22 factors like patient age, weight, race and prescription drug use.
The university reports that the algorithms’ predictions were significantly more accurate on every measure than doctors and that they were validated in tests with a separate and independent patient cohort from 60 health centres across the United States.
According to Trayanoya, this deep-learning concept could be developed for other fields of medicine that rely on visual diagnosis.