(ORDO NEWS) — A new AI-based approach can predict if and when a patient will die from cardiac arrest. Technology based on raw images of diseased hearts and patient biographies far exceeds physicians’ predictions and could revolutionize clinical decision making and improve survival for sudden and fatal cardiac arrhythmias, one of the most deadly and mysterious conditions in medicine.
The work, led by researchers at Johns Hopkins University, is detailed today in the journal Nature Cardiovascular Research.
“Sudden cardiac death caused by arrhythmia accounts for up to 20% of all deaths worldwide, and we know little about why it happens or how to determine who is at risk,” said study senior author Natalya Trayanova, professor of biomedical engineering and medicine.
“There are patients at low risk of sudden cardiac death who are receiving defibrillators that they may not need, and there are patients at high risk who are not receiving the necessary treatment and may die in their prime.
Our algorithm can determine who is in risk group for cardiac death and when it will occur, which will allow doctors to decide what exactly needs to be done.
The team is the first to use neural networks to build a personalized survival score for each patient with heart disease. These risk indicators provide a highly accurate estimate of the likelihood of sudden cardiac death within 10 years, as well as the time when this is most likely to occur.
The deep learning technology is called the Survival Study of Cardiac Arrhythmia Risk, or SSCAR. This name alludes to the scarring of the heart caused by heart disease, which often leads to fatal arrhythmias, and is the key to the algorithm’s predictions.
The team used contrast-enhanced images of the heart, visualizing the distribution of scars, taken from hundreds of real Johns Hopkins hospital patients with cardiac scarring, to train an algorithm to identify patterns and relationships not visible to the naked eye
The current analysis of clinical cardiac images reveals only simple scar characteristics such as volume and mass, which grossly underutilizes what this paper has shown to be critical data.
“The images convey important information that clinicians don’t have access to,” says first author Dan Popescu, a former graduate student at Johns Hopkins University. “These scars can be distributed in different ways, and this speaks to the patient’s chances of survival. There is hidden information in this.”
The team trained a second neural network based on 10 years of standard patient clinical data, 22 factors such as age, weight, race, and prescription drug use.
Not only were the predictions of the algorithms significantly more accurate across the board than physicians, they were validated in tests with an independent cohort of patients from 60 medical centers across the US, with different case histories and different imaging data, suggesting that the platform could be taken anywhere.
“This could have a significant impact on clinical decision making regarding arrhythmia risk and represents an important step towards bringing patient trajectory prediction into the age of artificial intelligence,” said Trayanova, co-director of the Cardiovascular Diagnosis and Treatment Innovation Alliance.
“This represents the trend of merging artificial intelligence, engineering and medicine as the future of healthcare.”
The team is now working on creating algorithms to detect other heart diseases. According to Trayanova, the concept of deep learning can be developed for other areas of medicine that rely on visual diagnostics.
—
Online:
Contact us: [email protected]
Our Standards, Terms of Use: Standard Terms And Conditions.