A new artificial-intelligence-based approach can predict if and when a patient will die from cardiac arrest far more accurately than a doctor. The technology, which is based on raw images of patients’ diseased hearts and patient histories, has the potential to revolutionize clinical decision-making and increase survival from sudden and lethal cardiac arrhythmias, one of medicine’s deadliest and most perplexing conditions.
The study, led by Johns Hopkins University researchers, was published 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 tell who’s at risk, said senior author Natalia Trayanova, the Murray B. Sachs Professor of Biomedical Engineering and Medicine.
Some patients are at low risk of sudden cardiac death and are getting defibrillators they don’t need, and then some high-risk patients aren’t getting the treatment they need and may die in their prime. Our algorithm can predict who is at risk of cardiac death and when it will occur, allowing doctors to determine exactly what needs to be done.
The researchers are the first to use neural networks to create a personalized survival assessment for each heart disease patient. These risk factors predict the likelihood of sudden cardiac death over 10 years, as well as when it is most likely to occur.
Survival Study of Cardiac Arrhythmia Risk (SSCAR) is the name of the deep learning technology. The name refers to cardiac scarring caused by heart disease, which frequently leads to lethal arrhythmias, and it is the key to the algorithm’s predictions.
The team trained an algorithm to detect patterns and relationships not visible to the naked eye by using contrast-enhanced cardiac images that visualize scar distribution from hundreds of real patients with cardiac scarring at Johns Hopkins Hospital. Current clinical cardiac image analysis extracts only simple scar features such as volume and mass, severely underutilizing what has been shown in this work to be critical data.
The images contain critical information that doctors haven’t been able to access, said first author Dan Popescu, a former doctoral student at Johns Hopkins. Scarring can be distributed in a variety of ways, and it can reveal information about a patient’s chances of survival. There’s information in there somewhere.
The researchers trained a second neural network to learn from 10 years of standard clinical patient data, which included 22 variables such as patients’ age, weight, race, and prescription drug use.
The algorithms’ predictions were not only significantly more accurate than doctors on every measure but they were also validated in tests with an independent patient cohort from 60 health centers across the United States, each with a different cardiac history and imaging data, implying that the platform could be used anywhere.
This has the potential to significantly shape clinical decision-making regarding arrhythmia risk and represents an important step toward bringing patient trajectory prognostication into the age of AI, said Trayanova, co-director of the Alliance for Cardiovascular Diagnostic and Treatment Innovation. It exemplifies the trend of fusing artificial intelligence, engineering, and medicine to shape the future of healthcare.
The team is now developing algorithms to detect other cardiac diseases. The deep-learning concept, according to Trayanova, could be applied to other fields of medicine that rely on a visual diagnosis.
Mauro Maggioni, Bloomberg Distinguished Professor of Data-Intensive Computation at Johns Hopkins, was also on the team, as were Julie Shade, Changxin Lai, Konstantino Aronis, and Katherine Wu. M. Vinayaga Moorthy and Nancy Cook of Brigham and Women’s Hospital; Daniel Lee of Northwestern University; Alan Kadish of Touro College and University System; and David Oyyang and Christine Albert of Cedar-Sinai Medical Center are among the other authors.
National Institutes of Health grants R01HL142496, R01HL126802, R01HL103812; Lowenstein Foundation, National Science Foundation Graduate Research Fellowship DGE-1746891, Simons Fellowship for 2020-2021, National Science Foundation grant IIS-1837991, Abbott Laboratories research grant supported this work. The National Heart, Lung, and Blood Institute research grant R01HL091069, St Jude Medical Inc, and St Jude Medical Foundation funded the PRE-DETERMINE study and the DETERMINE Registry.
Source link
