How is AI used in medicine?
Medical research makes substantial use of artificially intelligent computer systems.
Diagnosing patients, end-to-end drug research and development, increasing physician-patient communication, transcribing medical documents such as prescriptions, and remotely treating patients are all examples of common uses.
So can artificial intelligence predict whether or not someone will have a heart attack?
A novel artificial intelligence-based method may forecast whether or not a patient will have a cardiac arrest and when they will die.
The technology, which is based on raw images of patients' diseased hearts and backgrounds, outperforms doctors' predictions and 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 which was led by Johns Hopkins University researchers was published in the journal Nature Cardiovascular Research today.
"We have no idea why it happens or how to determine who is at risk," senior author Natalia Trayanova, a biomedical engineering and medicine professor, said.
"Arrhythmia-related sudden cardiac mortality accounts for up to 20% of all fatalities worldwide," she noted. "
Some people are at little risk of sudden cardiac death who are given defibrillators they don't need, and some high-risk patients aren't getting the care they require and may die in their prime.
Our system can predict who is in danger of cardiac death and when it will happen, allowing clinicians to pick the best course of action."
The researchers are the first to use neural networks to provide a personalized survival assessment for each patient with heart disease.
These risk variables can forecast the chance of sudden cardiac death over the next 10 years, as well as when it will happen.
Survival Study of Cardiac Arrhythmia Risk, or SSCAR, is the name of the deep learning technique.
The name relates to cardiac scarring induced by heart illness, which frequently leads to deadly arrhythmias and is also the cornerstone of the algorithm's forecasts.
"What our technology can do is forecast who is at risk of cardiac death and when it will occur, allowing doctors to make the best possible decisions." Natalia Trayanova
The researchers trained an algorithm to discover patterns and associations not evident to the human eye using contrast-enhanced cardiac pictures that reveal scar distribution from hundreds of real patients with cardiac scarring at Johns Hopkins Hospital.
Current clinical cardiac imaging analysis extracts only fundamental scar features like volume and mass, substantially underutilizing what our work has demonstrated to be crucial data.
"The photos include vital information that clinicians haven't been able to access," said Dan Popescu, a former Ph.D. student at Johns Hopkins.
Scarring may be dispersed in a variety of ways, and it can provide information about a patient's chances of survival. There's knowledge in there somewhere. "
The researchers trained a second neural network on 10 years of traditional clinical patient data, which contained 22 characteristics such as the patients' age, weight, race, and prescription medication use.
The algorithms 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 US, with a variety of cardiac histories and imaging data, indicating that the platform could be used anywhere.
"This has the potential to profoundly alter clinical decision-making around arrhythmia risk," said Trayanova, co-director of the Alliance for Cardiovascular Diagnostic and Treatment Innovation. "
As for the future of healthcare, it symbolizes the trend of combining artificial intelligence, engineering, and medicine."
The researchers are also developing algorithms to detect additional heart disorders. The deep-learning idea, according to Trayanova, might be applied to other sectors of medicine that rely on visual diagnosis.