A new AI model proved to be better than medical professionals at identifying patients at high risk of cardiac arrest, according to a federally funded study done at Johns Hopkins University published in Nature Cardiovascular Research.
The AI system was able to analyze underused heart imaging and medical records to reveal previously unseen information on a patient's heart health.
The study focused on hypertrophic cardiomyopathy, a common inherited heart disease that affects one in every 200 to 500 individuals worldwide and is a leading cause of sudden cardiac death in young people and athletes.
According to data from the study, many patients with hypertrophic cardiomyopathy could live everyday lives; however, a percentage are at significantly increased risk for sudden cardiac death and it is difficult for doctors to determine who those patients are.
Study authors noted that people with hypertrophic cardiomyopathy develop fibrosis or scarring across their heart, and it is the scarring that increases the risk of sudden cardiac death.
Although doctors haven't been able to make sense of the raw MRI images, data from the study concluded that the AI model zeroed in on critical scarring patterns.
The study's Multimodal AI for ventricular Arrhythmia Risk Stratification (MAARS) outperformed current clinical guidelines.
The researchers tested the model against real patients who were treated according to traditional clinical guidelines at Johns Hopkins Hospital and the Sanger Heart & Vascular Institute in North Carolina.
The researchers found that, compared to the clinical guidelines, which were accurate about half the time, the AI model was 89% accurate across all patients and 93% correct for individuals aged 40 to 60 years, the population among hypertrophic cardiomyopathy patients who are most likely to experience sudden cardiac death.
Additionally, researchers found that the AI model can explain why patients are at high risk, allowing doctors to design a personalized medical plan tailored to their specific needs.
Researchers noted that the data could save lives and spare people unnecessary medical interventions, such as the need for defibrillators.
The team plans to conduct further testing of the new model on additional patients and expand the latest algorithm to be used with other types of heart diseases, such as cardiac sarcoidosis and arrhythmogenic right ventricular cardiomyopathy.
According to the researchers, the study had several limitations, including that "MAARS was developed on a single tertiary-care center cohort and was potentially exposed to institutional or referral bias."
Additionally, despite being robust in a proof-of-concept study, the researchers concluded that "cohort sizes remain smaller than those of established risk prediction models being used clinically."
The study's authors noted that sudden cardiac death from arrhythmia is rare and that they had very few actual cases to analyze, just 19 in one group and 25 in another, a small number considering the complexity of the prediction model with 67 different medical factors and heart images.
THE LARGER TREND
Other institutions involved in predicting cardiac incidents is the University of Western Australia, which earlier this month announced it developed a new AI algorithm found to better predict the risk of cardiac events compared to existing methods.
The university's coronary artery calcification dispersion and density (CAC-DAD) score measures the burden of coronary calcification and the distance of each lesion from the coronary artery origin.
It can also reclassify the risk category of highly dense plaques that are otherwise considered stable and thus low risk.
The AI was developed via a collaboration between UWA, ASX-listed Artrya, South Metropolitan Health Service, Harry Perkins Institute of Medical Research, Victor Chang Cardiac Research Institute, the University of Ottawa Heart Institute in Canada and the Vision Group.
In 2021, the European Union approved an AI technology that can identify people at risk of a fatal heart attack years before it strikes.
The CaRi-Heart technology, developed by British Heart Foundation (BHF) spinout company Caristo Diagnostics, utilizes coronary computed tomography angiography (CCTA) scans already performed in clinical practice.
It uses AI and deep learning technology to produce a fat attenuation index score (FAI-Score), which accurately measures inflammation of blood vessels in and around the heart.
A BHF-funded study involving around 4,000 patients found those with an abnormal FAI were up to nine times more likely to die of a heart attack in the next nine years than those with normal FAI readings.
It also found that around one-third of patients initially considered low risk after a routine CCTA had a much higher risk after CaRi-Heart was applied to their scan.
