Explainable AI Tool Identifies High-Risk Patients Before Heart Attack Reperfusion

Intramyocardial hemorrhage is bleeding within the heart muscle that can occur after blood flow is restored during heart attack treatment. It affects about 40% of patients treated for ST-segment elevation myocardial infarction (STEMI) and is linked to higher risks of heart failure and death. Clinicians lack a reliable way to anticipate this complication before reopening an occluded artery. Researchers have developed an explainable artificial intelligence score to flag high-risk patients before reperfusion.

State University of New York Upstate Medical University (SUNY Upstate; Syracuse, NY, USA) led the development of an intrinsically explainable AI-based scoring system to estimate the likelihood of intramyocardial hemorrhage before an artery is reopened. The approach is designed for use by interventional cardiologists in cardiac catheterization laboratories. It aims to support decision-making at the point of care for patients presenting with ST-segment elevation myocardial infarction.

The risk score is intended for application during emergency angiography to provide real-time assessment before reperfusion. It can also assist in selecting patients for closer post-procedural monitoring once flow is restored. In addition, the tool may help determine who should undergo cardiac magnetic resonance imaging and who may qualify for clinical trials focused on limiting hemorrhagic injury.

The research describing the model was published in JACC: Advances. Collaborating institutions included the Indiana University School of Medicine, the University of Toledo College of Medicine and Life Sciences, Northern Ontario School of Medicine University, Cleveland Clinic, and Upstate Medical University. The team developed and tested the score to identify patients at high risk of bleeding into damaged heart muscle after a severe heart attack using explainable artificial intelligence.

"This is the kind of AI we need for accurate, interpretable and usable data at the point of care. The SNN method allows interventionalists to see exactly which factors are driving the prediction, rather than being asked to trust a black box," said Ankur Kalra, director of cardiac catheterization laboratories and chief of cardiology at Upstate Medical University. "These findings are actionable in an electronic health record-based calculator that can help identify these patients." 

Related Links
SUNY Upstate Medical University