In August 2026, Indiana University School of Medicine researchers developed and tested a six-point explainable AI score to identify patients at high risk of intramyocardial hemorrhage after a severe heart attack. The score uses three measurements from electrocardiography and angiography collected during cardiac catheterization, classifying scores of 4 or higher as high risk before a blocked artery is reopened.
Indiana University School of Medicine researchers have developed and tested a six-point explainable AI score for identifying patients at high risk of intramyocardial hemorrhage (IMH) after a severe heart attack. The work, published in JACC: Advances, is designed for use before clinicians reopen a blocked artery in the cardiac catheterization laboratory, according to IU School of Medicine and News-Medical.
IMH is internal bleeding within damaged heart muscle that can occur after blood flow is restored. IU School of Medicine reports that it affects about 40% of patients treated for ST-segment elevation myocardial infarction (STEMI), and is associated with higher risks of heart failure and death.
A bedside score from catheterization data
According to the IU report, the system uses three measurements already available during cardiac catheterization, drawn from electrocardiography and angiography data. The researchers converted the underlying model into a six-point clinical score:
- • 4 or higher: high risk of IMH - • 3 or lower: low risk of IMH
The research team applied Superposable Neural Networks (SNN), a traceable AI method. IU School of Medicine describes this as the first successful medical application of SNN. Lead author Khalid Youssef said the approach provides the reasoning behind a prediction and can be calculated before revascularization without waiting for a T2* cardiac MRI, which the report identifies as the current standard method for detecting IMH.
Why interpretability matters in acute care
The reported workflow addresses a practical constraint in interventional cardiology: decisions occur during a time-sensitive procedure, while definitive imaging may not be immediately available. The score's inputs are intended to be available at the bedside during that procedure, rather than requiring clinicians to wait for cardiac MRI.
More broadly, clinical AI systems used in acute settings often face an adoption barrier when they produce risk estimates without a transparent link to clinical variables. An interpretable point score can make model outputs easier to review alongside existing clinical judgment, although the provided reporting does not describe prospective deployment, patient-outcome effects, or external validation across additional hospital systems.
For ML practitioners, the study is an example of translating a traceable model into a compact decision rule. That conversion can improve usability in high-pressure clinical environments, but broader clinical evaluation remains important before such tools are relied on in routine care.
Key Points #
- 1Indiana University converted catheterization-lab ECG and angiography inputs into a six-point score for pre-reperfusion IMH risk classification.
- 2The study applies a traceable Superposable Neural Networks method, making model inputs and risk reasoning available to clinicians.
- 3In acute-care AI, compact interpretable scores can fit clinical workflows, though multi-site validation and outcome studies remain important.
Scoring Rationale #
This is a clinically focused application of explainable AI in an acute cardiology workflow, with a concrete bedside scoring output rather than a general research prototype. Its relevance is strongest for healthcare AI and clinical decision-support practitioners; the available reporting does not establish broad deployment or outcome improvement.
Sources #
Public references used for this report. Practice with real Health & Insurance data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Health & Insurance problems