Related Experiment Video
Updated: Jun 13, 2026
Echocardiography-guided Injection for Targeted and Reliable Intramyocardial Stem Cell Delivery in a Rat Model of Myocardial Infarction
Published on: July 25, 2025
Predicting Intramyocardial Hemorrhage Before Reperfusion in STEMI Patients With Intrinsically Explainable Artificial
Khalid Youssef1, Keyur P Vora1, Rajesh Gupta2
1Cardiovascular Imaging Research Center, Medical Imaging Research Institute, Indiana University School of Medicine, Indianapolis, Indiana, USA; Krannert Cardiovascular Research Center, Indiana University School of Medicine, Indianapolis, Indiana, USA.
Background:
Intramyocardial hemorrhage (IMH) complicates approximately 40% of reperfused ST-segment elevation myocardial infarctions (STEMIs) and is associated with worse outcomes. No method identifies patients at risk before reperfusion.
Objectives:
The objective of the study was to develop and evaluate an explainable artificial intelligence approach for pre-reperfusion IMH prediction and translate it into a practical bedside score.
Methods:
We enrolled 288 STEMI patients from the MIRON-PREDICT clinical study (NCT06423625) between June 2023 and December 2024, of whom 252 were used for model development and 36 were used for validation. Electrocardiographic, angiographic, and clinical variables were obtained during emergency coronary angiography. Cardiac magnetic resonance imaging 48 to 72 hours post-percutaneous coronary intervention served as the reference for IMH. A Superposable Neural Network identified key predictors and informed derivation of a point-based scoring model for bedside use.
Results:
The final 3-variable model comprised the presence of coronary collaterals, degree of coronary artery occlusion, and sum ST-score. Cardiac magnetic resonance imaging identified IMH in 142/288 patients. Coronary collaterals were present in 144/288 and total occlusion in 131/288. Sum ST-score was substantially higher in IMH-positive than IMH-negative patients. The resulting model achieved 84.9% accuracy, 82.3% sensitivity, and 87.3% specificity in identifying patients at risk of IMH before reperfusion. This was confirmed in the validation cohort with 83.3% accuracy, 81.3% sensitivity, and 85% specificity.
Conclusions:
Explainable artificial intelligence enabled rapid integration of clinical parameters from cardiac catheterization procedures to accurately predict IMH in STEMI before revascularization. The proposed approach could guide clinical decisions in real-time. This could inform future strategies aimed at reducing the incidence of IMH.