Explainable AI Predicts Intramyocardial Hemorrhage Prior to Reperfusion in STEMI Patients

Article Impact Level: HIGH
Data Quality: STRONG
Summary of  JACC: Advances  https://doi.org/10.1016/j.jacadv.2026.102864 
Dr. Khalid Youssef et al.

Points

Researchers from Upstate Medical University developed an explainable artificial intelligence scoring system to predict intramyocardial hemorrhage in STEMI patients before emergency arterial reperfusion.
Intramyocardial hemorrhage affects approximately forty percent of treated STEMI patients and significantly increases long term risks of progressive heart failure and cardiovascular mortality.
The transparent neural network model analyzes catheterization parameters in real time to show clinicians exact factors driving predictions rather than relying on uninterpretable algorithms.
Early identification using the clinical calculator enables targeted administration of investigational intravenous dexrazoxane evaluated during phase two clinical trials presented at European congresses.
Investigators concluded that point of care explainable artificial intelligence guides procedural decisions, post PCI cardiac MRI scheduling, and clinical trial enrollment for high risk patients.

Summary

This study evaluated an explainable artificial intelligence (XAI) scoring system using a scalable neural network (SNN) model to predict intramyocardial hemorrhage (IMH) prior to reperfusion in patients presenting with ST-segment elevation myocardial infarction (STEMI). Given that myocardial infarction accounts for over 800,000 cases annually in the United States alone and post-reperfusion IMH affects approximately 40% of treated STEMI patients, early risk identification is vital. The research sought to determine whether real-time integration of catheterization clinical parameters via interpretable XAI could accurately predict IMH to guide intra-procedural interventions and clinical trial selection.

Published in JACC: Advances and co-led by Dr. Ankur Kalra of Upstate Medical University, the scoring model rapidly processes emergency cardiac catheterization data before vessel recanalization. By leveraging transparent SNN architecture rather than opaque algorithms, interventional cardiologists can inspect the specific physiological variables driving individualized risk predictions at the point of care. The platform is designed for seamless deployment as an electronic health record-based risk calculator during emergency angiography to optimize post-procedure monitoring and cardiac MRI referral.

High-risk identification enables targeted adjuvant therapeutics, including investigational intravenous dexrazoxane evaluated during a Phase 2a clinical trial presented at the European Society of Cardiology Congress on August 31. The authors conclude that pre-reperfusion XAI risk stratification provides a usable, transparent tool in the catheterization lab to inform therapeutic decision-making, direct post-PCI management, and potentially mitigate reperfusion-induced microvascular myocardial injury in high-risk STEMI cohorts.

Link to the article: https://www.jacc.org/doi/10.1016/j.jacadv.2026.102864

References

Youssef, K., Vora, K. P., Gupta, R., Gruionu, G., Kumar, A., Puri, R., Reed, G. W., Kalra, A., & Dharmakumar, R. (2026). Predicting intramyocardial hemorrhage before reperfusion in stemi patients with intrinsically explainable artificial intelligence. JACC: Advances, 5(7), 102864. https://doi.org/10.1016/j.jacadv.2026.102864