A new study from the University of Missouri reveals that artificial intelligence could significantly improve the accuracy of cardiac risk prediction and revolutionize personalized cardiac care. Researchers led by FAHA’s Fares Alahdab, MD, MSc, MSc, used machine learning to analyze positron emission tomography (PET) scans of heart disease patients to identify those at highest risk for major adverse cardiac events (MACE). “Our model more accurately assigned a patient’s MACE risk than other predictive models that interpret the data,” said Aradab, an associate professor at Mizzou School of Medicine. Published prior to printing in Journal of Nuclear Cardiology January 28, 2026 This advancement overcomes the limitations of traditional statistical analysis, offering the potential to optimize individual treatment plans and improve patients’ quality of life.

PET scan data improves MACE risk prediction accuracy

This advancement promises to go beyond the limitations of traditional statistical analyzes currently used to predict outcomes such as readmission risk. The strength of new models lies in their ability to handle complex datasets and variable relationships beyond the capabilities of traditional approaches. “We trained our model based on information from advanced nuclear scans of patients with coronary artery disease, and some of these techniques may be applicable to other diseases,” said Alahdab. Identifying high-risk individuals is paramount to tailoring treatment plans and improving patients’ quality of life, and this goal is emphasized by study results. “Identifying patients who are most at risk for adverse health outcomes is critical to customizing their care plans and preserving their quality of life,” said Aradab.

Machine learning models overcome the limitations of traditional evaluation

Traditional methods of predicting cardiac risk, often relying on statistical analysis, are now being challenged by artificial intelligence. This new approach avoids the limitations inherent in traditional assessments that struggle with both data volume and complex variable interactions. The results of this research have recently Journal of Nuclear Cardiology.