AI developed to predict mortality risk in acute coronary syndrome

Scientists from Pitirim Sorokin Syktyvkar State University and their colleagues have developed a new system to predict mortality risk in patients with acute coronary syndrome (ACS). This machine learning-based model demonstrated significantly higher accuracy compared to the traditional GRACE scale. This was reported by Ixbt.com reports .

Data from over 14,000 patients were used for the analysis, with 13,300 included in the final study. The algorithm evaluated 28 clinical parameters, including age, hemodynamic indicators, and laboratory data. The CatBoost ensemble model showed the most effective result, with its predictive capability (AUC-ROC) reaching 0.961, while the GRACE scale showed 0.919.

Project leader Ilya Solovyov, Candidate of Biological Sciences, noted that this technology allows for the creation of a precise individual risk profile as soon as the patient is admitted to the hospital. Using the SHAP method, researchers identified key factors influencing the prognosis, specifically left ventricular ejection fraction, heart failure severity, and blood pressure.

The researchers noted that these results do not yet mean the technology is ready for widespread clinical practice. Multi-center clinical trials are planned for the next stage. If these tests are successful, it is expected that new generation clinical decision support systems will be created, reducing mortality from cardiovascular diseases.