Using artificial intelligence technologies to predict outcomes and support medical decision-making in patients with suspected coronary artery disease

Aim. To evaluate the role of clinical parameters, pre-test probability (PTP) of coronary artery disease (CAD), and its modifying factors in predicting the risk of cardiovascular events in a modern Russian cohort of patients with suspected CAD using conventional statistical methods and artificial intelligence (AI).

Material and methods. The prospective observational study included 210 patients (115 men (54,8%), aged 60,1±10,1 years). The PTP of coronary artery disease (CAD) was assessed. Glucose, lipid profile, and creatinine levels were determined. Electrocardiography, carotid and femoral artery ultrasound was performed. The prospective follow-up period was 21 [19-25] months. Cardiovascular endpoints (CVEs) included cardiovascular death, acute coronary syndrome, and myocardial revascularization. Statistical analysis was performed using Statistica for Windows, version 16.0 (StatSoft, USA). Prognostic models were built using the Python programming language (version 3.x) and the Scikit-learn, Pandas, and NumPy machine learning libraries.

Results. The PTP of CAD was 17% [11-26%]. Prognostic data were obtained for all patients (100%), with CVE recorded in 51 of them (24,3%). By a univariate analysis, a higher risk of CVE was associated with CAD PTP, carotid and femoral atherosclerosis; by a multivariate Cox regression — with CAD PTP and femoral atherosclerosis (sensitivity — 63%, specificity — 64%, accuracy — 64%, AUC — 0,68, p<0,001). AI methods revealed that CAD PTP and triglyceride (TG) levels were independent predictors of CVEs (sensitivity — 61%, specificity — 66,7%, accuracy — 69%, AUC — 0,70, p<0,001). The TG level had a direct impact on the prognosis in the cohort with CAD PTP of 16-23% as follows: the TG ≥1,7 mmol/L was a marker of unfavorable prognosis. The personalized prognosis was calculated using the equation "CVE probability=1/(1+exp(-z))", where z=-1,5674+(0,0592×CAP PTP, %)+(0,1871×TG); a value >0,53 (53%) indicated an unfavorable prognosis.

Conclusion. Using AI technologies, we established that CAD PTP serves as the most accurate indicator for the primary cardiovascular risk stratification and personalized referral for specific diagnostics in suspected CAD. In addition to CAD PTP, the detection of femoral atherosclerosis is associated with an increased risk of CVE. TG levels are of additional importance for decision making in the cohort with a CAD PTP of 16-23%.

1. Federal State Statistics Service. Health In Russia. 2025. Statistical book. M.: Rosstat; 2025. 149 p. (In Russ.)

2. Barbarash OL, Karpov YuA, Panov AV, et al. 2024 Clinical practice guidelines for Stable coronary artery disease. Russian Journal of Cardiology. 2024;29(9):6110. (In Russ.) doi:10.15829/1560-4071-2024-6110. EDN: HHJJUT.

3. Foldyna B, Udelson JE, Karády J, et al. Pretest probability for patients with suspected obstructive coronary artery disease: re-evaluating Diamond-Forrester for the contemporary era and clinical implications: insights from the PROMISE trial. Eur Heart J Cardiovasc Imaging. 2019;20(5):574-81. doi:10.1093/ehjci/jey182.

4. Vrints C, Andreotti F, Koskinas KC, et al. 2024 ESC Guidelines for the management of chronic coronary syndromes. European Heart Journal. 2024;45(36):3415-537. doi:10.1093/eurheartj/ehae177.

5. Steyerberg EW. Clinical Prediction Models: A Practical Approach to Development, Validation, and Updating. 2nd ed. Springer International Publishing. 2019. 558 p. doi:10.1007/9783-030-16399-0.

6. Vabalas A, Gowen E, Poliakoff E, et al. Machine learning algorithm validation with a limited sample size. PLoS One. 2019;14(11):e0224365. doi:10.1371/journal.pone.0224365.

7. Petmezas G, Stefanopoulos L, Kilintzis V, et al. State-of-the-art deep learning methods on electrocardiogram data: systematic review. JMIR Med Inform. 2022;10:e38454. doi:10.2196/38454.

8. Berchialla P, Foltran F, Bigi R, et al. Integrating stress-related ventricular functional and angiographic data in preventive cardiology: a unified approach implementing a Bayesian network. J Eval Clin Pract. 2012;18(3):637-43. doi:10.1111/j.1365-2753.2011.01651.x.

9. Kovelkova MN, Iakovleva EG. Artificial intelligence in the prevention and diagnosis of cardiovascular diseases In Russia (literature review). Siberian Journal of Clinical and Experimental Medicine. 2025;40(1):28-41. (In Russ.) doi:10.29001/2073-8552-2025-40-128-41. EDN: ZKPMNP.

10. Simonyan MA, Kalyuta TY, Genkal EN, et al. Pretest probability of coronary artery disease as a factor for optimizing invasive diagnostics in routine clinical practice. Russian Journal of Cardiology. 2022;27(1):4765. (In Russ.) doi:10.15829/1560-4071-2022-4765. EDN: JUNIYL.

11. Laclaustra M, Casasnovas JA, Fernandez-Ortiz A, et al. Femoral and Carotid Subclinical Atherosclerosis Association With Risk Factors and Coronary Calcium: The AWHS Study. J Am Coll Cardiol. 2016;67(11):1263-74. doi:10.1016/j.jacc.2015.12.056.

12. Dalager S, Falk E, Kristensen IB, et al. Plaque in superficial femoral arteries indicates generalized atherosclerosis and vulnerability to coronary death: an autopsy study. J Vasc Surg. 2008;47(2):296-302. doi:10.1016/j.jvs.2007.10.037.

13. Zhuravleva OA, Ryabova TR, Vrublevsky AV, et al. Stress Echocardiography by the ABCDE Protocol ln the Assessment of Prognosis of Stable Coronary Heart Disease. Kardiologiia. 2024;64(4):22-30. (In Russ.) doi:10.18087/cardio.2024.4.n2572. EDN: KVVBHN.

14. Ezhov MV, Kukharchuk VV, Sergienko IV, et al. Disorders of lipid metabolism. Clinical Guidelines 2023. Russian Journal of Cardiology. 2023;28(5):5471. (In Russ.) doi:10.15829/1560-4071-2023-5471. EDN: YVZOWJ.

15. Marston NA, Giugliano RP, Im K, et al. Association Between Triglyceride Lowering and Reduction of Cardiovascular Risk Across Multiple Lipid-Lowering Therapeutic Classes: A Systematic Review and Meta-Regression Analysis of Randomized Controlled Trials. Circulation. 2019;140(16):1308-17. doi:10.1161/circulationaha.119.041998.