MATHEMATICAL MODELS AND NUMERICAL METHODS IN CARDIOLOGY: OWN EXPERIENCE

Abstract

Technologies of machine learning and artificial intelligence are rapidly entering into medicine, including the diagnostic process of cardiovascular diseases. In this article, the authors share their own experience of using machine learning to create predictive models for cardiovascular pathology. Purpose of the study. To develop ways to effectively assess the risk of cardiovascular outcomes in patients with CHF, as well as in those with IHD. Patient characteristics and study methods. The study includes three options for predicting outcomes. 1st - risk assessment of the fatal annual cardiovascular events in patients with CHF, 2nd - risk assessment of the hospitalization due to decompensation of CHF within six months, 3rd - risk assessment of the combination of AF and IHD. Statistical analysis methods included machine learning methods: random forest (RF) and Shapley method (Shapley Values), free cross-platform visual programming system Orange, IBM SPSS Statistics package 28.0, C programming language console. Results and conclusion. The study demonstrated that the use of machine learning, the C programming language, and a console interface enables the development of cardiovascular outcome prediction models with high accuracy, interpretability, and robustness compared to traditional statistical models, providing a personalized approach to prediction.

Ähnliche Arbeiten