Analýza vlivu parametrů na výpočet hodnoty středního tlaku (MAP) pomocí strojového učení

Abstract

This thesis focuses on the prediction of mean arterial pressure (MAP) using machine learning methods, specifically regression models such as support vector regression (SVR), LASSO regression and Random Forest. These models were selected based on a search focused on machine learning in the field of cardiovascular diagnostics. The aim of this study was to compare the performance of these models in predicting MAP based on different parameters such as systolic and diastolic blood pressure, age, gender and BMI. The data used in this study were obtained from a web-based database containing measured blood pressure values and anthropometric parameters of 269 subjects. First, the SVR, LASSO regression and Random Forest models were compared with each other, and the best model (SVR) was then compared with commonly used methods of measuring MAP. The results showed that the SVR model showed more variability in predicted MAP values compared to the other models, but due to the lack of a reference value, it is not possible to determine whether it is more accurate than traditional methods. This work confirms that machine learning has the potential to improve MAP prediction and can be applied in clinical practice for more comprehensive cardiovascular health assessment.