Discriminant analysis of variational pulsometry parameters

Abstract

The article is devoted to an important applied problem - the exploration of heart rate variability using linear discriminant analysis. Statistical methods are often used in modern medicine, especially in cardiology. One of these methods is discriminant analysis. It allows studying the differences between groups of objects in several variables. When using the discriminant analysis method, the main indicator is the accuracy of classification, and this indicator can be determined by estimating the proportion of correctly classified patients using a predictive observation equation. The mathematical component of the rapid diagnosis of heart rhythm was developed. The durations of the RR cardio intervals of patients at the age of 60-70 years are the initial data in the research. The data were taken from the databases of medical signals of the open international resource Physionet. The calculated parameters of variational heart rate monitoring for healthy people are presented with different duration of ECG recordings from 5 to 30 minutes. Statistical processing was performed for small samples using the Student criterion with a confidence probability of 95% according to the standard method. Mathematical data processing was carried out by the method of discriminant analysis of variational pulsometry parameters of healthy and sick patients with heart failure. The parameters with weak cross-correlation with high predictive significance, the spread of the values of which corresponds to the normal distribution law, were selected for the discriminant analysis. A function which allows to identify the patient as “sick” or “healthy” (one of the categories) has been determined. The analysis performed for patients diagnosed with arrhythmia also showed a difference from healthy patients, although less pronounced. It was demonstrated that discriminant analysis in cardiology can be effectively used for instant diagnosis of heart rate variability.

Keywords

• discriminant analysis
• variational pulsometry
• heart rate variability
• diagnosis of cardiac activity
• mathematical processing