Estimation and Evaluation of New Features from Phonocardiogram for Detecting Cardiovascular Abnormalities

In the world, cardiovascular diseases are the major cause of death, as 31% of global mortality is from them. Due to problems such as the lack of technologies such as echocardiography, and limited access to cardiologists in deprived areas, automatic methods for detecting heart abnormalities in phonocardiogram (PCG) are used. In this study, to distinguish between normal and abnormal cases, three categories of features in PCG were estimated and evaluated. First, the extraction of the heart rate and heart rate variability; second, some of the features used in speech analysis and pattern recognition; and third, the time center, the frequency center, and the frequency variance of the signal. Some methods were proposed for extracting desired features, and the data were analyzed using t-test. The results of evaluation of the 10 proposed features, with the p-value of less than 0.010, showed that 8 features had significant distinction to detect abnormal cases from the normal ones. Regarding the patterns of the extracted features, the distinction between normal and abnormal signals was observed, which can be used to classify PCGs. Moreover, in the future, new features can be extracted from these patterns using some other analysis such as correlation.