Key points are not available for this paper at this time.
The time interval between two consecutive R-peaks in an electrocardiography signal is called the RR interval (RRI), and its non-Gaussianity is a valuable marker for assessing cardiovascular disease risk. While RRI non-Gaussianity can be quantified by fitting probability distributions, the selection of an appropriate distribution remains challenging. Although Gaussian scale mixture representation has proven effective for analyzing non-Gaussianity in various biological signals, its application to RRIs has not been explored. This paper examines the application of Gaussian scale mixture representation for quantifying these non-Gaussian characteristics in RRI data. Using open RRI databases from participants of different ages and with various cardiovascular diseases, we demonstrated that this representation, particularly using the Student’s t-distribution, provides better characterization of RRI distributions compared to conventional Gaussian and stable distributions. The non-Gaussianity indices derived from this approach, when combined with conventional heart rate variability measures, improved the classification of cardiovascular diseases such as congestive heart failure and hypertension, achieving an average area under the receiver operating characteristic curve of 0.835 across multiple classifiers. These findings establish Gaussian scale mixture representation as an effective framework for quantifying RRI non-Gaussianity, offering potential improvements in cardiovascular risk assessment.
Hagiyama et al. (Sat,) studied this question.