Session SA4.5

Does Sample Entropy Reflect Nonlinear Characteristics of Cardiovascular Murmurs?

SE Schmidt*, M Græbe, J Hansen, E Toft, JJ Struijk

Aalborg University
Aalborg , Denmark

Sample entropy is well suited for quantification of some nonlinear or chaotic signals. Several studies have successfully applied sample entropy for quantification of cardiovascular murmurs and some studies suggest that cardiovascular murmurs are nonlinear and chaotic in nature, but are sample entropy influenced by nonlinear and chaotic characteristics of cardiovascular murmurs? We obtained seven digital recordings from subjects with known carotid artery stenosis and audible carotid murmurs with an electronic stethoscope. Murmurs were isolated manually and sample entropies were calculated from both periods with murmur and silent periods of equal length. Sample entropy was calculated using predetermined embedding parameters (embedding dimension 2 and embedding lag 7) and recording specific embedding parameters determined by the false nearest neighbor method and the mutual information method. To test for the effect of non-linear signal components the sample entropy for each recording was compared with sample entropy from surrogates with randomized phase but otherwise identical linear properties as the recordings. Additionally the appropriateness of sample entropy for murmur quantification was evaluated by comparing sample entropy obtained with predetermined embedding parameters between periods with murmurs and silent periods. None of the sample entropy measures from the seven recordings differed significantly from their surrogates with known linear properties. The sample entropy calculated using fixed embedding parameters increased significantly in periods with audible murmurs compared with periods without, the mean increase was 0.21(95% CI: 0.11-0.32). According to a Mann–Whitney U test this increase was, however, not significantly (p=0.74) different from the difference between sample entropy from surrogates simulating murmur periods and surrogates simulating silent periods. The correlation between the sample entropy from recordings and surrogates was 0.95 (95% CI: 0.85-0.99). Our study confirms that sample entropy is suited for quantification of cardiovascular murmurs, but there is no evidence that the sample entropy is influenced by nonlinear signal components of the cardiovascular murmurs. This indicates that the changes observed in sample entropy under the presence of cardiovascular murmurs are due to changes in linear characteristics of the sounds.

(Abstract Control Number: 172)