Uncovering similarities in biomedical signals

Biomedical signals are generally complex, non-stationary, and non-periodic, making them difficult to analyze and to compare. Time-frequency correlation, in combination with information theoretic measures, can provide a clearer quantitative understanding of the relationships between these signals. Using the known correspondence between muscle sympathetic nerve activity (MSNA) and electrocardiography (ECG) (two signals that are frequently used in assessment and rehabilitation), these methods are employed to clarify MSNA-ECG relationships between differing experimental conditions. Results obtained with wavelet transforms based on both the standard Morlet function as well as on the selective discrete Fourier transform delineate areas of similarity and differences in both time and frequency during the lower-body negative pressure (LBNP) tests. Transfer entropy, computed with a GPU-based parallel algorithm, is used to quantify the lag between MSNA and ECG. It is shown that wavelet coherence and transfer entropy provide complementary information in uncovering common components in time and in frequency.