Quantifying physiological data with Lempel-Ziv complexity-certain issues

The oscillations observed in physiological data can be attributed largely to the presence of either a nonlinear deterministic or a nondeterministic component. The Lempel-Ziv [1976] complexity and its variants have been used successfully to quantify the regularity of these oscillations. The decrease in the complexity can be observed in the case of nontrivial deterministic patterns as well as correlated noise. Thus, any conclusion on the nature of the pattern based solely on the value of the Lempel-Ziv complexity is incomplete. In this paper, the use of the surrogate data technique is suggested to avoid spurious interpretation of this measure of complexity. The data sets considered include the uterine contraction obtained during active labor. The surrogates are generated using the amplitude adjusted Fourier transform and the iterated amplitude adjusted Fourier transform. The approximate entropy [S.M. Pincus, 1991] is used as an alternate measure to verify the results obtained.