Characterization of fetal heart rate irregularity using approximate entropy and wavelet filtering

Approximate entropy (ApEn) has been used by several researchers as a means to characterize the irregularity of fetal heart rate (FHR), a major contributor to fetal risk assessment. The present work starts by analyzing the influence of accelerative and decelerative (AD) components of the FHR signal in the computation of ApEn. Having shown that the AD components may completely distort the ApEn evaluation we proceed to present two methods of eliminating the influence of the AD components: wavelet filtering; AD detection using specific algorithms based on well-established physiological criteria. We obtained good results with the first method when applied to simulated signals; however it proved unreliable in the practical application of discriminating normal from pathological fetal FHR. With the second method we were able to achieve this discrimination with high statistical significance.