Effect of extreme data loss on heart rate signals quantified by entropy analysis

The phenomenon of data loss always occurs in the analysis of large databases. Maintaining the stability of analysis results in the event of data loss is very important. In this paper, we used a segmentation approach to generate a synthetic signal that is randomly wiped from data according to the Gaussian distribution and the exponential distribution of the original signal. Then, the logistic map is used as verification. Finally, two methods of measuring entropy—base-scale entropy and approximate entropy—are comparatively analyzed. Our results show the following: (1) Two key parameters—the percentage and the average length of removed data segments—can change the sequence complexity according to logistic map testing. (2) The calculation results have preferable stability for base-scale entropy analysis, which is not sensitive to data loss. (3) The loss percentage of HRV signals should be controlled below the range (p=30%), which can provide useful information in clinical applications.