Study On The Effect Of Data Lack On Tracing ECG Signal Characteristics

With the rapid development of information technology, how to centralize and distill the data information hidden in a large number of information which looks disorderly and unsystematicly, in order that find out the inherent laws of the object of study, and according to the prediction of the development trend in the future become the hotspot of theory research. However, in the analysis of environmental monitoring, census, archaeology and biomedical research and other large data, they would often appear loss data directly or collected data cannot be used, which bring many difficulties to the analysis and application of data. Therefore, a measure of the quality of a data analysis method, it is particularly important that the analysis result is able to maintain stability in the case of missing data.Take the heart rate variability signal as an example, many factors may result in data loss at the time of collection, such as the complex operation of equipment, wide variety of wires and electrodes, the longer the monitoring time, electrode loose contact, as well as body position change and so on. Moreover, once the data is missing, and generally has no secondary collection. So to judge the rationality and practicality of an algorithm, we should first consider whether the algorithm is more sensitive to data loss. If overly sensitive, it will cause great obstacles to detect and capture useful information in the actual signal, and affect the accuracy of analysis results.Based on this idea, we conducted a series of exploration and research, the main research work and innovations are as follows:(1)In order to simulate data loss, we first generate a binary series with the same length as the intact original series, using the segmentation approach. Compared with the direct artificially delete some data, this method is more random and universality. And it can more accurately analyze and describe the influence degree of the sequence complexity in case of data loss by setting the three parameters: the percentage data p of removed data,the average length u means of the removed data segments, the functional form P(L) of the distribution of the length L of the removed data segments.(2) A theoretically chaotic system-Logistic map as a verification is used, based on two different probability distribution of segment lengths L-exponential and Gaussian distribution, we calculate the base-scale entropy in the case of data loss with different degree,so that facilitate comparative analysis on the complete sequence and the situation of date loss. The results show that the base-scale entropy can totally capture and quantify the complexity changes of sequence and is not affected by data loss. Moreover, we find that the two key parameters which are the percentage and the average length of missing data segments can cause change to sequence complexity. But no matter which kind of distribution is the length of the missing data segments, it has little effect on the results of the analysis.(3) The actual HRV signal as a verification is used, based on two different probability distribution of segment lengths L, we calculate the base-scale entropy and approximate entropy in the case of data loss with different degree, so that facilitate comparative analysis on the complete sequence and the situation of date loss. The results show that the approximate entropy is more sensitive to data loss, the strong sensitivity does not apply to analysis the HRV signal in case of data loss, and often mislead to the judgment of physiological and pathological states for patients. In contrast, the method of base-scale entropy has its unique advantages in this regard. Its results have preferable stability and can provide useful information for accurately judging physiological and pathological status in the case of data loss in clinical applications.