Application of Dispersion Entropy to Healthy and Pathological Heartbeat ECG Segments

Augmentation of Dispersion Entropy for Handling Missing and Outlier Samples in Physiological Signal Monitoring

“…However, the second factor that leads to increased standard deviation values of the mean performance error is the small sample length of the analysed windows. An important advantage of the original DisEn algorithm is its capacity to acquire valuable insights, even when applied to time-series windows with small sample lengths [ 22 , 33 ]; for that reason, we chose to test the performance of the original DisEn algorithm and its variations using 360 samples per window, which is a considerably smaller sample length than what was commonly used in similar studies concerning the performance of ApEn and SampEn when applied to time-series containing missing and outlier samples [ 18 , 19 , 21 ]. Considering the observed fluctuations in the algorithmic performance recorded in our study, we recommend that, for field applications where the DisEn value of each window is considered individually, a larger sample length is used when the analysed time-series are expected to contain missing and outlier samples.…”

“…Furthermore, the DisEn algorithm can be implemented using a wide variety of mapping functions. Within the scope of this study, the logarithm sigmoid function is used due to its successful implementation in previous studies of physiological signal analysis [ 22 , 33 ]. However, as mentioned in Section 4.4 , outlier samples tend to disrupt the process of class allocation, which follows the mapping of the original time-series with the chosen mapping function.…”

“…Each original time series containing N points without missing samples is separated in non-overlapping windows of 360 samples each. The choice of window length is made with the aim to test algorithmic performance under the restriction of small sample lengths, which is considered one of the advantages of the original DisEn algorithm [ 22 , 33 ]. The original algorithm DisEn is used to calculate the ground truth DisEn value of each respective window.…”

“…The parameter values are selected after consulting the respective literature [ 14 , 22 ] and considering that each input window used in our study has a length of 360 samples [ 33 ]. The MATLAB codes for the implementation of the algorithmic variations presented in this paper are publicly available at .…”

Assessment of Dispersion and Bubble Entropy Measures for Enhancing Preterm Birth Prediction Based on Electrohysterographic Signals

“…Dispersion entropy is also faster than other related entropy measures due to the fact that it does not need to either sort the amplitude values of each embedding vector or calculate every distance between any two composite delay vectors with embedding dimensions m and m + 1 [ 28 ]. Kafantaris et al found that dispersion entropy obtained significantly higher values for both atrial premature beats and premature ventricular contraction electrocardiogram signals than healthy subjects’ [ 31 ]. Dispersion entropy of the discrete wavelet transformed electroencephalogram (EEG) has also been used for differential diagnosis of health control, mild cognitive impairment, and Alzheimer’s disease [ 32 ].…”