Traditional sample entropy fails to quantify inherent long-range dependencies among real data. Multiscale sample entropy (MSE) can detect intrinsic correlations in data, but it is usually used in univariate data. To generalize this method for multichannel data, we introduced multivariate multiscale entropy into multiscale signals as a reflection of the nonlinear dynamic correlation. But traditional multivariate multiscale entropy has a large quantity of computation and costs a large period of time and space for more channel system, so that it can not reflect the correlation between variables timely and accurately. In this paper, therefore, an improved multivariate multiscale entropy embeds on all variables at the same time, instead of embedding on a single variable as in the traditional methods, to solve the memory overflow while the number of channels rise, and it is more suitable for the actual multivariate signal analysis. The method was tested in simulation data and Bonn epilepsy dataset. The simulation results showed that the proposed method had a good performance to distinguish correlation data. Bonn epilepsy dataset experiment also showed that the method had a better classification accuracy among the five data set, especially with an accuracy of 100% for data collection of Z and S.
This paper explores a methodology used to discriminate the electroencephalograph (EEG) signals of patients with vegetative state (VS) and those with minimally conscious state (MCS). The model was derived from the EEG data of 33 patients in a calling name stimulation paradigm. The preprocessing algorithm was applied to remove the noises in the EEG data. Two types of features including sample entropy and multiscale entropy were chosen. Multiple kernel support vector machine was investigated to perform the training and classification. The experimental results showed that the alpha rhythm features of EEG signals in severe disorders of consciousness were significant. We achieved the average classification accuracy of 88.24%. It was concluded that the proposed method for the EEG signal classification for VS and MCS patients was effective. The approach in this study may eventually lead to a reliable tool for identifying severe disorder states of consciousness quantitatively. It would also provide the auxiliary basis of clinical assessment for the consciousness disorder degree.
In this paper, a feature extraction algorithm of weighted multiple multiscale entropy is proposed to solve the problem of information loss which is caused in the multiscale process of traditional multiscale entropy. Algorithm constructs the multiple data sequences from large to small on each scale. Then, considering the different contribution degrees of multiple data sequences to the entropy of the scale, the proportion of each sequence in the scale sequence is calculated by combining the correlation between the data sequences, so as to reconstruct the sample entropy of each scale. Compared with the traditional multiscale entropy the feature extraction algorithm based on weighted multiple multiscale entropy not only overcomes the problem of information loss, but also fully considers the correlation of sequences and the contribution to total entropy. It reduces the fluctuation between scales, and digs out the details of electroencephalography (EEG). Based on this algorithm, the EEG characteristics of autism spectrum disorder (ASD) children are analyzed, and the classification accuracy of the algorithm is increased by 23.0%, 10.4% and 6.4% as compared with the EEG extraction algorithm of sample entropy, traditional multiscale entropy and multiple multiscale entropy based on the delay value method, respectively. Based on this algorithm, the 19 channel EEG signals of ASD children and healthy children were analyzed. The results showed that the entropy of healthy children was slightly higher than that of the ASD children except the FP2 channel, and the numerical differences of F3, F7, F8, C3 and P3 channels were statistically significant (P<0.05). By classifying the weighted multiple multiscale entropy of each brain region, we found that the accuracy of the anterior temporal lobe (F7, F8) was the highest. It indicated that the anterior temporal lobe can be used as a sensitive brain area for assessing the brain function of ASD children.