Accelerating the Computation of Entropy Measures by Exploiting Vectors with Dissimilarity

Abstract

In the diagnosis of neurological diseases and assessment of brain function, entropy measures for quantifying electroencephalogram (EEG) signals are attracting ever-increasing attention worldwide. However, some entropy measures, such as approximate entropy (ApEn), sample entropy (SpEn), multiscale entropy and so on, imply high computational costs because their computations are based on hundreds of data points. In this paper, we propose an effective and practical method to accelerate the computation of these entropy measures by exploiting vectors with dissimilarity (VDS). By means of the VDS decision, distance calculations of most dissimilar vectors can be avoided during computation. The experimental results show that, compared with the conventional method, the proposed VDS method enables a reduction of the average computation time of SpEn in random signals and EEG signals by 78.5% and 78.9%, respectively. The computation times are consistently reduced by about 80.1~82.8% for five kinds of EEG signals of different lengths. The experiments further demonstrate the use of the VDS method not only to accelerate the computation of SpEn in electromyography and electrocardiogram signals but also to accelerate the computations of time-shift multiscale entropy and ApEn in EEG signals. All results indicate that the VDS method is a powerful strategy for accelerating the computation of entropy measures and has promising application potential in the field of biomedical informatics.

Keywords

• entropy
• approximate entropy
• sample entropy
• multiscale entropy
• EEG
• time-shift multiscale entropy