Ordinal Pattern-Based Dissimilarity Measure for Slow Cortical Potential Training in Stroke

Biological signals, such as EEG and ECG, generate complex fluctuations in correspondence with the underlying system dynamics. In this study, we propose a dissimilarity quantification, which is an improvement of information-based similarity for capturing the features of underlying dynamics from positivity or negativity trials in the neurofeedback training of chronic stroke patients. Simulated Gaussian white and pink noises are used to evaluate the validity of this measure by different embedding dimensions, time delays, and data lengths. Then, the method is applied to slow cortical potentials of chronic stroke patients. The results imply that the proposed dissimilarity measure characterizes the unique dynamical patterns of SCP signals. The dissimilarity measure is capable of capturing the underlying dynamics of SCPs that belong to positivity or negativity trials. Besides, as the session progressed, the dissimilarity showed an increasing trend.