Higher-Order Brain Network Analysis for Auditory Disease

摘要

Auditory diseases such as deafness and tinnitus have been plaguing people for a long time. On the one hand, although cochlear implantation may serve as a cure for deafness to some degree, the mechanism of developmental neuroplasticity in the auditory and visual systems has not been well understood. On the other hand, there is still no cure for tinnitus, and investigating the cause and then developing the cure of tinnitus is particularly necessary. EEG signals provide us insights into these auditory diseases and have been widely studied for developing the cure of auditory diseases, in particular from the brain network perspective. However, most of the existing methods either simply utilize lower-order features of the brain network at the level of local connections within selected brain regions or fail to analyze the EEG signals from the brain region connectivity perspective. In this paper, based on the EEG signals, we develop a new higher-order brain network analysis method termed HBNmining (higher-order brain network mining) based on the weighted motifs and colored motifs for deepening the understanding of the auditory diseases. In particular, after constructing brain network from EEG signals, both the weighted motifs and the colored motifs are extracted, from which subject classification and brain region connectivity analysis can be conducted respectively. The results have confirmed the effectiveness of our method, which may be helpful for clinical treatment of auditory diseases.