Effective Connectivity Analysis and Classification of Action Observation From Different Perspectives: An fMRI Study.

Abstract

Analyzing the effective connectivity characteristics of brain networks in the process of action observation is helpful for understanding the neurodynamic mechanisms during action observation. In this study, functional magnetic resonance imaging (fMRI) images were obtained from 20 participants who performed hand-object interaction observation tasks from the first-person perspective (1PP) and third-person perspective (3PP). On the basis of a meta-analysis, 11 key brain regions were extracted as nodes to build an action observation network. The weighted and directional connections between all of the nodes were investigated using partial directional coherence (PDC) analysis in five narrow frequency bands. The statistical analysis indicated that the ultra-low frequency band ( ≤ 0.04 Hz) exhibited significant activation compared with other frequency bands for both 1PP and 3PP. In addition, it was found that 3PP induced significantly stronger brain activation than 1PP in the ultra-low frequency band. Moreover, this study attempted to classify fMRI data corresponding to different perspectives using brain network features. A comparative analysis revealed that the weighted and binary PDC matrix methods achieved classification accuracies of 86.3% and 80.8%, respectively. The weighted PDC analysis exhibits a more comprehensive understanding of neural mechanisms during action observation in different visual perspectives. It also has potential applications value in human-computer interaction in the future.