Statistical Inferences on Brain Functional Networks Using Graph Theory and Multivariate Wavestrapping: An fNIRS Hyperscanning Illustration

Over the past two decades, there has been a rapid advancement in biomedical technology for instrumentation. This progress has not only revolutionized various aspects of our daily lives but has also significantly enhanced the capabilities of neuroscientific tools designed for monitoring brain signals. Among the various modalities of signal acquisition, one that is particularly promising for social Neuroscience is hyperscanning based on functional near-infrared spectroscopy (fNIRS). This modality makes it possible to estimate the temporal changes on oxyhemoglobin and deoxyhemoglobin concentrations across multiple brain regions simultaneously, concurrently in two or more individuals. In certain conditions, these hemodynamic fluctuations can be regarded as indirect measures of local neuronal activity. Consequently, through the application of fNIRS hyperscanning, researchers can investigate interbrain coupling dynamics during tasks that involve social interactions. In the current chapter, we present a novel non-parametric statistical test to evaluate interbrain coupling. The proposal is based on a combination of discrete wavelet transform and bootstrap. We illustrate the usefulness of the proposed method both in synthetic (Monte Carlo simulations) and real fNIRS hyperscanning data. The real data was acquired in the educational context of a teacher interacting with a child during a computational thinking activity. The results suggest that the wavelet bootstrap approach is indeed suitable for the analysis of hyperscanning data.