Classification of Working Memory Loads via Assessing Broken Detailed Balance of EEG-FNIRS Neurovascular Coupling Measures

Human brain breaks the detailed balance to drive a variety of cognitive functions, such as memory. Recently, a promising classification framework of working memory loads has been proposed based on functional magnetic resonance imaging (fMRI) data with relative entropy (RE) measurement to quantify the broken detailed balance of human brain. However, there are limitations in previousely developed methods. First, single-modality fMRI can only detect the cerebral hemodynamics but not the neuronal activity, lacking detailed information of the neurovascular coupling process in the brain. Second, the RE measurement utilized to quantify the broken detailed balance may be biased and have no finite upper bound, limiting its application in high dimensional signal domains. In this study, a neurovascular coupling strategy based on concurrent electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) recordings was proposed to take both the cerebral hemedynamics and neuronal activity into consideration in assessing broken detailed balance of the brain. Furthermore, the generalized relative entropy (GRE) was employed to reduce the bias associated with the conventional RE measure. Our results demonstrated that the proposed framework showed higher classification accuracy (82.48%) to identify different levels of working memory loads than conventional methods. In addition, our results revealed that the broken detailed balance was significantly stronger when subjects performed cognitively demanding tasks (
P
<0.05) and was highly correlated with the neurovascular coupling models derived from the EEG
$\theta$
and
$\alpha$
bands, respectively. In conclusion, our findings provide an advanced framework to accurately classify various levels of working memory with the broken detailed balance of human brain and can be extended to explore the underlying broken detailed balance related to other cognitive behaviors and diseases.