Decoding time-course of saliency network of fMRI signals by EEG signals using optimized forward variable selection: a concurrent EEG-fMRI study

Tung Dang, Kentaro Ono, Takafumi Sasaoka, Shigeto Yamawaki, Maro G. Machizawa

Abstract: Coinciding with the rapid growth of neural engineering technologies, the social demand for digital mental health is rapidly rising to date. While neuroimaging techniques, such as functional MRI or MEG, allow us to monitor neural activities in depth brain regions and sparse networks, direct application of those tools is limited in real life to assess neural well-being in real life. Literature suggests human interoceptive awareness is associated with activities and connections from the insular cortex as a part of the saliency network. Accumulating evidence suggests monitoring and modulation of these functional connectivities may reflect or improve our mental states. The neurophysiological link between signals from fMRI and EEG is isolated due to varieties of reasons; however, here we propose a novel signal processing approach to decode the time-course of functional connectivity obtained from fMRI by the concurrently recorded EEG signals (n = 28). To exploit the feasibility of this fusion network decoding, we present a novel machine learning method, "optimized forward variable selection decoder," that achieves higher decoding accuracy over classical methods. We propose a unified EEG model for brain-computer-interface applications in real life that can be applied for monitoring mental well-being in daily life.