Advanced treatments for depressive symptoms, such as a real-time functional MRI (fMRI) neurofeedback have been proven by several studies. In particular, regularization of functional connectivity (FC) between executive control network (ECN) and default mode network (DMN) during fMRI neurofeedback have been proposed to reduce depressive symptoms. However, it is difficult to install this system anywhere and repetitively provide the treatment in practice, because the cost is high and no practical signal processing techniques have existed so far to extract FC-related features from EEG, particularly when no physical forward models are available. In this regard, stacked pooling and linear components estimation (SPLICE; Hirayama et al., 2017), recently proposed as a multilayer extension of independent component analysis (ICA) and related independent subspace analysis (ISA; Hyvarinen & Hoyer, 2000), can be a promising alternative, although its application to the resting-state EEG have never been investigated so far. We expected that if the extracted EEG network features were correlated with fMRI network activity corresponding to DMN or ECN, it may help to modulate the target FC in the EEG-based neurofeedback training.
Here, we describe a real-time EEG neurofeedback paradigm for improvement of depressive symptoms by using an EEG network features estimated by SPLICE. We hypothesized upregulation of the dorsolateral prefrontal cortex (DLPFC)/middle frontal cortex or downregulation of precuneus/posterior cingulate cortex (PCC) related to fMRI biomarker for depression (Ichikawa et al., Sci Rep, 2020) should specifically predict decreases in depressive symptoms during the neurofeedback training. We conducted a single-blind design for neurofeedback group (n=8; NF group) and sham group (n=9) groups for three days. To this end, we found large effect size in the rumination response scale score (total, brooding and reflection) in the comparison between NF and sham groups. Additionally, brain signals of the tasked fMRI (a contrast of 2-back > 0-back) in the neurofeedback group were significantly decreased in the right cuneus and DLPFC compared to the control group. We demonstrated a feasibility of EEG neurofeedback treatment for depressive symptoms using EEG network features extracted by SPLICE in the subclinical trials.