Kavli Affiliate: David Linden
| Authors: Xuelei Wang, Assunta Ciarlo, Michael Luehrs, Alexander Atanasyan, David Böken, Jürgen Roßmann, Michael Schluse, Maren Jäger, Marisa Nordt, Fengyu Cong, Klaus Mathiak, David Linden, Rainer Goebel, David MA Mehler and Jana Zweerings
| Summary:
Real-time functional magnetic resonance imaging neurofeedback (rt-fMRI-NF) is a promising non-invasive brain-computer-interface (BCI) technique for enhancing self-regulation of affective states in the brain. However, conventional univariate rt-fMRI-NF approaches are limited in their ability to distinguish neural patterns of distinct emotions that involve overlapping brain regions. In this study, we applied an rt-fMRI semantic neurofeedback (rt-fMRI-sNF) paradigm, incorporating real-time representational similarity analysis (rt-RSA) to enable navigation between emotional states. Four emotional patterns were first derived from functional localizer runs, each designed to evoke a specific emotion, and then applied as target patterns during neurofeedback. Using an RSA-informed circular semantic map (CSM), participants received real-time visual feedback indicating both the similarity and intensity of their current brain activity relative to target emotional patterns. Participants were instructed to use mental imagery to shift their brain activity toward the specific target pattern and increase its intensity. Twenty-four healthy participants completed the localizer runs, and two consecutive neurofeedback runs in the same session. Ten participants successfully engaged with both the similarity and intensity components of the CSM, showing effective modulations of their mental states. Analyses of the localizer runs revealed overlapping regional activations across emotions and demonstrated that RSA outperformed univariate analysis in distinguishing between them. For the neurofeedback runs, linear mixed-effects model (LMM) analyses across multiple performance metrics indicated consistent within-run improvements and higher initial performance in the second run, while significant between-run learning effects emerged only in exploratory models with quadratic time terms. A block-wise comparison also showed significantly higher performance at the end of each run compared to the beginning based on the intensity metric. These findings support the usability of RSA in differentiating multiple emotional states and demonstrate the feasibility of the rt-fMRI-sNF paradigm for emotion regulation.