Contribution of In-Scanner Thoughts to Resting-State Functional Connectivity: How participants rest matters

Resting-state fMRI (rs-fMRI) scans are often used to identify aberrant patterns of functional connectivity (FC) in clinical populations and to reveal the neural correlates of specific phenotypes. To minimize interpretational uncertainty, researchers control for age, gender, co-morbidities, and motion. Yet, rarely considered is the role of systematic differences of in-scanner experience (i.e., what subjects are thinking during the scan). To evaluate this prospect, we used 471 publicly available rs-fMRI scans (MPI Mind-Brain-Body dataset) annotated with self-reports about the content and form of in-scanner thoughts, and perceived levels of wakefulness. Based on these self-reports, we subdivided our sample into groups with different in-scanner experience controlling for age, gender, and wakefulness. Group G1 is characterized by reporting thoughts in the form of images, of positive valence and about other people. Group G2 includes scans with thoughts focused primarily on the environment and of negative valence. For all scans, we estimated FC using the 400 ROI Schaefer Atlas augmented with 8 subcortical ROIs. Significant differences in FCacross groups were estimated using Network Based Statistics. We found stronger FC between the DMN and somatosensory and attentional networks for the contrast G1 > G2. In addition, we observed significantly stronger FCbetween sensory regions and attentional regions for the contrast G2 > G1. These results show that internally vs. externally-oriented thought engagement modulates FC between attentional regions and the rest of the brain. Next, we asked if we could predict aspects of in-scanner experience using FC. Prediction targets included: wakefulness, individual descriptors of thought form (images, words, intrusive, vague) and content (surroundings, other people, oneself, future/past events, positive/negative valence). Using connectome-predictive modeling, we were able to significantly predict wakefulness, reported levels of visual imagery, and focus on surroundings and past events. Inspection of FC models contributing to each prediction agree with our current understanding of how these state-level aspects of cognition manifest in the brain. Together, these results highlight the key role of in-scanner experience in shaping rs-fMRI FC and motivate the practice of annotating rs-fMRI scans with first-person descriptions of in-scanner experience. Future work should elucidate if accounting for these state-level effects help characterize sources of inter- and intra-subject variability that hinder our ability to interpret FC differences and develop rs-fMRI biomarkers of disease.