Identifying brain targets for real-time fMRI neurofeedback in chronic pain: Insights from functional neurosurgery

Abstract

Background: The lack of clearly defined neuromodulation targets has contributed to the inconsistent results of real-time fMRI-based neurofeedback (rt-fMRI-NF) for the treatment of chronic pain. Functional neurosurgery (funcSurg) approaches have shown more consistent effects in reducing pain in patients with severe chronic pain. Objective: This study aims to redefine rt-fMRI-NF targets for chronic pain management informed by funcSurg studies. Methods: Based on independent systematic reviews, we identified the neuromodulation targets of the rt-fMRI-NF (in acute and chronic pain) and funcSurg (in chronic pain) studies. We then characterized the underlying functional networks using a subsample of the 7 T resting-state fMRI dataset from the Human Connectome Project. Principal component analyses (PCA) were used to identify dominant patterns (accounting for a cumulative explained variance >80%) within the obtained functional maps, and the overlap between these PCA maps and canonical intrinsic brain networks (default, salience, and sensorimotor) was calculated using a null map approach. Results: The anatomical targets used in rt-fMRI-NF and funcSurg approaches are largely distinct, with the middle cingulate cortex as a common target. Within the investigated canonical rs-fMRI networks, these approaches exhibit both divergent and overlapping functional connectivity patterns. Specifically, rt-fMRI-NF approaches primarily target the default mode network (P value range 0.001-0.002) and the salience network (P = 0.002), whereas funcSurg approaches predominantly target the salience network (P = 0.001) and the sensorimotor network (P value range 0.001-0.023). Conclusion: Key hubs of the salience and sensorimotor networks may represent promising targets for the therapeutic application of rt-fMRI-NF in chronic pain.