Schedule of Events | Symposia

Targeting Region-Specific Cerebrospinal Fluid Noise to Enhance Subcortical Neural Estimates in fMRI

Poster Session D - Monday, March 31, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom

Alexandra Fischbach¹ (fischbach.a@northeastern.edu), Hallee Shearer¹, Ajay Satpute¹, Karen Quigley¹, Jordan Theriault¹, Lisa Feldman Barrett¹, Stephanie Noble¹; ¹Northeastern University

Distinguishing neural signals from noise is a major challenge in neuroimaging, with physiological noise being a dominant contributor. Subcortical structures are disproportionately affected by physiological noise due to low signal-to-noise ratio, small size, and proximity to cerebrospinal fluid (CSF). Traditional CSF correction methods, pool signals from all CSF compartments into a single regressor which is applied to all regions of interest (ROIs). However, these methods may fail to account for the spatial heterogeneity of CSF-noise, potentially allowing residual noise to persist in the data. To investigate this, resting-state data (N=83) were analyzed to compute voxel-wise spatial correlations between CSF time-series and a weighted CSF average. The analysis revealed distinct spatial patterns across the inferior-superior axis, with correlations increasing towards the superior (top)regions of the brain, emphasizing the anatomical and functional heterogeneity of CSF distribution. We then explored whether a localized correction approach could better account for region-specific noise. To examine this, we developed a method to model CSF signals adjacent to each subcortical ROI. We evaluated its impact on resting-state functional connectivity (FC) estimates in comparison to standard global correction methods. Global correction introduced more negative correlations compared to local correction. A paired two-tailed t-test (FDR, α < 0.05, p < 0.05), revealed significant differences in edge strength for 85% of connections, with 80% demonstrating stronger FC following local CSF correction. Our results indicate that targeting region-specific CSF-noise increases sensitivity to subcortical neural estimates, highlighting the critical role of spatially-informed noise correction in preserving neural signals while reducing unwanted variance.

Topic Area: METHODS: Neuroimaging