Prediction of Q-CHAT scores based on Functional Connectivity in Healthy Newborns

Neonatal resting-state functional connectivity may provide early markers of later variation in Q-CHAT scores, measured dimensionally within a non-clinical population, but the large-scale systems carrying the most robust predictive signal remain unclear. Using resting-state fMRI data from 397 infants in the Developing Human Connectome Project (277 term-born, 120 preterm-born), we applied a stability-driven, ROI-constrained connectome-based predictive modeling framework to predict 18-month Quantitative Checklist for Autism in Toddlers (Q-CHAT) scores. Significant prediction was observed in the whole cohort and in term-born infants, but did not reach statistical significance in the preterm-only group. Across the statistically significant models (whole cohort and term-born infants), the most prominent hubs were located in occipital and adjacent cortical regions, including the middle occipital gyrus, lingual gyrus, calcarine gyrus, and rolandic operculum. At the network level, the strongest predictive connections linked visual and visual-association systems with auditory networks, with additional contributions from medial motor, temporoparietal, and prefrontal systems. These findings suggest that later variation in Q-CHAT scores, is associated with neonatal large-scale functional organization, particularly in sensory and multisensory pathways. Keywords neonatal rs-fMRI; functional connectivity; Q-CHAT; connectome-based predictive modeling; preterm birth