Abstract
OBJECTIVE: This study investigates the predictive associations between motor abilities and executive functions in early childhood by examining brain functional connectivity patterns and their predictive value for developmental trajectories. METHODS: A longitudinal study recruited 256 healthy preschool children aged 3-6 years from kindergartens affiliated with Shandong Sport University, China. Participants underwent resting-state fMRI, standardized motor assessments (MABC-2), and cognitive testing at baseline, 6-month, and 12-month follow-up (with primary analyses focusing on baseline to 12-month changes). A novel machine learning framework integrated multimodal neuroimaging and behavioral data using graph neural networks and feature fusion architectures to model motor-cognitive developmental relationships. RESULTS: Motor skills showed progressive maturation, with fine motor percentiles increasing from 38.2±23.7 to 56.3±27.1. Sensorimotor network connectivity increased systematically (0.15±0.08 to 0.22±0.09), while attention networks followed inverted-U developmental patterns. The multimodal machine learning model achieved 76.8±4.3% accuracy for motor and 74.2±3.9% for executive function outcomes, outperforming single-domain models. Brain connectivity features contributed 58% of predictive variance, indicating that baseline neural patterns predict subsequent developmental changes, though causal relationships cannot be established from these observational data. CONCLUSIONS: These results highlight early brain functional connectivity-especially sensorimotor networks-as a key predictor of motor and executive function development. Findings support the identification of early neural biomarkers of developmental risk and inform evidence-based strategies in early childhood education and targeted motor interventions.