Abstract
OBJECTIVE: There is evidence that complex relationships exist between motor functions, brain structure, and cognitive functions, particularly in the aging population. However, whether such relationships observed in older adults could extend to other age groups (e.g., younger adults) remains to be elucidated. Thus, the current study addressed this gap in the literature by investigating potential associations between motor functions, brain structure, and cognitive functions in a large cohort of young adults. METHODS: In the current study, data from 910 participants (22-35 yr) were retrieved from the Human Connectome Project. Interactions between motor functions (i.e., cardiorespiratory fitness, gait speed, hand dexterity, and handgrip strength), brain structure (i.e., cortical thickness, surface area, and subcortical volumes), and cognitive functions were examined using linear mixed-effects models and mediation analyses. The performance of different machine-learning classifiers to discriminate young adults at three different levels (related to each motor function) was compared. RESULTS: Cardiorespiratory fitness and hand dexterity were positively associated with fluid and crystallized intelligence in young adults, whereas gait speed and handgrip strength were correlated with specific measures of fluid intelligence (e.g., inhibitory control, flexibility, sustained attention, and spatial orientation; false discovery rate [FDR] corrected, p < 0.05). The relationships between cardiorespiratory fitness and domains of cognitive function were mediated by surface area and cortical volume in regions involved in the default mode, sensorimotor, and limbic networks (FDR corrected, p < 0.05). Associations between handgrip strength and fluid intelligence were mediated by surface area and volume in regions involved in the salience and limbic networks (FDR corrected, p < 0.05). Four machine-learning classifiers with feature importance ranking were built to discriminate young adults with different levels of cardiorespiratory fitness (random forest), gait speed, hand dexterity (support vector machine with the radial kernel), and handgrip strength (artificial neural network). CONCLUSIONS: In summary, similar to observations in older adults, the current study provides empirical evidence (i) that motor functions in young adults are positively related to specific measures of cognitive functions, and (ii) that such relationships are at least partially mediated by distinct brain structures. Furthermore, our analyses suggest that machine-learning classifier has a promising potential to be used as a classification tool and decision support for identifying populations with below-average motor and cognitive functions.