0426 Graph Convolutional Network (GCN)-Based Prediction of Brain Age in Individuals with Insomnia

Abstract Introduction The impact of insomnia on brain aging and cognitive function remain undisclosed. The Brain Age Index has arisen as a robust biomarker for assessing an individual’s brain health, quantifying the deviations from normative aging trajectories. This study investigates accelerated brain aging in insomnia suffers compared to non-suffers using the Brain Age Index, exploring its correlation with the neurocognitive function. Methods Forty chronic insomnia patients (31 females, mean age 51.2 years) and 80 healthy controls (57 females, mean age 52.0 years) were analyzed. Graphical convolutional networks (GCNs) constructed regional brain age prediction models, integrating cortical morphology and topology data for comprehensive graph structure. Cortical thickness and gray-/white matter intensity ratio mapped on the cortical surface mesh were input to the GCNs. Ten brain network regions (sensorimotor, frontoparietal, dorsal attention, ventral attention, default mode, salience, language, auditory, visual, and limbic) were employed in the analysis of regional brain ages, utilizing automated anatomical labeling. Each model was trained on data from 6,563 healthy controls, with five-fold cross-validation indicating approximately three years of mean absolute error. These models computed the Brain Age Index for 80 healthy sleepers and 40 insomnia individuals. Linear regression models, adjusted for age, sex, and Beck Depression Inventory, scrutinized group disparities in regional Brain Age Indices. An additional analysis explored the association Brain Age Index and neuropsychological test utilizing linear regression analysis. Results While the global Brain Age Index revealed no significant group differences, notable regional Brain Age Index distinctions emerged in insomnia subjects. The regional Brain Age Index were significantly elevated in three distinct brain network regions—limbic, frontoparietal, and language (in descending significance order)—compared to normal subjects. The Brain Age Index within the language network exhibited a discernible negative correlation with neuropsychological deficits in attention and visuospatial domains. Conclusion This pioneering study validates neuroimaging-driven brain age in chronic insomnia, unveiling accelerated brain aging in distinct regions. Such revelations signify the potential impact of insomnia on the brain aging trajectory independently of chronological age. The potential association between accelerated brain aging in the language network and cognitive impairments has the potential to elucidate the intricate interplay between brain aging patterns and cognitive performance. Support (if any)