Deep Learning-Based Pediatric Brain Region Segmentation and Volumetric Analysis for General Growth Pattern in Healthy Children.

Abstract

To establish a quantitative reference for brain structural changes in children with neurological disorders, we employed deep learning technique to brain region segmentation and volumetric analysis within a cohort of healthy children. In this study, we recruited 312 participants aged 1.5 to 14.5 years (210 boys and 102 girls), dividing them into five age groups. High-resolution structural T1-weighted images were obtained, and an established toolkit utilizing deep learning algorithms was employed for brain region segmentation. For each age group, the volumes of gray matter and white matter, along with the thickness and surface area of the cortex, were calculated and compared between boys and girls. The results indicated that the volumes of gray matter and white matter in both bilateral cerebral hemispheres, as well as the total brain volume, increased with age. Furthermore, the volumes of the left and right hippocampus, amygdala, and thalamus also demonstrated an increase as age progressed. Conversely, cortical thickness and surface area decreased with age. Our findings provide a quantitative reference for understanding brain structural changes in children with neurological disorders.