VOLUMETRIC AND VBM MEASURES IN 1.5-5 YEARS OLD CHILDREN OBTAINED USING ADULT SEGMENTATION TOOLS

Abstract

MRI-based volumetric and morphometric studies in a healthy pediatric population give a unique opportunity to investigate brain development, potentially leading to development of structural markers for neurological and psychiatric diseases. However, pediatric data analysis presents significant challenges for established processing tools, which were initially developed for adult population. This study aimed to investigate sexual dimorphism and age-related changes in neural tissues in healthy 1.5-5-years-old children and to critically assess the feasibility of the use of popular software such as CAT-12, FSL SIENAX and FSL VBM to obtain volumetric and VBM measures in this age group. Results showed inter-method inconsistency in estimations of total intracranial (TIV), grey (GM) and white matter (WM) volumes. Nonetheless, TIV and GM measures proved to be highly correlated with each other regardless of the chosen processing tool. As tissue segmentation is an essential part of the VBM analysis, quality of the GM and WM segmentations were assessed using Dice coefficients against manually corrected, curated FreeSurfer segmentations. Regardless of the used method, the quality of the segmentation was higher for the group of children of age 5 compared to 1.5-2-years-old group (toddlers); and for GM compared to WM. The amount of statistically significant voxels for FSL VBM results was noticeably higher than for CAT-12. FSL VBM analysis revealed higher GM volumes in females compared to males in the left auditory cortex, while CAT-12 showed no statistically significant difference. CAT-12 and FSL VBM agreed on increased GM volumes in toddlers compared to 5-year-olds in the frontal lobe, lingual gyri and cerebellum; and in putamina in 5-year-olds compared to toddlers. The results indicate that we need to be cautious when interpreting the neuroimaging findings in younger children as they may significantly vary due to the differences in used preprocessing methods and statistical analysis.