A comprehensive processing framework crafted for pediatric brain MRI

Processing raw brain magnetic resonance (MR) data is essential for converting it into a standard format to facilitate further investigations. While adult-centric brain MR scan processing pipelines are well-established, the distinct characteristics of pediatric brains necessitate a customized approach to improve scan quality and reliability for comparative analysis. This study introduces a comprehensive processing framework tailored for pediatric T1-weighted brain MR scans (ages 2 to 8 years) by incorporating several processing steps: skull stripping, registration, bias field correction, normalization, resizing, and tissue segmentation. The framework utilizes various neuroimaging tools and Python-based libraries (FSL, Freesurfer, SimpleITK, SciPy, etc.). Our work integrated an atypical pediatric dataset containing 30 scans from children with prenatal alcohol exposure (PAE) along with a primary dataset of 279 scans from unexposed controls. For skull stripping, we recommend using FSL-BET and SynthStrip. For registration, an age-specific template is suggested to better align with the characteristics of developing pediatric brains. Bias correction and normalization are essential for addressing non-uniform and distinct voxel intensities, respectively. Depending on post-processing requirements, resizing, and tissue segmentation are available as optional steps. We developed the framework using unexposed controls, while the PAE data were utilized to validate its generalizability. Our study presents an advisory framework for processing pediatric data, where most outcomes were evaluated through visual inspection as a qualitative approach, along with some quantitative analyses in skull stripping, normalization, and tissue segmentation steps.