Evaluation of methods for volumetric analysis of pediatric brain data: The childmetrix pipeline versus adult-based approaches

Area of Science:

• Neuroimaging
• Pediatric Neuroscience
• Medical Image Analysis

Background:

• Pediatric brain volumetric analysis using MRI is crucial for understanding development and neurodevelopmental disorders.
• Existing methods face bias from head motion and anatomical disparities in pediatric brains, as adult-based atlases are inadequate.
• Accurate segmentation of gray matter (GM) and white matter (WM) is essential for reliable volumetric analysis.

Purpose of the Study:

• To introduce childmetrix, a novel tool for pediatric neuroimaging data analysis.
• To overcome limitations of adult-based methods in pediatric brain MRI analysis, particularly concerning motion artifacts and anatomical differences.
• To improve the accuracy, reproducibility, and robustness of gray matter and white matter segmentation in children's brains.

Main Methods:

• Development of childmetrix, utilizing an age-specific atlas and a probabilistic model-based approach for GM and WM segmentation.
• Validation on 55 pediatric scans (ages 5-8, including developmental dyslexia) and 10 test-retest pairs.
• Comparison with state-of-the-art adult-atlas-based methods: icobrain (probabilistic) and Freesurfer (surface-based).

Main Results:

• Childmetrix demonstrated superior reproducibility in GM and WM segmentations compared to Freesurfer.
• Childmetrix showed better robustness to head motion in GM volume estimation than Freesurfer.
• Childmetrix achieved 82-84% overlap with manual segmentation, outperforming adult-based methods, especially on low-quality data.
• Adult-based methods required twice as many subjects to detect morphological differences between dyslexic and typical readers.

Conclusions:

• Childmetrix provides more accurate and reliable pediatric brain volumetric measures than current adult-based methods.
• The tool's age-specific approach enhances robustness to motion and anatomical variations in pediatric neuroimaging.
• Childmetrix has the potential to significantly aid in the examination and understanding of children's brain development and disorders.