Related Experiment Video
Updated: Nov 8, 2025
06:48
Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
Published on: January 7, 2019
9.1K
Cortically constrained shape recognition: Automated white matter tract segmentation validated in the pediatric brain
Kesshi M Jordan1,2, Michael Lauricella1,2, Abigail E Licata1,2
1Department of Neurology, University of California, UCSF Memory and Aging Center, Sandler Neurosciences Center, San Francisco, California, USA.
Summary
A new automated tool, Cortically Constrained Shape Recognition (CCSR), accurately segments white matter (WM) bundles in pediatric brains. This method offers a repeatable and reliable approach for large-scale neuroimaging studies.
Area of Science:
- Neuroimaging
- Computational Neuroscience
- Developmental Neuroscience
Background:
- Manual segmentation of white matter (WM) bundles is time-consuming and requires extensive expertise.
- Automated methods are crucial for reproducible, large-scale studies, especially given developmental changes in pediatric brains.
- Existing methods may lack flexibility for capturing developmental variations in WM tracts.
Purpose of the Study:
- To introduce Cortically Constrained Shape Recognition (CCSR), a novel automated tool for WM bundle segmentation.
- To evaluate the performance and repeatability of CCSR in a pediatric cohort.
- To enable reproducible large-scale tractography analyses in neurodevelopmental research.
Main Methods:
- CCSR combines anatomical connectivity priors (FreeSurfer ROIs) with 3D streamline bundle atlases (RecoBundles).
- The tool was tested on main language WM tracts in a pediatric cohort (n=59) from the UCSF Dyslexia Center.
- Performance was assessed by comparing CCSR segmentations with expert manual segmentations using Dice Similarity Coefficient (DSC) and Lin's Concordance Correlation Coefficient (CCC).
Main Results:
- CCSR demonstrated high agreement with manual segmentations, achieving an average DSC of 0.76 for spatial overlap.
- Fractional anisotropy (FA) analysis showed an average CCC of 0.81, indicating strong agreement.
- Repeatability testing yielded an average DSC of 0.92, highlighting the method's reliability.
Conclusions:
- CCSR is a promising automated tool for segmenting WM bundles in pediatric populations.
- The approach facilitates reproducible, large-scale tractography analyses.
- CCSR aids in the quantitative assessment of structural connections relevant to neurodevelopmental disorders.
