Robust Cortical Thickness Morphometry of Neonatal Brain and Systematic Evaluation Using Multi-Site MRI Datasets.

Area of Science:

• Neuroscience
• Medical Imaging
• Developmental Biology

Background:

• Neonatal brain development is rapid, making it vulnerable to structural anomalies from pre-term birth or injury.
• Accurate characterization of cortical thickness is vital for understanding developmental trajectories.
• Existing methods struggle with small neonatal brains, leading to inaccurate cortical surface extraction.

Purpose of the Study:

• To develop a novel framework for reconstructing neonatal white matter (WM) and pial surfaces.
• To enable accurate cortical thickness measurements in developing brains using T1-weighted MRI.
• To address challenges posed by large partial volumes in small neonatal brains.

Main Methods:

• A novel framework, NEOCIVET 2.0, utilizing deep neural networks for neonatal MRI segmentation.
• Enhancement of cortical boundary delineation using CSF/GM boundary detection and edge gradient information.
• A new skeletonization method for sulcal folding in regions lacking visible CSF voxels.
• Evaluation on three independent datasets (736 pre-term, 97 term neonates).

Main Results:

• NEOCIVET 2.0 demonstrated high accuracy (86.9% rated accurate) and robustness across diverse datasets.
• Mean displacement of reconstructed surfaces was less than one voxel size (0.532 ± 0.035 mm).
• Cortical thickness positively correlated with post-menstrual age (PMA), with significant regional growth differences observed.

Conclusions:

• NEOCIVET 2.0 provides a robust and reproducible method for neonatal cortical surface reconstruction and thickness measurement using T1-weighted MRI.
• The pipeline is valuable for studying early brain development and identifying potential anomalies.
• NEOCIVET 2.0 is publicly available via the CBRAIN platform for broader research application.