Construction of 4D infant cortical surface atlases with sharp folding patterns via spherical patch-based group-wise...

Area of Science:

• Neuroscience
• Medical Imaging
• Developmental Biology

Background:

• Understanding dynamic early brain development requires high-resolution 4D (spatial + temporal) infant cortical surface atlases.
• Existing atlases often lack longitudinal consistency and detailed cortical attribute patterns, hindering accurate analysis.

Purpose of the Study:

• To construct a comprehensive set of 4D infant cortical surface atlases covering 11 time points from 1 to 72 months of age.
• To ensure longitudinal consistency, preserve sharp cortical attribute patterns, and facilitate accurate normalization of dynamic infant brain surfaces.

Main Methods:

• Utilized a two-stage group-wise surface registration for longitudinal consistency and unbiasedness.
• Employed spherical patch-based sparse representation with an augmented dictionary to maintain sharp cortical patterns and mitigate registration errors.
• Jointly represented geometric attributes, cortical thickness, and myelin content using group-wise sparsity for attribute consistency.

Main Results:

• Developed 4D infant cortical surface atlases with longitudinally consistent and sharp attribute patterns across 11 time points.
• The novel atlases preserve finer details of cortical folding compared to existing methods.
• Achieved improved accuracy in registering individual infant cortical surfaces to the atlas.

Conclusions:

• The constructed 4D infant cortical surface atlases provide a valuable tool for studying dynamic early brain development.
• These atlases enhance the precision of neuroimaging analysis by offering superior longitudinal consistency and pattern preservation.
• The inclusion of parcellations facilitates region-based analysis for a deeper understanding of infant brain maturation.