An end-to-end infant brain parcellation pipeline.

Area of Science:

• Neuroimaging
• Developmental Neuroscience
• Medical Image Analysis

Background:

• Infant brain parcellation is vital for understanding early development.
• Challenges include low contrast, noise, and partial volume effects in infant MRIs.

Purpose of the Study:

• To develop an end-to-end pipeline for accurate infant brain MRI parcellation.

Main Methods:

• A two-stage global-to-local approach was developed.
• Transformer and convolution operations were used for global ROI localization.
• Position priors and raw MRIs refined local ROI boundaries.

Main Results:

• The pipeline demonstrated superior accuracy and robustness on 263 subjects.
• Evaluated using the Dice ratio across multiple datasets (NDAR, BCP).
• Outperformed competing infant brain parcellation methods.

Conclusions:

• The developed end-to-end pipeline accurately parcellates 6-month-old infant brain MRIs.
• This tool can advance research in early brain development and disorders.