A Comparative Study of Automatic Approaches for Preclinical MRI-based Brain Segmentation in the Developing Rat

Area of Science:

• Neuroscience
• Medical Imaging
• Biomedical Engineering

Background:

• Reproducible pre-clinical study reporting necessitates validated processing tools.
• Automated brain extraction is a critical, yet underdeveloped, step in preclinical neuroimaging.
• Developing robust tools for immature rat brain segmentation is essential for developmental studies.

Purpose of the Study:

• To evaluate and compare the accuracy of automatic brain extraction tools for immature rat T2-weighted MRI data.
• To identify the optimal software package for rodent brain segmentation in developmental imaging research.

Main Methods:

• Assessed skull stripping modules from AFNI, PCNN-3D, and RATS software.
• Compared automated segmentation results against manual segmentation using the Dice coefficient of similarity.
• Quantified false positive and false negative rates for each software package.

Main Results:

• The RATS package demonstrated a lower percentage of false positive non-brain voxels within the segmented brain mask.
• The AFNI package exhibited a lower percentage of false negative voxels, indicating better inclusion of brain tissue.
• Both automated methods significantly streamline data processing but have limitations.

Conclusions:

• Neither AFNI nor RATS provided perfect rodent brain segmentation for immature rats.
• The study highlights the necessity of supervised quality control for automated brain segmentation in developmental neuroimaging.
• Validated tools and quality control are vital for enhancing data reproducibility in preclinical research.