Brain Parcellation Repeatability and Reproducibility Using Conventional and Quantitative 3D MR Imaging.

Area of Science:

• Radiology
• Neuroimaging
• Medical Physics

Background:

• Automatic brain parcellation typically requires dedicated MRI sequences, increasing examination time.
• Quantitative MRI (qMRI) sequences can acquire R1, R2, and proton density maps.

Purpose of the Study:

• To synthesize T1-weighted (T1w) images from qMRI data for brain volume measurement and parcellation.
• To evaluate the repeatability and reproducibility of brain parcellation using synthetic vs. conventional T1w images.

Main Methods:

• Twelve subjects underwent 1.5T and 3T MRI scans using 3D-QALAS and conventional T1w sequences.
• Synthetic T1w images were generated from R1, R2, and proton density maps using SyMRI.
• Brain parcellation was performed on both conventional and synthetic T1w images using NeuroQuant.

Main Results:

• High correlations were observed between conventional and synthetic T1w images (medians of 0.97 at 1.5T, 0.92 at 3T).
• Repeatability was high, with median coefficients of variation of 1.2% (1.5T) and 1.5%-4.4% (3T).
• Significant biases were noted between conventional and synthetic methods, and across field strengths.

Conclusions:

• qMRI data can be used to synthesize T1w image stacks for automatic brain parcellation.
• Further investigation into synthetic parameter settings is needed to minimize observed biases.