Improvement of brain segmentation accuracy by optimizing non-uniformity correction using N3.

Area of Science:

• Medical Imaging
• Neuroimaging
• Biomedical Engineering

Background:

• Artifactual intensity variations in MR images reduce automated brain segmentation accuracy.
• The N3 (nonparametric non-uniformity intensity normalization) method is widely used for correcting these variations.
• Previous studies suggest N3 performance can be improved on 3 T scanners with multichannel coils by optimizing its bias field smoothness parameter.

Purpose of the Study:

• To confirm and further investigate the optimization of the N3 method's bias field smoothness parameter.
• To evaluate the impact of reduced parameter values on brain structure measurements using FreeSurfer.
• To enhance the precision of neuroimaging analyses, particularly for cortical thickness estimation.

Main Methods:

• The study confirmed findings that optimizing the N3 bias field smoothness parameter improves performance.
• The research demonstrated benefits by reducing the smoothness parameter values to 30-50 mm (default 200 mm).
• White matter surface estimation and cortical/subcortical structure measurements were performed using FreeSurfer.

Main Results:

• Optimizing the N3 method's smoothness parameter significantly improved accuracy.
• Reduced parameter values (30-50 mm) enhanced white matter surface estimation.
• Measurements of cortical and subcortical structures showed increased precision.

Conclusions:

• Reducing the N3 bias field smoothness parameter to 30-50 mm enhances automated brain segmentation and structure measurement accuracy.
• This optimization is particularly beneficial for studies requiring precise estimation of cerebral cortex thickness.
• The findings contribute to more reliable neuroimaging analyses and inferences in neurological research.