Quantitative Evaluation of Intensity Inhomogeneity Correction Methods for Structural MR Brain Images

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Biomedical Engineering

Background:

• Intensity non-uniformity (INU) correction is critical for reliable magnetic resonance (MR) imaging, impacting both within- and across-subject analyses.
• T1-weighted MR images are standard for structural brain imaging, making INU correction techniques for these images particularly important.
• Existing software packages like FreeSurfer, BrainVoyager, SPM, and FSL offer integrated INU correction methods, but their comparative performance is not well-defined.

Purpose of the Study:

• To objectively compare the performance of INU correction techniques integrated into widely used MR data analysis software.
• To evaluate and optimize configurations for INU correction methods to achieve superior reconstruction accuracy.
• To provide guidance on selecting the most appropriate INU correction method based on specific MR data characteristics.

Main Methods:

• Utilized simulated MR data with controlled inhomogeneity magnitudes and noise levels to assess INU field reconstruction.
• Investigated INU correction methods within FreeSurfer, BrainVoyager, SPM, and FSL, evaluating a range of input parameters.
• Defined and compared 'enhanced' configurations against default settings to determine optimal performance parameters for each method.

Main Results:

• Enhanced configurations generally yielded more accurate INU field reconstructions compared to default settings across all evaluated methods.
• SPM and FSL, which combine INU correction with brain segmentation, demonstrated superior performance over FreeSurfer and BrainVoyager for most INU magnitudes and noise levels.
• FreeSurfer and BrainVoyager showed potential for accurate INU reconstruction under specific conditions: low noise for FreeSurfer and smooth inhomogeneity profiles for BrainVoyager.

Conclusions:

• The choice of INU correction method significantly impacts the reliability of structural brain imaging analysis.
• SPM and FSL offer robust INU correction, particularly beneficial when integrated with segmentation, while FreeSurfer and BrainVoyager have specific use cases.
• This comparative study provides valuable insights for researchers to select optimal INU correction strategies tailored to their MR data characteristics.