Distortion correction of EPI data using multimodal nonrigid registration with an anisotropic regularization

Area of Science:

• Medical Imaging
• Neuroimaging
• Image Processing

Background:

• Echo-planar imaging (EPI) is susceptible to geometric and intensity distortions.
• Accurate image registration is crucial for analyzing MRI data, especially in multimodal studies.
• Existing methods may struggle with local distortions caused by rapid magnetic field changes.

Purpose of the Study:

• To present a novel strategy for correcting geometric and intensity distortions in EPI MRI data.
• To introduce an improved multimodal registration framework for estimating dense displacement fields.
• To develop a new quality measure for quantifying geometric distortions.

Main Methods:

• Utilized an improved multimodal registration framework with normalized mutual information (NMI) and multi-scale techniques.
• Employed a novel anisotropic regularization functional to ensure robustness in high-dimensional inverse problems.
• Introduced standardized contour distance (SCD) using outer structure shape (OSS) information to quantify geometric distortions.

Main Results:

• The new registration method achieved results comparable to state-of-the-art techniques in monomodal phantom data.
• In multimodal human brain data, the strategy significantly improved the mean SCD.
• Achieved mean SCD improvements from 0.96±0.11 to 0.60±0.13 for SPM Subject data and 0.92±0.07 to 0.78±0.11 for BrainSuite data.

Conclusions:

• The proposed method effectively corrects geometric and intensity distortions in EPI MRI.
• The novel registration framework and SCD measure enhance the accuracy and reliability of neuroimaging analysis.
• This approach offers a robust solution for improving the quality of multimodal brain MRI data.