Phase Vector Incompressible Registration Algorithm for Motion Estimation From Tagged Magnetic Resonance Images

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Computational Anatomy

Background:

• Tagged magnetic resonance imaging (MRI) is crucial for quantifying tissue motion and strain.
• Estimating 3D motion from 2D slices presents challenges due to acquisition time constraints and interpolation inaccuracies.

Purpose of the Study:

• To develop a novel phase-based 3D motion estimation technique for tagged MRI.
• To improve the accuracy, speed, and robustness of 3D tissue motion quantification.

Main Methods:

• Computed harmonic phase volumes from interpolated tagged MRI slices.
• Employed a diffeomorphic image registration framework with a novel symmetric similarity metric for harmonic phase volumes.
• Utilized harmonic magnitude to enforce tissue incompressibility.

Main Results:

• Achieved dense, incompressible, diffeomorphic, and inverse-consistent 3D motion fields at the voxel level.
• Demonstrated comparable accuracy to existing methods across diverse datasets (phantoms, brain, speech, cardiac).
• Exhibited robustness to tag fading and noise, with reduced computation time.

Conclusions:

• The proposed phase-based 3D motion estimation method offers a significant advancement in tagged MRI analysis.
• This technique provides accurate and efficient quantification of tissue deformation, applicable to various clinical and research scenarios.