Automatic 4-D registration in dynamic MR renography based on over-complete dyadic wavelet and Fourier transforms

Area of Science:

• Medical Imaging
• Radiology
• Image Processing

Background:

• Dynamic contrast-enhanced 4-D MR renography offers significant clinical potential for kidney assessment.
• Respiratory motion, including out-of-plane translation and rotation, severely limits the analysis and interpretation of these 4-D MR images.
• Manual image registration is impractical due to the large number of serial 3-D images generated per examination.

Purpose of the Study:

• To develop and evaluate a novel, robust, and automated technique for correcting out-of-plane translation and rotation in 4-D dynamic MR images.
• To achieve sub-voxel accuracy in motion correction for improved image analysis.

Main Methods:

• A new automated registration method was developed to address out-of-plane motion and rotation in 4-D dynamic MR datasets.
• The technique was validated using simulated motion data derived from clinical patient data.
• Performance was further assessed on 24 clinical kidney datasets and compared against a mutual information-based registration method.

Main Results:

• The automated method demonstrated high accuracy in removing out-of-plane translation and rotation.
• Comparison with manual co-registration showed good agreement in time-intensity curves.
• The novel technique achieved sub-voxel accuracy, outperforming existing methods.

Conclusions:

• The presented automated technique effectively corrects for out-of-plane motion and rotation in 4-D dynamic MR renography.
• This method significantly enhances the potential for broad clinical applications of dynamic contrast-enhanced 4-D MR renography.
• The sub-voxel accuracy and robustness of the technique offer a valuable solution for motion artifacts in abdominal MR imaging.