Correcting scaling errors in tomographic images using a nine degree of freedom registration algorithm

Area of Science:

• Medical Imaging
• Image Registration
• Magnetic Resonance Imaging

Background:

• Clinical imaging systems, particularly MR scanners, often exhibit voxel dimension errors.
• Accurate voxel dimensions are crucial for precise medical image analysis and interpretation.

Purpose of the Study:

• To evaluate a nine-degree-of-freedom registration algorithm that maximizes mutual information for detecting and correcting scaling errors in MR imaging.
• To compare the efficacy of patient-specific scaling versus phantom-based scaling for reducing voxel dimension errors.

Main Methods:

• The study employed a mutual information-based registration algorithm with nine degrees of freedom.
• Two scaling methods were evaluated: patient scaling (using MR and CT images) and phantom scaling (using MR images and a phantom model).
• Validation involved assessing the fiducial registration error (FRE) using bone-implanted markers localized intraoperatively.

Main Results:

• Both patient and phantom scaling methods significantly reduced the average FRE in MR to CT and MR to physical space registration (p < 0.05).
• The phantom scaling method demonstrated a greater reduction in voxel scaling errors compared to patient scaling.
• This indicates successful reduction of scaling errors in MR imaging through the evaluated registration methods.

Conclusions:

• A nine-degree-of-freedom registration algorithm maximizing mutual information effectively reduces scaling errors in MR imaging.
• Phantom-based scaling offers a more robust approach for correcting voxel dimension inaccuracies in clinical MR data.
• The findings support the integration of such algorithms to enhance the precision of clinical imaging systems.