Validation of motion correction techniques for liver CT perfusion studies.

Area of Science:

• Medical Imaging
• Radiology
• Image Processing

Background:

• Motion artifacts in computed tomography perfusion (CTp) imaging can compromise data accuracy.
• Image registration is crucial for correcting anatomical misalignment in CTp datasets.
• Validation of different registration techniques is necessary for reliable CTp analysis.

Purpose of the Study:

• To compare the effectiveness of manual, rigid, and non-rigid image registration techniques.
• To assess the ability of these methods to correct anatomical misalignment in liver CTp data.
• To determine the optimal registration strategy for improving CTp data quality.

Main Methods:

• 17 liver CTp datasets from patients with liver tumors were analyzed.
• Semi-automated rigid and non-rigid, plus manual, registration techniques were applied to align Phase 2 images to a Phase 1 reference.
• Four independent observers assessed alignment using a semi-quantitative visual scale, with statistical comparison via ordinal probit regression.

Main Results:

• A total of 306 registrations and 2448 observer scores were evaluated.
• Significant differences were found among the three registration techniques (p=0.03).
• Semi-automated techniques outperformed manual registration; non-rigid was superior to rigid (p<0.0001), and rigid was superior to manual (p=0.04).

Conclusions:

• Semi-automated registration, particularly non-rigid, provides superior liver anatomy alignment compared to manual methods.
• Improved anatomical alignment is expected to lead to more dependable CTp analyses.
• This study validates advanced registration techniques for enhancing CTp imaging accuracy.