Computing global minimizers to a constrained B-spline image registration problem from optimal l1 perturbations to...

Area of Science:

• Medical imaging
• Image registration
• Computational anatomy

Background:

• Block matching is a robust image registration technique but lacks spatial accuracy.
• B-spline registration offers global smoothness but can be sensitive to noise and large displacements.
• A gap exists in combining the strengths of both methods for improved medical image analysis.

Purpose of the Study:

• To develop a novel numerical method for constrained B-spline registration.
• To integrate the robustness of block matching with the global smoothness of B-splines.
• To achieve high spatial accuracy in medical image registration.

Main Methods:

• Reformulated B-spline registration as a basis pursuit problem.
• Used block matching for optimization involving image data.
• Identified a voxel subset for global B-spline minimization.

Main Results:

• Achieved high spatial accuracy in registration of thoracic CT image pairs.
• Demonstrated performance on complex image datasets (COPDgene).
• Validated against expert-determined landmark points.

Conclusions:

• The proposed methodology achieves high spatial accuracy.
• The framework is generalizable to various displacement field parameterizations.
• It allows flexible combinations of block matching metrics and physical modeling.