SPARE: Symmetrized Point-to-Plane Distance for Robust Non-Rigid 3D Registration.

Area of Science:

• Medical image analysis
• Computer vision
• Computational geometry

Background:

• Current non-rigid registration methods often rely on point-to-point or point-to-plane distances, leading to slow convergence and loss of anatomical detail.
• These traditional metrics can struggle with accurate geometric approximation, hindering precise alignment in medical imaging.

Purpose of the Study:

• To develop a novel and robust non-rigid registration formulation for enhanced accuracy and efficiency.
• To improve the approximation of underlying geometry during the registration process.

Main Methods:

• Proposed SPARE (Symmetrized Point-to-plane distance for Accurate Registration) utilizing a novel symmetrized point-to-plane distance metric.
• Incorporated an as-rigid-as-possible regularization term for estimating deformed normals.
• Employed an alternating minimization solver with a majorization-minimization strategy and deformation graph-based coarse alignment for initialization.

Main Results:

• The symmetrized point-to-plane distance, leveraging both positions and normals, provides a more accurate geometric approximation.
• SPARE demonstrates significantly improved accuracy in non-rigid registration tasks compared to existing methods.
• The method maintains high solution efficiency, validated through extensive experimental evaluations.

Conclusions:

• SPARE offers a superior approach to non-rigid registration, achieving higher accuracy through a refined distance metric.
• The proposed optimization strategy and initialization enhance both the quality and efficiency of the registration process.
• The publicly available code facilitates further research and application in medical image registration.