Model-based rigid and nonrigid volumetric image registration for image-guided interventions

Area of Science:

• Neurosurgery
• Medical Imaging
• Computational Anatomy

Background:

• Accurate image registration is critical for image-guided neuro interventions.
• Misalignments in preoperative and intraoperative scans can compromise navigation and patient safety.

Purpose of the Study:

• Introduce a model-based framework for rigid and nonrigid volumetric brain image registration.
• Establish anatomical point-based correspondence using shape-constrained deformable brain segmentation.
• Provide an alternative to conventional image-intensity-based registration methods.

Main Methods:

• Registration relies solely on anatomical geometry, eliminating image intensity dependence.
• Rigid registration aligns centroids and estimates transformations between segmented meshes.
• Nonrigid registration fits B-spline surfaces to mesh vertices for smooth deformation fields.

Main Results:

• Quantitative validation using synthetic MR scans demonstrated high accuracy.
• Rigid registration performed comparably to an FDA-cleared approach.
• Nonrigid registration effectively captured realistic brain deformations like brain shift.
• Both methods showed high accuracy and computational efficiency.

Conclusions:

• The proposed framework offers a robust, anatomically driven registration alternative.
• Eliminates reliance on image intensity, improving registration accuracy.
• Shows strong clinical potential for enhancing precision, safety, and outcomes in interventional workflows.