Robust H-K Curvature Map Matching for Patient-to-CT Registration in Neurosurgical Navigation Systems

Area of Science:

• Medical imaging
• Computer-assisted surgery
• Geometric registration

Background:

• Image-to-patient registration aligns real patients with medical images (e.g., CT) for surgical guidance.
• Markerless registration methods use patient scan data and 3D CT data.
• Iterative Closest Point (ICP) algorithms require accurate initial locations to avoid convergence issues and local minima.

Purpose of the Study:

• To develop an automatic and robust 3D data registration method for improved image-to-patient registration.
• To address the limitations of conventional ICP algorithms by finding a proper initial location.
• To enhance the accuracy and efficiency of surgical navigation using medical imaging.

Main Methods:

• A markerless approach utilizing patient scan data and 3D CT data.
• Conversion of 3D CT and scan data into 2D curvature images.
• Curvature matching to identify corresponding areas for robust 3D registration.
• Iterative Closest Point (ICP) algorithm for precise 3D alignment after initial localization.

Main Results:

• The proposed curvature matching method effectively finds accurate initial locations for ICP.
• Curvature features demonstrate robustness against translation, rotation, and deformation.
• Successful implementation of precise 3D registration using the enhanced ICP approach.

Conclusions:

• The developed method provides an automatic and robust solution for initial localization in image-to-patient registration.
• Curvature matching significantly improves the performance of ICP algorithms in medical applications.
• This technique enhances the utility of medical images during surgical procedures.