Robust methods for automatic image-to-world registration in cone-beam CT interventional guidance

Area of Science:

• Medical Imaging
• Surgical Navigation
• Image-to-World Registration

Background:

• Real-time surgical navigation requires precise image-to-world registration, aligning imaging and patient coordinate systems.
• Conventional manual registration is time-consuming, prone to errors, and shows variability between operators.
• This study explores automatic image-to-world registration using an automatic registration marker (ARM) with C-arm cone-beam CT (CBCT).

Purpose of the Study:

• To introduce and evaluate two novel automatic image-to-world registration methods: Known-Model and Free-Form.
• To compare the accuracy and reproducibility of these automatic methods against conventional manual registration.
• To assess the performance across various anatomical sites and under challenging conditions with interventional tools.

Main Methods:

• Developed a prototype C-arm for CBCT and a surgical tracking system.
• Designed an ARM with markers detectable in X-ray and infrared tracking.
• Implemented Known-Model (predefined ARM, 3D-2D registration) and Free-Form (individual marker localization, Hough transform, backprojection) methods.
• Tested robustness and accuracy across head, thorax, and abdomen, comparing with manual registration.

Main Results:

• Automatic methods showed comparable or superior accuracy (Target Registration Error - TRE) to manual registration within ~200 mm of C-arm isocenter.
• Marker localization was robust across all anatomical sites, even with interventional tools present.
• The Free-Form method demonstrated significantly improved reproducibility (0.22 mm vs. 0.30 mm) compared to manual registration.
• TRE increased with distance from isocenter, but performance was independent of anatomical site.

Conclusions:

• Automatic image-to-world registration methods offer potential for enhanced accuracy, reproducibility, and workflow efficiency in CBCT-guided surgery.
• The Free-Form method proved robust across anatomical sites, matching or exceeding manual registration accuracy.
• Both automatic methods, particularly the configurable Free-Form approach, show promise for advancing surgical navigation.