Self-Calibration of Cone-Beam CT Geometry Using 3D-2D Image Registration: Development and Application to Task-Based...

Area of Science:

• Medical Imaging
• Robotics
• Computer Vision

Background:

• Robotic C-arm systems offer flexible imaging trajectories for various medical tasks.
• Accurate geometric calibration is crucial for reconstructing images from these systems, especially with non-circular orbits.
• Existing calibration methods can be challenging or irreproducible for arbitrary C-arm geometries.

Purpose of the Study:

• To propose and validate a unique geometric self-calibration method for arbitrary robotic C-arm orbits.
• To determine system geometry by registering 2D projections to a 3D image.
• To enable accurate 3D imaging in cone-beam CT (CBCT) systems with non-standard geometries.

Main Methods:

• A cone-beam CT (CBCT) bench system and a robotic C-arm were used with three phantoms.
• A 3D-2D registration process computed the 9 degree of freedom (DOF) transformation between projections and a 3D volume.
• Normalized gradient information was maximized using digitally reconstructed radiographs (DRRs) for calibration.

Main Results:

• Self-calibration produced CBCT images with spatial resolution statistically indistinguishable from standard calibration methods (p > 0.05).
• Differences between self-calibrated and standard-calibrated CBCT images were minimal (order of 10^-3 mm^-1).
• Maximum error in magnification was 3.2%, and back-projection ray placement error was within 0.5 mm.

Conclusions:

• The proposed geometric self-calibration method is effective for 3D imaging on general non-circular orbits in CBCT.
• This technique is valuable for systems lacking available or reproducible geometric calibration.
• The method supports the development of advanced, task-based 3D imaging for robotic C-arms.