Online geometric calibration of cone-beam computed tomography for arbitrary imaging objects

Area of Science:

• Medical Imaging
• Computational Imaging
• Geometric Calibration

Background:

• Accurate geometric parameter estimation is crucial for high-quality image reconstruction in cone-beam computed tomography (CBCT).
• Traditional methods often require dedicated calibration phantoms and can be sensitive to noise and specific imaging geometries.
• There is a need for robust, phantom-less online calibration methods for CBCT systems.

Purpose of the Study:

• To propose and validate a novel online method for determining geometric parameters in circular trajectory CBCT.
• To assess the method's accuracy, noise sensitivity, and impact on reconstructed image quality.
• To demonstrate the applicability of the method to arbitrary imaging objects and compare its performance with offline techniques.

Main Methods:

• Development of an online method based on the symmetry property of the sum of projections (SOP).
• Deduction of an objective function dependent on geometric parameters, minimized to find their true values.
• Validation through numerical phantom studies with varying noise levels and small animal imaging experiments.

Main Results:

• The proposed method accurately determines detector skew, roll, and rotation axis without requiring a calibration phantom.
• The method demonstrates high insensitivity to noise in numerical simulations.
• Geometric parameter accuracy for mid-plane and source-to-detector distance is slightly lower but does not significantly degrade reconstructed image quality.
• Comparable image quality to offline methods was achieved in reconstructed images from small animal studies.

Conclusions:

• The novel online SOP symmetry-based method provides accurate geometric parameter estimation for CBCT, eliminating the need for calibration phantoms.
• The method is robust to noise and suitable for circular trajectory CBCT with arbitrary cone angles.
• The technique achieves high-quality image reconstruction, comparable to offline methods, and is applicable to diverse imaging objects.