Joint Cone-Beam CT Reconstruction and Rigid Motion Compensation Using A Differentiable Projector

Area of Science:

• Medical Imaging
• Image Reconstruction
• Computational Science

Background:

• Cone Beam Computed Tomography (CBCT) is crucial for medical imaging, but motion artifacts degrade image quality.
• Existing reconstruction methods struggle to accurately compensate for complex motion during scanning.
• Model-based iterative reconstruction (MBIR) offers potential but requires robust motion estimation.

Purpose of the Study:

• To develop and assess a feasible method for integrating model-based CBCT reconstruction with rigid motion estimation.
• To utilize a projection-domain data fidelity objective and a differentiable forward projector for joint optimization.
• To evaluate the effectiveness of the proposed framework in reducing motion artifacts.

Main Methods:

• A joint image reconstruction and motion estimation framework was developed, integrating motion compensation into MBIR.
• Six degrees of freedom (DoF) rigid motion was modeled as geometric perturbations.
• A differentiable forward projector with analytical gradients enabled efficient backpropagation for simultaneous optimization of motion parameters and attenuation volume.

Main Results:

• The proposed motion compensation method significantly enhanced image quality across various motion amplitudes.
• In simulations, Structural Similarity Index (SSIM) improved by up to 48% for increasing motion.
• Physical phantom studies demonstrated restored sharpness and anatomical continuity, confirming efficacy.

Conclusions:

• The discrepancy between estimated and acquired projections is sufficient for joint image and motion parameter estimation.
• This approach potentially eliminates the need for image-based sharpness criteria in motion-compensated reconstruction.
• The method effectively recovers anatomical structures with minimal residual errors, even under significant motion.