Dual-Energy Cone-Beam CT Using Two Orthogonal Projection Views: A Phantom Study

Area of Science:

• Medical Imaging
• Computational Imaging
• Radiotherapy Physics

Background:

• Dual-energy CT (DECT) provides material decomposition capabilities crucial for radiation therapy.
• Traditional DECT methods often require extensive data acquisition, leading to longer scan times and higher radiation doses.
• Developing efficient and accurate DECT reconstruction techniques is essential for improving image quality and patient safety.

Purpose of the Study:

• To develop a novel imaging and reconstruction framework for dual-energy cone-beam CT (DECT) using only two orthogonal X-ray projections (2V-DECT).
• To enable fast and low-dose DECT volumetric imaging with high spectral fidelity and structural accuracy for DECT-guided radiation therapy.
• To demonstrate the feasibility of using physics-informed diffusion models for sparse-view DECT reconstruction.

Main Methods:

• A physics-informed dual-domain diffusion model framework was developed for 2V-DECT reconstruction.
• A cycle-domain training strategy was employed with a differentiable physics-informed module to ensure projection-volume consistency.
• A spectral-consistency loss was introduced to maintain inter-energy contrast during image generation.
• The model was trained and validated using 4D XCAT phantom data simulating realistic anatomical motion.

Main Results:

• The proposed framework successfully reconstructed high-fidelity DECT volumes from only two views.
• The method accurately preserved anatomical boundaries and effectively suppressed image artifacts.
• Subtraction maps derived from reconstructed energy volumes demonstrated strong agreement with ground truth data.
• The diffusion model approach achieved accurate structural and spectral recovery from extremely sparse projection data.

Conclusions:

• This study presents the first diffusion model-based framework for 2V-DECT reconstruction.
• The developed method enables accurate structural and spectral recovery from highly undersampled datasets.
• The framework holds significant potential for fast, low-dose DECT imaging in image-guided radiation therapy.