Physics-Informed Score-Based Diffusion Model for Limited-Angle Reconstruction of Cardiac Computed Tomography

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Science

Background:

• Cardiac computed tomography (CT) is crucial for cardiovascular disease diagnosis and monitoring.
• High temporal resolution in cardiac CT is vital for accurate diagnosis.
• Limited-angle data acquisition improves scan time and temporal resolution but causes image degradation, necessitating advanced reconstruction methods.

Purpose of the Study:

• To introduce a novel physics-informed score-based diffusion model (PSDM) for reconstructing high-quality cardiac CT images from limited-angle data.
• To enhance image quality by combining diffusion model priors with iterative algorithm and Fourier fusion techniques.

Main Methods:

• Developed a novel physics-informed score-based diffusion model (PSDM).
• Integrated the primal-dual hybrid gradient (PDHG) algorithm with score-based diffusion models.
• Employed a combination of data prior from a diffusion model and a model prior from iterative algorithm and Fourier fusion at the sampling time.

Main Results:

• Successfully reconstructed high-quality cardiac CT images from limited-angle data using the proposed PSDM.
• Demonstrated the effectiveness of the approach through numerical simulations and real data experiments.
• Achieved improved image quality compared to conventional methods when dealing with limited-angle data.

Conclusions:

• The proposed physics-informed score-based diffusion model (PSDM) effectively reconstructs high-quality cardiac CT images from limited-angle data.
• The integration of PDHG algorithm and diffusion models offers a promising solution for overcoming image degradation in limited-angle cardiac CT.
• This approach holds potential for improving diagnostic accuracy in cardiovascular imaging.