Image reconstruction from incomplete data via approximated pseudo-inverse and dual-domain coupled diffusion posterior...

Area of Science:

• Medical Imaging
• Computational Imaging
• Artificial Intelligence

Background:

• Sparse-view (SV) and limited-angle (LA) computed tomography (CT) are crucial for reducing patient radiation dose.
• Data incompleteness in SV and LA CT causes severe artifacts, impacting diagnostic accuracy and clinical use.
• Developing methods to improve CT image quality under incomplete data is essential for clinical translation.

Purpose of the Study:

• To enhance CT image quality under incomplete-data conditions by suppressing artifacts while preserving anatomical details.
• To investigate an optimized dual-domain coupled diffusion posterior sampling (DPS) framework integrated with an approximated pseudo-inverse method.

Main Methods:

• Developed an approximated pseudo-inverse-guided dual-domain stochastic diffusion model (API-DDM).
• Coupled sinogram and image domains in the reverse diffusion stage using a stochastic process.
• Leveraged the approximated pseudo-inverse of the measurement matrix and a filtered back-projection operator to guide the diffusion process.

Main Results:

• The API-DDM method demonstrated superior reconstruction quality on public CT datasets for both SV (≤50 views) and LA (≤120°) conditions.
• Achieved high Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) values, e.g., PSNR of 41.21 dB and SSIM of 0.9527 for SV CT.
• Maintained robust performance even under extreme undersampling conditions (e.g., 9 views or 60° range).

Conclusions:

• The API-DDM algorithm effectively bridges physical measurement consistency and deep generative priors for incomplete CT data reconstruction.
• The method's robustness in SV and LA CT suggests potential clinical applications in low-dose screening and intraoperative imaging.
• This approach offers a promising solution for improving diagnostic reliability in radiation-sensitive CT imaging scenarios.