Structure-guided computed tomography reconstruction from limited-angle projections.

Area of Science:

• Medical Imaging
• Image Reconstruction
• Computational Imaging

Background:

• Limited-angle computed tomography (CT) poses challenges in image reconstruction, often resulting in artifacts and degraded structures.
• Existing methods utilizing image structure priors have improved quality but still struggle with residual artifacts.
• The need for advanced reconstruction techniques to enhance image fidelity in limited-angle CT remains critical.

Purpose of the Study:

• To develop a novel reconstruction model for limited-angle CT that minimizes artifacts and preserves image structures.
• To introduce a weighted relative structure (wRS) approach guided by image gradients.
• To validate the efficacy of the proposed model against established reconstruction methods.

Main Methods:

• A new reconstruction model based on weighted relative structure (wRS) derived from image gradients was established.
• An efficient algorithm employing a surrogate function was developed to solve the proposed model.
• The method was rigorously compared with anisotropic total variation and L0 norm minimization techniques.

Main Results:

• Experiments on digital phantoms and real-world data (cheese, walnut) demonstrated the superiority of the wRS method.
• Quantitative metrics (RMSE, PSNR, SSIM) for the FORBILD head phantom (90°-data) were 0.0120, 43.52, and 0.9961, respectively.
• Reconstructed images showed fewer artifacts and less structure degradation compared to other methods, closely matching reference images.

Conclusions:

• The proposed weighted relative structure (wRS) model significantly enhances image quality in limited-angle CT reconstruction.
• The developed algorithm provides an efficient solution for artifact reduction and structure preservation.
• This method represents a substantial advancement for limited-angle CT imaging applications.