Two-Stage Sparse Angle CT Reconstruction Combining Group Sparsity and Relativity-of-Gaussian.

Area of Science:

• Medical Imaging
• Computational Imaging
• Image Reconstruction

Background:

• Sparse angle CT is an advanced imaging technology with image reconstruction as a key research area.
• Existing algorithms struggle with noise and artifacts in low-contrast regions.
• Group sparse regularization methods offer potential but require further refinement.

Purpose of the Study:

• To propose a novel two-stage CT reconstruction model, PLS-GSR-RoG, for sparse angle CT.
• To address noise and artifacts in low-contrast areas while preserving image details.
• To improve the overall performance of sparse angle CT image reconstruction.

Main Methods:

• Developed a dual regularization model combining Group Sparsity (GSR) and Relativity-of-Gaussian (RoG).
• Implemented a two-stage iterative reconstruction process: GSR and RoG in the first stage, GSR only in the second.
• Validated the model using FORBILD head, thoracic, and pelvic phantoms/images at varying projection angles.

Main Results:

• The PLS-GSR-RoG model significantly reduced noise and artifacts in low-contrast areas.
• Achieved superior image quality compared to SART-TV, SART-RTV, SART-GSR, and SART-GSR-WIGF.
• Reconstructed images showed high Peak Signal-to-Noise Ratio (PSNR) up to 48.03 dB and Feature Similarity Index (FSIM) up to 0.9996.

Conclusions:

• The two-stage PLS-GSR-RoG model effectively enhances sparse angle CT image reconstruction.
• Combining GSR and RoG in a staged iterative manner improves noise suppression and detail preservation.
• The proposed method offers a promising advancement for clinical applications of sparse angle CT.