[Sinogram restoration for low-dose cerebral perfusion CT images]

Area of Science:

• Medical Imaging
• Radiology
• Computational Imaging

Background:

• Cerebral perfusion CT (CPCT) imaging requires high radiation doses for adequate image quality.
• Reducing radiation dose in CPCT by lowering scanning current leads to significant image noise and artifacts, compromising clinical diagnosis.
• Existing noise reduction methods often degrade image details crucial for accurate perfusion analysis.

Purpose of the Study:

• To develop and evaluate a novel penalized weighted least-square (PWLS) method for restoring projection data in low-dose cerebral perfusion CT.
• To improve image quality, reduce radiation exposure, and enhance diagnostic accuracy in CPCT examinations.
• To investigate the effectiveness of PWLS in noise reduction and artifact suppression while preserving anatomical details.

Main Methods:

• A penalized weighted least-square (PWLS) method was developed to recover projection data, incorporating statistical properties of brain perfusion CT data.
• The Gauss-Seidel (GS) iterative method was employed for solving the PWLS reconstruction.
• Adaptive weighting was implemented between original and restored projection data to optimize image quality.

Main Results:

• The PWLS-based sinogram restoration method demonstrated superior noise reduction compared to conventional techniques.
• Significant suppression of artifacts was observed, leading to clearer image reconstruction.
• The method effectively preserved anatomical edges and fine details, crucial for accurate perfusion mapping.
• Improved cerebral perfusion maps were generated from the restored low-dose CT data.

Conclusions:

• The proposed PWLS method offers a promising approach for high-quality low-dose cerebral perfusion CT imaging.
• This technique effectively balances radiation dose reduction with diagnostic image quality.
• PWLS-based sinogram restoration enhances the clinical utility of CPCT by providing clearer perfusion maps with reduced artifacts and noise.