Prior-image-based CT reconstruction using attenuation-mismatched priors.

Area of Science:

• Medical Imaging
• Computed Tomography
• Image Reconstruction

Background:

• Prior-image-based reconstruction (PIBR) enhances low-dose CT but struggles with attenuation mismatches between prior and current images due to varying scanner or beam qualities.
• Attenuation mismatches, alongside anatomical changes, pose significant challenges to the accuracy of PIBR methods, potentially introducing artifacts.

Purpose of the Study:

• To investigate and enhance the statistical image reconstruction using normal-dose image-induced nonlocal means regularization (SIR-ndiNLM) for prior-image-based reconstruction (PIBR) in low-dose CT.
• To address the challenges posed by attenuation-mismatched priors in PIBR and enable quantitative low-dose CT imaging.
• To develop and validate corrective schemes for SIR-ndiNLM to handle attenuation and anatomical differences between prior and current images.

Main Methods:

• Proposed two corrective schemes for SIR-ndiNLM: a global histogram-matching approach and a local attenuation correction approach.
• Validated the schemes using dual-energy CT data to simulate attenuation mismatches and different CT slices for anatomical changes.
• Evaluated the performance of the original SIR-ndiNLM and the proposed corrective schemes in the presence of attenuation and anatomical variations.

Main Results:

• The original SIR-ndiNLM method produced artifacts when using attenuation-mismatched priors, with artifact severity increasing with greater mismatch.
• The two proposed corrective schemes effectively handled both attenuation mismatches and anatomical changes, successfully eliminating artifacts.
• The enhanced SIR-ndiNLM demonstrated robust performance in quantitative low-dose CT reconstruction from low-flux and sparse-view data.

Conclusions:

• The proposed corrective schemes significantly improve the robustness of SIR-ndiNLM for PIBR, enabling reliable low-dose CT reconstruction even with attenuation-mismatched priors.
• This work allows for the effective utilization of priors acquired with different beam settings, enhancing the applicability of PIBR.
• The developed methods facilitate quantitative low-dose CT imaging in scenarios with varying acquisition parameters, improving diagnostic accuracy and reducing radiation exposure.