Blind deconvolution in model-based iterative reconstruction for CT using a normalized sparsity measure.

Area of Science:

• Medical Imaging
• Computational Imaging
• Image Reconstruction

Background:

• Model-based iterative reconstruction enhances CT image quality by incorporating noise statistics and physical forward models.
• Including system blur models in reconstruction can improve resolution but requires precise blur characterization, which is often difficult and can degrade image quality if inaccurate.

Purpose of the Study:

• To develop a novel objective function for jointly estimating system blur parameters during computed tomography (CT) reconstruction.
• To create a less blur-biased objective function using a normalized sparsity measure based on total-variation regularization.

Main Methods:

• Introduced a novel objective function designed for joint estimation of system blur parameters within tomographic reconstruction.
• Developed a solving strategy for simultaneously recovering low-dimensional blur parameters and performing CT reconstruction.
• Utilized a new normalized sparsity measure based on total-variation regularization, aiming to reduce bias related to blur.

Main Results:

• Extensive simulations demonstrated successful recovery of simulated blur parameters across various regularization strengths and system blurs.
• The proposed strategy was validated, showing the dependency of the objective function on the number of iterations.
• Experimental validation on human wrist phantom data showed good agreement between estimated and visually inspected blur parameters.

Conclusions:

• The novel objective function effectively enables joint estimation of system blur parameters during CT reconstruction, improving accuracy and image quality.
• The approach, validated through simulations and experiments, offers a robust method for handling system blur in CT imaging.
• Findings are applicable beyond attenuation-based CT and can aid in recovering more complex imaging model parameters.