Total variation norm for three-dimensional iterative reconstruction in limited view angle tomography.

Area of Science:

• Medical Imaging
• Image Reconstruction
• Computational Imaging

Background:

• Limited view angle tomography (ectomography) often suffers from artifacts and noise.
• Existing 2D filtered backprojection algorithms have limitations in addressing these issues.
• The 3D total variation (TV) norm is known for noise reduction and edge preservation in 2D imaging.

Purpose of the Study:

• To develop and evaluate an iterative Bayesian reconstruction algorithm for limited view angle tomography.
• To incorporate the 3D total variation (TV) norm into an expectation maximization (EM) algorithm for improved ectomography.
• To assess the algorithm's effectiveness in reducing artifacts, noise, and improving depth resolution.

Main Methods:

• Developed an iterative Bayesian reconstruction algorithm using the 3D total variation (TV) norm.
• Integrated the TV norm into an ordered subsets expectation maximization (EM) algorithm, termed TV3D-EM.
• Evaluated TV3D-EM using modeled point spread functions and digital phantoms, comparing results to 2D filtered backprojection.

Main Results:

• TV3D-EM demonstrated a substantial reduction in artifacts specific to limited view angle geometry.
• Noise levels in the reconstructed images were significantly improved.
• Depth resolution was enhanced by at least 45% compared to the conventional algorithm.
• Overall perceived image quality was notably improved.

Conclusions:

• The developed TV3D-EM algorithm effectively addresses limitations in limited view angle tomography.
• The use of the 3D TV norm provides superior image quality by reducing artifacts and noise.
• This advanced reconstruction technique offers significant improvements in depth resolution and diagnostic potential.