Tomographic image reconstruction via estimation of sparse unidirectional gradients

Area of Science:

• Medical Imaging
• Computational Mathematics
• Image Reconstruction

Background:

• Computed tomography (CT) image quality is continuously improved by new algorithms.
• Sparse solution theory is increasingly applied to CT reconstruction from undersampled data for low-dose imaging.
• Total Variation (TV) minimization is a popular method for CT image reconstruction.

Purpose of the Study:

• To propose a novel method for CT image reconstruction using sparse gradients estimation (SGE).
• To compare the SGE method with existing techniques like TV minimization.
• To evaluate the performance of SGE in reconstructing images from low-dose and undersampled CT data.

Main Methods:

• A new CT image reconstruction method, sparse gradients estimation (SGE), was developed.
• SGE estimates 1D gradients in four directions using an iterative reweighting algorithm.
• Numerical simulations with Shepp-Logan and brain phantoms were conducted under varying projection numbers and noise levels.

Main Results:

• The proposed SGE algorithm demonstrated higher accuracy than TV methods when sparsity conditions were met.
• SGE effectively preserved sharp edges in reconstructed images.
• Improved image quality was achieved with SGE using a limited number of projections and low measurement noise.

Conclusions:

• The SGE method offers a promising approach for high-quality CT image reconstruction from undersampled and low-dose data.
• SGE outperforms traditional TV methods in preserving image details and accuracy.
• This technique has the potential to reduce radiation exposure in CT scans without compromising diagnostic quality.