Image reconstruction from few-view CT data by gradient-domain dictionary learning.

Area of Science:

• Medical Imaging
• Image Reconstruction
• Computational Science

Background:

• Reducing radiation dose in computed tomography (CT) imaging necessitates fewer projections.
• Incomplete projection data in CT reconstruction leads to artifacts and image distortions.

Purpose of the Study:

• To introduce a novel dictionary learning algorithm operating in the gradient-domain (Grad-DL) for few-view CT reconstruction.
• To enhance image quality and reduce artifacts in low-dose CT imaging.

Main Methods:

• The Grad-DL algorithm trains dictionaries from horizontal and vertical gradient images.
• Image reconstruction is performed by solving the least-square method using sparse representations of gradients.
• This approach leverages the inherent sparsity of gradient images for improved reconstruction.

Main Results:

• Qualitative and quantitative studies were conducted using computer simulations and real data.
• Experiments covered both fan-beam and cone-beam CT geometries.
• The Grad-DL algorithm demonstrated superior performance compared to existing methods.

Conclusions:

• The proposed Grad-DL algorithm effectively improves image quality in few-view CT reconstruction.
• This method offers a promising solution for reducing radiation dose while maintaining diagnostic image quality.