Usefulness of a Metal Artifact Reduction Algorithm in Digital Tomosynthesis Using a Combination of Hybrid Generative...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Metal artifacts pose a significant challenge in digital tomosynthesis, degrading image quality.
• Reducing radiation dose is crucial for patient safety in medical imaging procedures.

Purpose of the Study:

• To develop and evaluate a novel hybrid generative adversarial network (CGpM-MAR) for metal artifact reduction (MAR) and radiation dose reduction in digital tomosynthesis.
• To compare the performance of CGpM-MAR against conventional methods like filtered back projection (FBP) and convolutional neural network MAR.

Main Methods:

• A hybrid generative adversarial network (GAN) combining cycle-consistent GAN, pix2pix, and mask pyramid network (MPN) was developed (CGpM-MAR).
• The algorithm was tested using projection data and a prosthesis phantom at various radiation doses.
• Metal artifact reduction was quantified using the artifact index (AI) and Gumbel distribution analysis.

Main Results:

• CGpM-MAR demonstrated effective metal artifact reduction, achieving good performance in terms of AI.
• The algorithm successfully reduced artifacts by 55% radiation dose, independent of metal type.
• CGpM-MAR showed superior MAR compared to conventional algorithms at reduced radiation doses, closely resembling reference FBP images.

Conclusions:

• The novel CGpM-MAR algorithm is a promising method for reducing metal artifacts and radiation dose in digital tomosynthesis.
• This approach offers significant potential for improving clinical practice by enhancing image quality and patient safety.