Devising a novel evaluation method for computed tomography images containing metal artifacts from titanium seed...

Area of Science:

• Medical Imaging
• Radiology
• Image Processing

Background:

• Metal artifacts in CT scans hinder image quality, particularly with metallic implants.
• Existing metal artifact reduction (MAR) techniques are insufficient for complete artifact elimination.
• Optimizing acquisition parameters and post-processing is crucial for improving CT image fidelity.

Purpose of the Study:

• To develop a novel evaluation method for CT images with metal artifacts from titanium seed implants.
• To identify the optimal virtual monochromatic imaging (VMI) energy level for artifact reduction and signal enhancement.
• To assess the efficacy of deep learning (DL) and MAR algorithms in conjunction with VMI.

Main Methods:

• A phantom simulating post-brachytherapy pelvic CT scans with radioactive seeds was utilized.
• Dual-energy CT was employed to investigate monochromatic energy levels from 35-200 keV.
• Artifacts were quantified using the Gumbel method; signal detectability was assessed via contrast-to-noise ratio (CNR) and a new contrast-to-artifact ratio (CAR).

Main Results:

• The lowest artifact index was observed at 65 keV.
• Optimal signal detectability, measured by CNR and CAR, occurred at 70 keV and 65 keV, respectively.
• VMI at 65 keV demonstrated the best balance between artifact reduction and signal detection.

Conclusions:

• Virtual monochromatic imaging (VMI) at 65 keV, utilizing deep learning (DL) and metal artifact reduction (MAR) reconstructions, is optimal for reducing metal artifacts.
• The newly developed contrast-to-artifact ratio (CAR) effectively evaluates images affected by metal artifacts.
• This approach enhances signal detectability in CT images with metallic implants.