Removal of streaking artifact in computed tomography

Area of Science:

• Medical Imaging
• Computerized Tomography
• Image Processing

Background:

• Transmission computed tomography (CT) reconstructs internal body structures from X-ray attenuation data.
• Dense materials (e.g., surgical clips) cause artifacts (streaks) that obscure diagnostic information in CT images.
• Current methods may render CT scans unusable due to these artifacts.

Purpose of the Study:

• To develop and evaluate a novel technique for removing imaging artifacts in CT scans.
• To address the loss of diagnostic information caused by high-density streaks.
• To improve the utility of CT scans in the presence of metallic or dense objects.

Main Methods:

• A new artifact removal technique was developed, treating affected projection data as misinformation.
• The technique involves assigning new values to erroneous data before image reconstruction.
• Monte Carlo simulation generated projection data for a simulated head section with a lead fragment.
• Three methods for generating replacement data were investigated, including nearest-neighbor pattern recognition.

Main Results:

• The developed technique successfully removed high-density streaks from simulated CT projection data.
• Nearest-neighbor pattern recognition proved to be an effective method for generating replacement data.
• Image reconstruction was performed without requiring changes to existing hardware or software.

Conclusions:

• The proposed technique offers a viable solution for mitigating artifacts in CT imaging.
• Nearest-neighbor pattern recognition is a promising approach for correcting projection data in artifact-affected CT scans.
• This method enhances diagnostic accuracy by preserving internal structure information in challenging cases.