Nonlinearly scaled prior image-controlled frequency split for high-frequency metal artifact reduction in computed...

Area of Science:

• Medical Imaging
• Radiology
• Image Processing

Background:

• Metal implants in computed tomography (CT) scans cause high-frequency artifacts, degrading image quality and hindering diagnosis.
• Existing metal artifact reduction (MAR) techniques often struggle to preserve anatomical details while effectively removing these artifacts.

Purpose of the Study:

• To introduce and evaluate a novel approach for dedicated reduction of high-frequency metal artifacts in CT images.
• To preserve edge information and anatomical detail by incorporating prior image information into the artifact reduction process.

Main Methods:

• A nonlinear scaling (NLS) transfer function was applied in the sinogram domain to suppress high-frequency streak artifacts specifically in metal projections.
• Prior image information from tissue classification was used to exclude anatomical regions from scaling, ensuring detail preservation.
• The corrected high-frequency sinogram was reconstructed and combined with the low-frequency component of a normalized metal artifact reduction (NMAR) image.

Main Results:

• The prior image-controlled NLS method successfully removed streak artifacts while preserving anatomical detail in bone and soft tissue.
• Qualitative analysis of clinical datasets demonstrated significant enhancement in images with dental fillings, neuro coils, spinal screws, and hip implants.
• Phantom studies confirmed the effectiveness, showing the lowest Hounsfield unit (HU) deviation and best data visualization with NLS correction.

Conclusions:

• Prior image-controlled nonlinear scaling (NLS) offers an effective method for reducing high-frequency streaks in metal-corrupted CT data.
• This technique enhances diagnostic accuracy by improving the visualization of anatomical structures adjacent to metal implants.