Frequency split metal artifact reduction (FSMAR) in computed tomography.

Area of Science:

• Medical Imaging
• Computed Tomography
• Image Processing

Background:

• Metal artifacts significantly degrade CT image quality and diagnostic value.
• Existing metal artifact reduction (MAR) methods often lack a universally accepted solution.
• Severe artifacts arise from metal implants within the CT measurement field.

Purpose of the Study:

• Introduce a novel frequency split metal artifact reduction (FSMAR) method.
• Achieve efficient MAR with high image quality and detail preservation near implants.
• Address the long-standing challenge of metal artifacts in CT imaging.

Main Methods:

• FSMAR combines raw data inpainting with an image-based frequency split approach.
• It integrates high frequencies from uncorrected images with low frequencies from corrected images.
• The method was evaluated with normalized MAR (NMAR) and standard inpainting MAR, using phantom data and patient scans (hip prostheses, dental fillings, neurocoil, spine fixation).

Main Results:

• FSMAR demonstrates sharp edges and superior preservation of anatomical details compared to inpainting-based MAR alone.
• The method effectively reduces metal artifacts without the typical blurring observed near implants.
• Quantitative evaluations on phantom and patient data confirmed its efficacy.

Conclusions:

• The combination of FSMAR with NMAR provides accurate correction for both high and low frequencies.
• FSMAR is computationally inexpensive and suitable for clinical routine application.
• The algorithm requires no manual parameter tuning, facilitating immediate clinical integration.