A fully 3D approach for metal artifact reduction in computed tomography

Area of Science:

• Medical Imaging
• Image Processing
• Computational Science

Background:

• Metal objects in computed tomography (CT) imaging cause significant data inconsistencies.
• These inconsistencies lead to artifacts like star-shaped or stripe-like patterns in reconstructed images.
• Reducing these metal artifacts is crucial for enhancing diagnostic image quality.

Purpose of the Study:

• To propose a novel metal artifact reduction (MAR) approach for computed tomography (CT) imaging.
• The method aims to reduce image artifacts caused by metal objects.
• This is achieved by recomputing inconsistent projection data using Fourier-based interpolation.

Main Methods:

• The core of the MAR strategy involves recomputing inconsistent projection data using nonequispaced fast Fourier transforms (NFFT).
• NFFT interpolation is applied in one, two, and three dimensions, overcoming challenges with irregular grids.
• The method allows for incorporating prior knowledge through transform damping and is applicable to various CT data types.

Main Results:

• The study presents results from MAR using one-, two-, and three-dimensional NFFT interpolations.
• Performance is evaluated by comparing results in both projection data and spatial domains.
• Comparisons are made against the established one-dimensional linear interpolation technique.

Conclusions:

• Increasing the dimensionality of NFFT interpolation significantly enhances image quality.
• Incorporating more spatial information into the recomputation process is key for effective artifact reduction.
• The proposed NFFT-based MAR method offers considerable improvements in CT image quality.