Metal artifact reduction for CT based on sinusoidal description

Computed tomography has played a key role in bone structure imaging for over two decades. However, when a metal implant is present in the sample, the reconstructions are seriously distorted by artifact, and no method has successfully met the clinical demands. This paper presents a new method for metal artifact reduction in Computed Tomography based on sinusoidal description with the concentration of clinical applications. A piece of pig's leg with a lead nail placed inside the bone was scanned, generating 224 slices, in 177 of which the metal implant was present. The method includes detection of the correspondence of metal implants, fitting, amendment, and reconstruction based on sinusoidal description. Simulation and statistical error analysis show that the method improves PSNR (Peak Signal-to-Noise Ratio). A 3D modeling based on the reconstruction using the sinusoidal amendment method for a real case demonstrates that most of the metal artifact has been removed, which is compared with that based on the default output of the scanner. Metal artifacts in CT can be reduced effectively by the method based on the sinusoidal description, which isolates the correspondence of a metal implant from the original projection, so that a high quality reconstruction can be obtained.