Research In CT Image Metal Artifact Reduction Based On Hybrid Interpolation

Computer tomography(CT) is the most important support method of medical diagnosis nowadays.It measures the cumulative attenuation coefficient(or projector) of X-rays in all directions through the human body and calculated the distribution of X-ray attenuation coefficient by a computer,at last the image is displayed in order to help doctors diagnose the lesion.CT images can provide more high soft tissue resolution than ordinary X-ray images,and resolved the problem of overlapping three-dimensional structure,so make a big significance.The artifact is one of the important factors which impact of CT image quality and result the CT image useless for diagnosis.Metal as a high density of objects,have a higher attenuation coefficient than the other organizations of the human body,thus the X-ray passing it rapid decay, result the corresponding projection data distortion,artifact will appear on reconstruction image.These artifacts caused by metal implants are referred to as the metal artifacts.Because of its impact on clinical diagnosis as well as the diversification of manifestations,Metal artifacts are an important focus in current CT imaging studiesAt present,the elimination of the metal artifact method is mostly manipulate as the following:first of all obtain damaged original projection data by the filter back projection algorithm(Filtered Back Projection);then divide metal objects by segmentation algorithm,namely searches the exact boundary of metal object by border search algorithm in the projection data;followed fill the deviation of projection data by the application of interpolation algorithm;Finally reconstruct the date by FBP algorithm.Linear interpolation is the most commonly used interpolation methods.Because of the traditional algorithm itself,not only fail to accurately outline the borders of metal objects,can not restore a good metallic appearance.This is caused by a single interpolation methods and segmentation algorithm:though linear interpolation is most effective for the artifact removal strip,its inadequacy is at fault will occur has long stripes,and absorption Change the rate of severe regional edge and metal objects simultaneously tangent to the boundary,this projection value at both ends exactly at the point gap when the most obvious dramatic changes,which will affect the entire image quality.Polynomial interpolation and Spline interpolation makes image smoothness because of its interpolation characteristics which can be very good information on metal internal reflection.But at the same time,these methods are wake at smoothing strip which contained in metal artifacts.So a single interpolation method can not completely meet the clinical requirements.In this paper,the image is divided to two parts,one is metal and another is artifact.The main point is the metal itself four times a polynomial interpolation,while for radiation stripe metal interpolation artifacts.First of all mean-shift algorithm is been used for segmentation,through the division,the image is divided into the metal part and other parts.Because non-metal part of the image contained in the stripe artifacts,therefore,this part can also be used as part of artifacts.In this paper,the segmentation is done in projection images.Then,put the two separate parts projection separately transformed into sinogram space.Using the above algorithm to string map interpolation,that is,metal parts to do four polynomial interpolation,artifacts in part (non-metallic part) to do linear interpolation.Finally,put the sum of these two parts and make back-projection has to get corrected images.Through the phantom images and clinical experiments,one can see the mixed interpolation method combined with linear interpolation and four polynomial interpolation which can fully eliminate the radiation and smooth transition images,more successful.