基于正弦图融合的CT金属伪影校正算法研究OACSTPCD

Objective To propose a metal artifact correction algorithm based on sinusoidal graph fusion to solve the problem of transition discontinuity between repaired projection data and surrounding projection data.Methods By preprocessing and K-means clustering technology,organizations with the same spatial information were gathered together to generate prior images,and the corrected projection data were obtained by correcting the original image projections based on the projection differences between the metal regions and the a priori images.Finally,the filtered backprojection algorithm was used to reconstruct the corrected CT image.Results In the verification of CT simulation data,the peak signal-to-noise ratio of fusion prior-based metal artifact reduction(FP-MAR)algorithm based on prior interpolation in single-metal correction and multi-metal correction were 0.943 and 0.915,respectively.Compared with linear interpolation based MAR(LI-MAR)algorithm,it increased 28.65% and 44.55%,respectively.The structural similarity of FP-MAR algorithm in single-metal correction and multi-metal correction were 0.984 and 0.961,which were 48.41% and 64.27% higher than that of LI-MAR algorithm.In clinical CT artifact image validation,the subjective evaluation of FP-MAR-corrected CT metal artifacts was higher than that of LI-MAR-corrected CT metal artifact images,and the difference was statistically significant.Conclusion The algorithm proposed in this study can effectively solve the transition discontinuity between the repaired projection data and the surrounding projection data,and better preserve the information near the metal structure.