医用X-CT图像中金属伪影校正算法的研究进展OA

Metal implants such as intracranial aneurysm embolisation,oral metal implants,spinal endoprostheses,and hip arthroplasty are increasingly used in clinical practice with the continuous progress of medical technology.However,the high-density metal implant absorbed a large number of X-ray photon counts resulting in missing projection data in the postoperative CT image review.Stripe artifacts appear on the reconstructed CT images,reducing the contrast of the images,making subtle anatomical structures poorly displayed,and even causing clinicians to make incorrect diagnoses.Interpolation algorithm to correct metal artifacts is to reduce the artifacts by estimating the missing data in the metal projection area.Iterative reconstruction technique is to approximate the real image step by step through several iterations of computation.Prior image correction algorithms use known metal object information or models to pre-process images to reduce the impact of metal artifacts.Deep learning algorithms are able to automatically identify and correct metal artifacts with high accuracy and efficiency by training neural network models.However,there are significant differences in the results demonstrated by different algorithms for metal artifact suppression.The interpolation algorithm is not effective for the suppression of polymetallic artifacts.Iterative reconstruction algorithms require multiple iterations to optimize the image reconstruction process,which can partially reduce the impact of metal artifacts,but may take longer to compute.Therefore,the application of interpolation algorithm,iterative reconstruction,prior image correction algorithm and deep learning algorithm to metal artifact correction is reviewed.