Development Of Technologies For Limited-view Artifact And Metal Artifact Correction In Oral Cone-beam CT Images

Oral cone-beam computed tomography technology can directly and quickly obtain threedimensional structural images of the human oral cavity.In order to reduce ionizing radiation damage to patients,the scanning angle is limited during oral cone-beam CT scanning,but limited angle artifacts appear in reconstructed CT images.Also,when metal implants are present in the patient’s mouth,metal artifacts can be generated in the reconstructed CT images.Both of these artifacts can seriously affect the diagnosis of clinical dentists.Therefore,the correlative study of correction of limited-angle reconstruction artifacts and metal artifacts has important clinical value in oral cavity.To correct the limited-view reconstruction artifacts,this paper proposes a limited-view cone-beam CT reconstruction algorithm based on cycle-consistent generative adversarial networks.The algorithm constructs a set of recurrent generative adversarial networks,including two sets of generative networks and discriminative networks.The generative network uses a U-shaped structure and combines the attention mechanism to extract and transform the features of the input image,and the generative network is responsible for the prediction of the missing data in the two-dimensional projection.The discriminative network uses pixel-level convolution blocks for feature extraction to discriminate the authenticity of the output data of the generated network.At the same time,cycle consistency loss and content loss are introduced,combined with adversarial loss to constrain the network output results,so as to obtain better prediction results.The trained generative network can predict the missing part of the CT projection data,and use the FDK(Feldkamp Davis Kress)algorithm to reconstruct the CT image from the complementary cone beam CT 3D projection data.The experimental results show that this method can effectively improve the reconstruction quality of cone beam CT images when a certain scanning angle is missing.Secondly,in order to effectively correct metal artifacts in oral cone beam CT images,this paper develops a metal artifact correction technology based on guided filtering in oral cone beam CT images.The algorithm is mainly divided into FDK reconstruction,metal region segmentation,orthographic projection,interpolation and guided filtering.First,the FDK algorithm is used for CT image reconstruction,image threshold segmentation and morphological processing,and then the segmentation result is subjected to forward projection and projection data interpolation of the corresponding area,and then the interpolated projection data is subjected to guided filtering processing,and finally the The processed projection data is FDK reconstructed again and stitched with the original reconstructed image.The experimental results show that the method can effectively correct the metal artifacts existing in the oral cone beam CT images.Finally,according to the proposed oral cone beam CT metal artifact correction algorithm and the requirements of related medical equipment companies,a set of oral cone beam CT image metal artifact correction function modules is developed.This module uses CUDA parallel operation programming,and parallelizes the FDK algorithm,bilinear interpolation algorithm,and guided filtering algorithm,so that it can use GPU for parallel operation and speed up its running speed.