Metal artifact reduction in computed tomography images based on developed generative adversarial neural network

Abstract

Abstract Objectives Metal artifacts are one of the major issues encountered in computed tomography (CT) images since they may make distinguishing healthy and tumor organs and computing dose distribution through radiotherapy very difficult. Accordingly, designing generative adversarial neural networks (GANs) will help reduce metal artifacts. Methods Training and validating images with and without metal artifacts were simulated in MATLAB. Then, these images were used as the input data for the GAN, while CT images of 30 patients with head and neck cancer were used as testing data for the GAN. Finally, the quality metrics of denoised images were compared with those of the noisy images. Results The images of patients with one dental implant have shown more improvement in the oral cavity area (16.81%), which is very important in treatment planning. Simulated images were used for the validation of the GAN's ability for metal artifact reduction. Moreover, the GAN was affected by the density and position of the metal artifacts in the CT images. Conclusions Corrected images allow us to improve the quality of the CT images of patients with metal artifacts.