Decrease in destructive effects of metal artifacts in mouth area images by optimized neural network to improve treatment quality through radiotherapy

In this study, in order to reduce the effects of metal artifacts caused by metal objects in CT scan images of the mouth area, we investigated the quality parameters of head and neck CT scan images of patients before and after the presence of artifacts and evaluated the changes in quality image parameters to improve the quality of radiotherapy after modifying the images. For this purpose, first, we provided CT scan images of 20 patients with head and neck cancers with and without metal objects in the mouth area and compared the absorbed dose in patients with and without metal objects. Then, in order to prevent the destructive effects of images with artifacts in diagnosis and treatment process in radiotherapy, we created modified images by two different neural network models and evaluated the performance of neural networks by image quality parameters to find the effective neural network. By generative adversarial neural network, in some places around salivary glands and teeth with metal, up to 94% improvement has been achieved in image quality metrics, which is up to 70% better than the convolutional neural network. This study can be done to improve the quality of treatment in radiotherapy on patients with different types of cancer, especially with metal prostheses, in order to improve the quality of CT scan images for better diagnosis and contouring of the therapeutic area and reduce the dose received by patients through radiotherapy.