Metal Artifact Recognition using Deep Convolutional Neural Network in Abdomen and Pelvis CT Image

In medicine, radiotherapy image obtained from CT or MRI is an important part in planning to find dosage for radiation treatment for cancer patients. Those images that obtained from CT may have artifacts caused by many factors such as motion artifact, metal artifact, scatter, ring artifact, pseudo enhancement and cone beam effect which is a component that makes image analysis worse. For improvement the quality of image that have artifact, image processing technique is used to manage and reduce it before bring to apply in radiotherapy of medicine. However, for finding the artifacts of 3D image is so difficultly and take more time because 3D image has to split into 2D image before. The aim of this research is to identify the artifact noise in 2D slice by using Deep Convolutional Neural Network (DCNN) model of metal artifact for reduce time. In this experiment have divided dataset into two types of image include 100 image of artifacts and 100 image Non-artifacts for training and 20 images for testing. The result has been shown that accuracy of artifacts and Non-artifacts are 76% and 62.28% respectively. In addition, this studied has found that a small amount of data from Thailand individual results in less accuracy.