Deep Learning Reconstruction for X-ray CT

Abstract

Image reconstruction for X-ray CT is an inverse problem of Radon transform. The analytical and the iterative image reconstruction are well known methods for the solution. In recent years, ill-posed problems when sufficient projection data cannot be collected have been actively discussed, and there has been active research on iterative image reconstruction that incorporates a priori or physical models into algorithms. On the other hand, the third boom in machine learning has arrived. In the image reconstruction, AUTOMAP has been reported, and it has become to be possible to transform from projection domain to image domain using deep learning. The deep learning reconstruction has newly born as a third method after analytical and iterative image reconstruction. In this paper, we have explained the basic knowledge of deep learning, and the latest papers for deep learning reconstruction.