FRAMELET DENOISING FOR LOW-DOSE CT USING DEEP LEARNING

Abstract

Recently, deep leaning algorithms have been extensively studied for low-dose X-ray computed tomography (CT) thanks to its superior performance with very little run-time computational complexity. Here, inspired by the recent interpretation of the deep convolutional neural network (CNN) as deep convolutional framelets, we provide a novel iterative framelet-based denoising algorithm which synergistically combines the expressive power of deep learning and the performance guarantee from the framelet-based denoising algorithms. Numerical results show that the proposed deep convolutional framelet denoising algorithm provides superior reconstruction performance.