Statistical Image Restoration for Low-Dose CT using Convolutional Neural Networks

Abstract

Deep learning has recently attracted widespread interest as a means of reducing noise in low-dose CT (LDCT) images. Deep convolutional neural networks (CNNs) are typically trained to transfer high-quality image features of normal-dose CT (NDCT) images to LDCT images. However, existing deep learning approaches for denoising LDCT images often overlook the statistical property of CT images. In this paper, we propose an approach to statistical image restoration for LDCT using deep learning (StatCNN). We introduce a loss function to incorporate the noise property in the image domain derived from the noise statistics in the sinogram domain. In order to capture the spatially-varying statistics of axial CT images, we increase the receptive fields of the proposed network to cover full-size CT slices. In addition, the proposed network utilizes z-directional correlation by taking multiple consecutive CT slices as input. For performance evaluation, the proposed network was thoroughly trained and tested by leave-one-out cross-validation with a dataset consisting of LDCT-NDCT image pairs. The experimental results showed that the denoising networks successfully reduced the noise level and restored the image details without adding artifacts. This study demonstrates that the statistical deep learning approach can transfer the image style from NDCT images to LDCT images without loss of anatomical information.