A Model-Based Unsupervised Deep Learning Method for Low-Dose CT Reconstruction

Abstract

Low-dose CT (LDCT) is of great significance due to the concern about the potential radiation risk. With the fast development of deep learning, neural networks have become powerful tools in LDCT enhancement. Current deep neural networks for LDCT reconstruction are often trained with paired LDCT dataset and normal-dose CT (NDCT) dataset. However, high quality NDCT dataset paired with LDCT dataset is expensive to acquire or even not available sometimes in reality. In this work, we proposed an unsupervised model-based deep learning (MBDL) for LDCT reconstruction. The network is trained based on group-wise maximum a posterior (G-MAP) loss function with LDCT dataset only. The MBDL is a general framework. It also allows us to combine with supervised training if a small number of paired NDCT dataset accessible to help optimizing the network parameters, i.e. works in a semi-supervised mode. During inference, LDCT images are reconstructed end-to-end by the trained network. We verified the proposed method with simulated projection data from clinical CT images. The proposed method restrained noise well while restoring anatomical structures and it achieved better results than model-based iterative reconstruction (MBIR) with significantly less computational cost. The performances of MBDL were further enhanced by integrating a small paired NDCT dataset for semi-supervised training. The results suggested that MBDL is an efficient and flexible method for LDCT deep learning based reconstruction in the situations lacking of enough high quality NDCT data.

The use of deep learning methods in low-dose computed tomography image reconstruction: a systematic review

“…Semi-supervised learning falls between unsupervised learning (with no labeled training data) and supervised learning (with only labeled training data). Some of semi-supervised and unsupervised studies reviewed are [39,66,74,[127][128][129][130][131][132][133][134][135][136][137][138][139][140]. For example, [129] proposed an unsupervised model-based deep learning (MBDL) for LDCT reconstruction.…”

“…Some of semi-supervised and unsupervised studies reviewed are [39,66,74,[127][128][129][130][131][132][133][134][135][136][137][138][139][140]. For example, [129] proposed an unsupervised model-based deep learning (MBDL) for LDCT reconstruction. The network was trained with only the LDCT data set using the in-group maximum a posteriori (G-MAP) loss function.…”

“…Then, the entire network became a type of generative adversarial network (GAN) with this complementary structure. Another current trend applied more complex loss functions so that observed smoothing artifacts can be overcome [18,54,71,95,100,109,123,129,130,134,146,150,151,154,[156][157][158][159]. Especially in [159], the loss function utilized in comparison has two pixel-level losses (mean square error and mean absolute error), the perceptual loss was based on the Visual Geometry Group network (VGG loss), and the one generated by Wasserstein for training gradient penalty adversarial network (WGAN-GP) adversarial loss, and their weighted sum.…”

Accurate and robust sparse‐view angle CT image reconstruction using deep learning and prior image constrained compressed sensing (DL‐PICCS)

“…The following three general strategies have been developed in the last 5 years to address the sparse‐view CT reconstruction problem using deep learning methods: One can use a deep neural network to transform the aliasing artifact‐contaminated images into the desired artifact‐free images, 15–19 or use a deep neural network to transform the sparse‐view data set into a dense‐view data set and then apply the conventional filtered backprojection (FBP) to reconstruct images, 20,21 or use a deep neural network to directly transform the sparse‐view data set into artifact‐free images 22–24 . It is important to emphasize that, in addition to the sparse‐view CT reconstruction problems, the powerful statistical regression capacity in deep learning can also be exploited to address many other scientific problems in CT such as noise reduction in general low‐dose CT applications, 25–29 or flexible regularizer design in the aforementioned first paradigm 30–48 …”

“…As a matter of fact, numerical solvers of any iterative image reconstruction algorithm can be un-rolled and incorporated into a deep neural network architecture and the reconstruction parameters can then be learned using training data, or alternatively, the hand-crafted regularizers used in the conventional iterative reconstruction algorithms can be learned from the available training data as shown in a large body of literature. [30][31][32][33][34][35][36][37][38][39][40][41][42][43][44][45][46][47][48]52,53 In this paper, we propose a new pathway to combine a deep learning reconstruction strategy with the previously published prior image-constrained CS (PICCS) algorithm 13 to improve reconstruction accuracy for individual patients and enhance generalizability for sparse-view reconstruction problems. This method is referred to as deep learning based PICCS (DL-PICCS), and we will show that the proposed DL-PICCS framework provides us a natural method to take advantage of both deep learning and CS reconstruction methods to address the aforementioned fundamental challenges encountered in current deep-learning-based reconstruction methods.…”