Search

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 radiation risk of X-ray CT gained increasing concern in the past decades. Lowering CT scan dose leads to noisy raw data as well as streak artifacts after reconstruction. Extensive studies have been conducted to reduce noise and artifacts for low-dose CT (LDCT). As deep learning has achieved great success in computer vision tasks, it also become a powerful tool in LDCT denoising. Commonly used deep learning methods such as supervised learning and generative adversarial learning have a strong dependence on large normal-dose CT (NDCT) dataset. While in real cases, the NDCT dataset is often expensive or not accessible, which limits the implementation of deep learning. In recent studies, multiple deep learning methods have been proposed for LDCT denoising without NDCT data. Among them, a popular type of methods is noisy label training (NLT) which use LDCT data as labels for network supervised training. Noise2Void is an easily implementable and representative method of NLT and has achieved great results in pixel-independent noise denoising. Another type is distribution learning methods which reduce LDCT noise-level by learning NDCT distribution. Deep distribution learning from noisy samples (DDLN) learns the NDCT distribution from LDCT data only and adopts MAP estimation for LDCT denoising with the learned NDCT distribution prior. It is effective for LDCT projection data denoising. In this work, the two representative methods are compared for LDCT projection data denoising under different noise-levels to seek for their suitable application scenarios.

Because of the growing concern over the radiation dose delivered to patients, X-ray cone-beam CT (CBCT) imaging of low dose is of great interest. It is difficult for traditional reconstruction methods such as Feldkamp to reduce noise and keep resolution at low doses. A typical method to solve this problem is using optimization-based methods with careful modeling of physics and additional constraints. However, it is computationally expensive and very time-consuming to reach an optimal solution. Recently, some pioneering work applying deep neural networks had some success in characterizing and removing artifacts from a low-dose data set. In this study, we incorporate imaging physics for a cone-beam CT into a residual convolutional neural network and propose a new end-to-end deep learning-based method for slice-wise reconstruction. By transferring 3D projection to a 2D problem with a noise reduction property, we can not only obtain reconstructions of high image quality, but also lower the computational complexity. The proposed network is composed of three serially connected sub-networks: a cone-to-fan transformation sub-network, a 2D analytical inversion sub-network, and an image refinement sub-network. This provides a comprehensive solution for end-to-end reconstruction for CBCT. The advantages of our method are that the network can simplify a 3D reconstruction problem to a 2D slice-wise reconstruction problem and can complete reconstruction in an end-to-end manner with the system matrix integrated into the network design. Furthermore, reconstruction can be less computationally expensive and easily parallelizable compared with iterative reconstruction methods.

Low dose CT is of great interest in these days. Dose reduction raises noise level in projection and decrease image quality in reconstructions. Model based image reconstructions can combine noise statistical model together with prior knowledge in an optimization problem so that significantly reduce noise and artefacts. In this work, we propose a deep learning neural network for low lose CT reconstruction in the sense of penalized weighted least-squares (PWLS) so that a self-supervised learning can be done with no ground-truth information needed. Instead of minimizing cost function for each image, the network learns to minimize the cost function for the whole training set. Features learned can be robust. No iterative processes are needed for real data reconstruction with such a trained network. We carried on our experiments on data from a practical dental CT. Applausible reconstructions with great noise reductions are obtained.

X-ray cone beam CT (CBCT) is widely used in many fields such as clinic diagnosis, public safety inspection, and nondestructive testing. In recent years, radiation risk gets growing concern by the public. As a result, X-ray cone-beam CT (CBCT) imaging of low dose is becoming more and more helpful in many fields especially clinic diagnosis and public safety inspection. To get better reconstruction results, a typical idea is to add useful and meaningful constraints in reconstruction procedure. The evaluation criterion of reconstruction could be a general target function concerning both physical models and specific constraints. As computer hard-ware evolves greatly, optimization-based iterative algorithms to minimizing these functions have been proposed and proved helpful for low-dose case. However, these methods are usually time consuming. Nowadays, deep learning method has been utilized in image processing. In CT fields, it has been shown that reconstruction algorithms based on convolutional neural network (CNN) is effective in solving problems with low-dose or sparse-view case. In our work, we proposed a novel CNN structure which is designed for transferring cone-beam projection to fan-beam projection of each cross section attenuation map of an object with noise reduction. After the 3D problem is turned to be a 2D one, we simply use the system matrix to reconstruct the object. Because of its slice-wise character, the method can effectively simplify the 3D reconstruction problem.

Purpose Metal artifact is a quite common problem in diagnostic dental computed tomography (CT) images. Due to the high attenuation of heavy materials such as metal, severe global artifacts can occur in reconstructions. Typical metal artifact reduction (MAR) techniques segment out the metal regions and estimate the corrupted projection data by various interpolation methods. However, interpolations are not accurate and introduce new artifacts or even deform the teeth in the reconstructed image. This work presents a new strategy to take advantage of the power of deep learning for metal artifact reduction. Method The analysis first uses coarse reconstructions from simulated locally interpolated data affected by metal fillings as a starting point. A deep learning network is then trained using the simulated data and applied to practical data. Thus, an easily implemented three-step MAR method is formed: Firstly, use the acquired projection data to create a preliminary image reconstruction with linearly interpolated data for the metal-related projections. Secondly, a deep learning network is used to remove the artifacts from the linear interpolation and recover the nonmetal region information. Thirdly, the method adds the ROI reconstruction of the metal regions. The structures behind the shading artifacts in the direct filtered back-projection (FBP) reconstruction can be partially recovered by interpolation-based MAR (I-MAR) with the network further correcting for interpolation errors. The key to this method is that the linear interpolation reconstruction errors can be easily simulated to train a network and the effectiveness of the network can be easily generalized to I-MAR results in real situations. Results We trained a network with a simulation dataset and validated the network against a separate simulation dataset. Then, the network was tested using simulation data that did not overlap with the training/validation datasets and real patient datasets. Both tests gave encouraging results with accurate tooth structure recovery and few artifacts. The relative root mean square error and structure similarity index method indexes were significantly improved in the tests. The method was also evaluated by two experienced dentists who gave positive evaluations. Conclusions This work presents a strategy to build a transferable learning from simulations to practical systems for metal artifact reduction using a supervised deep learning method. The system transforms the MAR analyses to an interpolation-artifact reduction problem to recover structural details from the coarse interpolation reconstruction. In this way, training data from simulations with ground truth labels can easily model the similar features in real data with I-MAR as the bridge. The network can seamlessly optimize both simulations and real data. The whole method is easily implemented with little computational cost. Test results demonstrated that this is an effective MAR method applicable to practical dental CT systems.

Sparse-view CT imaging has been a hot topic in the medical imaging field. By decreasing the number of views, dose delivered to patients can be significantly reduced. However, sparse-view CT reconstruction is an illposed problem. Serious streaking artifacts occur if reconstructed with analytical reconstruction methods. To solve this problem, many researches have been carried out to optimize in the Bayesian framework based on compressed sensing, such as applying total variation (TV) constraint. However, TV or other regularized iterative reconstruction methods are time consuming due to iterative process needed. In this work, we proposed a method of angular resolution recovery in projection domain based on deep residual convolutional neural network (CNN) so that projections at unmeasured views can be estimated accurately. We validated our method by a disjointed data set new to trained networks. With recovered projections, reconstructed images have little streaking artifacts. Details corrupted due to sparse view are recovered. This deep learning based sinogram recovery can be generalized to more data insufficient situations.