In this paper, we present a semi-supervised deep learning approach to accurately recover high-resolution (HR) CT images from low-resolution (LR) counterparts. Specifically, with the generative adversarial network (GAN) as the building block, we enforce the cycle-consistency in terms of the Wasserstein distance to establish a nonlinear end-to-end mapping from noisy LR input images to denoised and deblurred HR outputs. We also include the joint constraints in the loss function to facilitate structural preservation. In this process, we incorporate deep convolutional neural network (CNN), residual learning, and network in network techniques for feature extraction and restoration. In contrast to the current trend of increasing network depth and complexity to boost the imaging performance, we apply a parallel ${1}\times {1}$ CNN to compress the output of the hidden layer and optimize the number of layers and the number of filters for each convolutional layer. The quantitative and qualitative evaluative results demonstrate that our proposed model is accurate, efficient and robust for super-resolution (SR) image restoration from noisy LR input images. In particular, we validate our composite SR networks on three large-scale CT datasets, and obtain promising results as compared to the other state-of-the-art methods.

CT Super-Resolution GAN Constrained by the Identical, Residual, and Cycle Learning Ensemble (GAN-CIRCLE)

The COVID-19 coronavirus has spread rapidly around the world and has caused global panic. Chest CT images play a major role in confirming positive COVID-19 patients. The computer aided diagnosis of COVID-19 from CT images based on artificial intelligence have been developed and deployed in some hospitals. But environmental influences and the movement of lung will affect the image quality, causing the lung parenchyma and pneumonia area unclear in CT images. Therefore, the performance of COVID-19's artificial intelligence diagnostic algorithm is reduced. If chest CT images are reconstructed, the accuracy and performance of the aided diagnostic algorithm may be improved. In this paper, a new aided diagnostic algorithm for COVID-19 based on super-resolution reconstructed images and convolutional neural network is presented. Firstly, the SRGAN neural network is used to reconstruct super-resolution images from original chest CT images. Then COVID-19 images and Non-COVID-19 images are classified from super-resolution chest CT images by VGG16 neural network. Finally, the performance of this method is verified by the pubic COVID-CT dataset and compared with other aided diagnosis methods of COVID-19. The experimental results show that improving the data quality through SRGAN neural network can greatly improve the final classification accuracy when the data quality is low. This proves that this method can obtain high accuracy, sensitivity and specificity in the examined test image datasets and has similar performance to other state-of-the-art deep learning aided algorithms.

/pdf/classification-of-covid-19-pneumonia-from-chest-ct-images-2gyoeshqmp.pdf

Classification of COVID-19 pneumonia from chest CT images based on reconstructed super-resolution images and VGG neural network.

While X-ray computed tomography (XCT) is pushed further into the micro- and nanoscale, the limitations of various tool components and object motion become more apparent. For high-resolution XCT, it is necessary but practically difficult to align these tool components with sub-micron precision. The aim is to develop a novel reconstruction methodology that considers unavoidable misalignment and object motion during the data acquisition in order to obtain high-quality three-dimensional images and that is applicable for data recovery from incomplete datasets. A reconstruction software empowered by sophisticated correction modules that autonomously estimates and compensates artefacts using gradient descent and deep learning algorithms has been developed and applied. For motion estimation, a novel computer vision methodology coupled with a deep convolutional neural network approach provides estimates for the object motion by tracking features throughout the adjacent projections. The model is trained using the forward projections of simulated phantoms that consist of several simple geometrical features such as sphere, triangle and rectangular. The feature maps extracted by a neural network are used to detect and to classify features done by a support vector machine. For missing data recovery, a novel deep convolutional neural network is used to infer high-quality reconstruction data from incomplete sets of projections. The forward and back projections of simulated geometric shapes from a range of angular ranges are used to train the model. The model is able to learn the angular dependency based on a limited angle coverage and to propose a new set of projections to suppress artefacts. High-quality three-dimensional images demonstrate that it is possible to effectively suppress artefacts caused by thermomechanical instability of tool components and objects resulting in motion, by center of rotation misalignment and by inaccuracy in the detector position without additional computational efforts. Data recovery from incomplete sets of projections result in directly corrected projections instead of suppressing artefacts in the final reconstructed images. The proposed methodology has been proven and is demonstrated for a ball bearing sample. The reconstruction results are compared to prior corrections and benchmarked with a commercially available reconstruction software. Compared to conventional approaches in XCT imaging and data analysis, the proposed methodology for the generation of high-quality three-dimensional X-ray images is fully autonomous. The methodology presented here has been proven for high-resolution micro-XCT and nano-XCT, however, is applicable for all length scales.

/pdf/deep-learning-based-inaccuracy-compensation-in-4a4op925ke.pdf

Deep Learning-based Inaccuracy Compensation in Reconstruction of High Resolution XCT Data.

Deep Implicit Functions (DIFs) represent 3D geometry with continuous signed distance functions learned through deep neural nets. Recently DIFs-based methods have been proposed to handle shape reconstruction and dense point correspondences simultaneously, capturing semantic relationships across shapes of the same class by learning a DIFs-modeled shape template. These methods provide great flexibility and accuracy in reconstructing 3D shapes and inferring correspondences. However, the point correspondences built from these methods do not intrinsically preserve the topology of the shapes, unlike mesh-based template matching methods. This limits their applications on 3D geometries where underlying topological structures exist and matter, such as anatomical structures in medical images. In this paper, we propose a new model called Neural Diffeomorphic Flow (NDF) to learn deep implicit shape templates, representing shapes as conditional diffeomorphic deformations of templates, intrinsically preserving shape topologies. The diffeomorphic deformation is realized by an autodecoder consisting of Neural Ordinary Differential Equation (NODE) blocks that progressively map shapes to implicit templates. We conduct extensive experiments on several medical image organ segmentation datasets to evaluate the effectiveness of NDF on reconstructing and aligning shapes. NDF achieves consistently state-of-the-art organ shape reconstruction and registration results in both accuracy and quality. The source code is publicly available at https://github.com/Siwensun/Neural_Diffeomorphic_Flow-NDF.

Topology-Preserving Shape Reconstruction and Registration via Neural Diffeomorphic Flow

Medical image quality is highly relative to clinical diagnosis and treatment, leading to a popular research topic of medical image denoising. Image denoising based on deep learning methods has attracted considerable attention owing to its excellent ability of automatic feature extraction. Most existing methods for medical image denoising adapted to certain types of noise have difficulties in handling spatially varying noise; meanwhile, image detail losses and structure changes occurred in the denoised image. Considering image context perception and structure preserving, this paper firstly introduces a medical image denoising method based on conditional generative adversarial network (CGAN) for various unknown noises. In the proposed architecture, noise image with the corresponding gradient image is merged as network conditional information, which enhances the contrast between the original signal and noise according to the structural specificity. A novel generator with residual dense blocks makes full use of the relationship among convolutional layers to explore image context. Furthermore, the reconstruction loss and WGAN loss are combined as the objective loss function to ensure the consistency of denoised image and real image. A series of experiments for medical image denoising are conducted with the denoising results of PSNR = 33.2642 and SSIM = 0.9206 on JSRT datasets and PSNR = 35.1086 and SSIM = 0.9328 on LIDC datasets. Compared with the state-of-the-art methods, the superior performance of the proposed method is outstanding.

/pdf/a-novel-medical-image-denoising-method-based-on-conditional-3n4vohqej0.pdf

A Novel Medical Image Denoising Method Based on Conditional Generative Adversarial Network.

Purpose Cone-beam (CB) CT is a powerful noninvasive imaging modality, and is widely used in many applications. Accurate geometric parameters are essential for high-quality image reconstruction. Usually, a CBCT system with higher spatial resolution, particularly on the order of microns or nanometers, will be more sensitive to the parametric accuracy. Here, we propose a novel calibration method combining a simple phantom containing ball bearing markers and an advanced optimization procedure. This method can be applied to CBCT with reproducible geometry and frame-to-frame invariant geometric parameters. Methods Our proposed simplex-simulated annealing procedure minimizes the cost function that associates the geometrical parameters with the degree to which the back projections of the ball bearings in projections from various viewing angles converge, and the global minimum of the cost function corresponds to the actual geometric parameters. Specifically, six geometric parameters can be directly obtained by minimizing the cost function, and the last parameter, the distance from source to rotation axis (SRD), can be obtained using prior knowledge of the phantom - the spacing between the two ball bearings. Results Numerical simulation was performed to validate that the proposed method with various noise levels. With the proposed method, the mean errors and standard deviations can be reduced to ∼10% and less than 1/3 of a competing benchmark method in the case of strong Gaussian noise (sigma = 200% of the pixel size) and large tilt angle (tilt angle = - 4 ∘ ). The calibration experiments with micro-CT and high-resolution CT scanners demonstrate that the proposed method recovers imaging parameters accurately, leading to superior image quality. Conclusion The proposed method can obtain accurate geometric parameters of a CBCT system with a circular trajectory. While in the case of micro-CT the proposed method has a performance comparable to the competing method, for high-resolution CT, which is more sensitive to geometric calibration, the proposed method demonstrates higher calibration accuracy and more robustness than the benchmark algorithm.

A novel calibration method incorporating nonlinear optimization and ball-bearing markers for cone-beam CT with a parameterized trajectory.

A C-GAN denoising algorithm in Projection Domain for Micro-CT

PURPOSE The misalignment correction in cone beam computed tomography (CBCT), which is usually carried out in an offline manner, is a difficult and tedious process. It becomes even more challenging in microscopic CBCT due to the much higher requirements on spatial resolution. In practice, however, an offline approach for misalignment correction may not be readily implementable, especially in the situations where either time is of the essence or the process needs to be carried out repetitively. Thus, an online self-calibration (i.e., data sustained misalignment correction without the involvement of specific alignment phantom) would be more practical. In this work, we investigate the data sustained misalignment correction in microscopic CBCT via optimization under the Grangeat Epipolar Consistence Condition and evaluate its performance via phantom and specimen studies. METHODS With the cost function defined according to the Grangeat Epipolar Consistency Condition (G-ECC) and by minimizing the cost function using the simplex-simulated annealing algorithm (SIMPSA), we evaluate and verify the G-ECC optimization-based online self-calibration method's performance. Performance is measured in sensitivity, robustness, and accuracy using the projection data of phantoms generated by computer simulation and botanical specimens acquired by a prototype microscopic CBCT. RESULTS The online data sustained misalignment correction in microscopic CBCT via G-ECC optimization works very well in sensitivity and robustness, in addition to its accuracy of 0.27%, 0.48%, and 0.34% relative errors, respectively, in obtaining the three geometric parameters that are the most critical to image reconstruction in CBCT. Quantitatively, the performance in meeting the requirements on spatial resolution is comparable to, if not better than, that of the offline misalignment correction method, in which a specific alignment phantom has to be used. CONCLUSIONS The G-ECC optimization-based online self-calibration approach provides a practical solution (as long as no latitudinal (lateral) data truncation occurs) for misalignment correction in microscopic CBCT, an application that demands high accuracy in geometric alignment for biological (cellular) imaging at super high spatial resolutions in the order of micrometers (2.1 µm).

Data sustained misalignment correction in microscopic cone beam CT via optimization under the Grangeat Epipolar consistency condition.

In multi-lens microscopic cone-beam CT system, different resolutions can be achieved by tuning different objective lens with different magnifications, following the scan field of view (FOV) changed...

Geometric Calibration Based on a Simple Phantom for Multi-Lens Microscopic CT

For most offline calibration methods, a specific phantom scanning is often required before the following imaging tasks. It’s time-consuming and unable to be applied to the unstable systems. Some online approaches don’t have such drawbacks but their accuracy cannot fulfil requirements compared with the offline ones. This paper proposes an Epipolar geometry consistency based online geometric calibration for cone beam CT(CBCT). Four parameters: the detector skew, rotation axis, mid-plane and the source to detector distance are employed for geometrical modeling of CBCT in this paper. A cost function is built by exploiting the Epipolar geometry consistency among the projective views. By taking advantage of the simplex-simulated annealing algorithm(SIMPSA) algorithm to minimize the cost function, we can obtain the geometric parameters of our practical CBCT systems for image reconstruction. In simulation, different noise levels are added to projection images respectively and the experimental results show that the proposed method is insensitive to noise. What’s more, the accuracies of the parameters of the detector skew and rotation axis is slightly higher than that of the other two. In the practical situation, we scanned a thin bamboo stick and a Chinese parasol tree branch. Compared with the image reconstructed with the geometric parameters calculated by the conventional offline calibration approach, the ones reconstructed with the proposed method show no apparent difference and geometric artifacts, which shows that the accuracy of our method is comparable to that of the offline calibration ones.

Liang Zheng

Papers

A novel calibration method incorporating nonlinear optimization and ball-bearing markers for cone-beam CT with a parameterized trajectory.

A C-GAN denoising algorithm in Projection Domain for Micro-CT

Data sustained misalignment correction in microscopic cone beam CT via optimization under the Grangeat Epipolar consistency condition.

Geometric Calibration Based on a Simple Phantom for Multi-Lens Microscopic CT

Self-geometric calibration of circular cone beam CT based on epipolar geometry consistency