14 June 2019 in IEEE Transactions on Medical Imaging
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.
1 June 2016 in Pattern Recognition Letters
1 February 2000 in Computer Vision and Image Understanding
This paper extends a previously reported theory and algorithms for object definition based on fuzzy connectedness. In this approach, a strength of connectedness is determined between every pair of image elements. This is done by considering all possible connecting paths between the two elements in each pair. The strength assigned to a particular path is defined as the weakest affinity between successive pairs of elements along the path. Affinity specifies the degree to which elements hang together locally in the image. Although the theory allowed any neighborhood size for affinity definition, it did not indicate how this was to be selected. By bringing object scale into the framework in this paper, not only the size of the neighborhood is specified but also it is allowed to change in different parts of the image. This paper argues that scale-based affinity, and hence connectedness, is natural in object definition and demonstrates that this leads to more effective object segmentation. The approach presented here considers affinity to consist of two components. The homogeneity-based component indicates the degree of affinity between image elements based on the homogeneity of their intensity properties. The object-feature-based component captures the degree of closeness of their intensity properties to some expected values of those properties for the object. A family of non-scale-based and scale-based affinity relations are constructed dictated by how we envisage the two components to characterize objects. A simple and effective method for giving a rough estimate of scale at different locations in the image is presented. The original theoretical and algorithmic framework remains more-or-less the same but considerably improved segmentations result. The method has been tested in several applications qualitatively. A quantitative statistical comparison between the non-scale-based and the scale-based methods was made based on 250 phantom images. These were generated from 10 patient MR brain studies by first segmenting the objects, then setting up appropriate intensity levels for the object and the background, and then by adding five different levels for each of noise and blurring and a fixed slow varying background component. Both the statistical and the subjective tests clearly indicate that the scale-based method is superior to the non-scale-based method in capturing details and in robustness to noise. It is also shown, based on these phantom images, that any (global) optimum threshold selection method will perform inferior to the fuzzy connectedness methods described in this paper.
1 February 2008 in Journal of Bone and Mineral Research
Trabecular plates and rods are important microarchitectural features in determining mechanical properties of trabecular bone. A complete volumetric decomposition of individual trabecular plates and rods was used to assess the orientation and morphology of 71 human trabecular bone samples. The ITS-based morphological analyses better characterize microarchitecture and help predict anisotropic mechanical properties of trabecular bone. Introduction: Standard morphological analyses of trabecular architecture lack explicit segmentations of individual trabecular plates and rods. In this study, a complete volumetric decomposition technique was developed to segment trabecular bone microstructure into individual plates and rods. Contributions of trabecular type-associated morphological parameters to the anisotropic elastic moduli of trabecular bone were studied. Materials and Methods: Seventy-one human trabecular bone samples from the femoral neck (FN), tibia, and vertebral body (VB) were imaged using μCT or serial milling. Complete volumetric decomposition was applied to segment trabecular bone microstructure into individual plates and rods. The orientation of each individual trabecula was determined, and the axial bone volume fractions (aBV/TV), axially aligned bone volume fraction along each orthotropic axis, were correlated with the elastic moduli. The microstructural type-associated morphological parameters were derived and compared with standard morphological parameters. Their contributions to the anisotropic elastic moduli, calculated by finite element analysis (FEA), were evaluated and compared. Results: The distribution of trabecular orientation suggested that longitudinal plates and transverse rods dominate at all three anatomic sites. aBV/TV along each axis, in general, showed a better correlation with the axial elastic modulus (r2 = 0.95∼0.99) compared with BV/TV (r2 = 0.93∼0.94). The plate-associated morphological parameters generally showed higher correlations with the corresponding standard morphological parameters than the rod-associated parameters. Multiple linear regression models of six elastic moduli with individual trabeculae segmentation (ITS)-based morphological parameters (adjusted r2 = 0.95∼0.98) performed equally well as those with standard morphological parameters (adjusted r2 = 0.94∼0.97) but revealed specific contributions from individual trabecular plates or rods. Conclusions: The ITS-based morphological analyses provide a better characterization of the morphology and trabecular orientation of trabecular bone. The axial loading of trabecular bone is mainly sustained by the axially aligned trabecular bone volume. Results suggest that trabecular plates dominate the overall elastic properties of trabecular bone.
1 October 2006 in Journal of Bone and Mineral Research
The roles of microarchitecture and types of trabeculae in determining elastic modulus of trabecular bone have been studied in μCT images of 29 trabecular bone samples by comparing their Young's moduli calculated by finite element analysis (FEA) with different trabecular type‐specific reconstructions. The results suggest that trabecular plates play an essential role in determining elastic properties of trabecular bone.Introduction: Osteoporosis is an age‐related disease characterized by low bone mass and architectural deterioration. Other than bone volume fraction (BV/TV), microarchitecture of bone is also believed to be important in governing mechanical properties of trabecular bone. We quantitatively examined the role of microarchitecture and relative contribution of trabecular types of individual trabecula in determining the elastic property of trabecular bone.Materials and Methods: Twenty‐nine human cadaveric trabecular bone samples were scanned at 21‐μm resolution using a μCT system. Digital topological analysis (DTA) consisting of skeletonization and classification was combined with a trabecular type‐specific reconstruction technique to extract the skeleton and identify topological type of trabeculae of the original trabecular bone image. Four different μCT‐based finite element (FE) models were constructed for each specimen: (1) original full voxel; (2) skeletal voxel; (3) rod‐reconstructed, preserving rod volume and plate skeleton; and (4) plate‐reconstructed, preserving plate volume and rod skeleton. For each model, the elastic moduli were calculated under compression along each of three image‐coordinate axis directions. Plate and rod tissue fractions directly measured from DTA‐based topological classification were correlated with the elastic moduli computed from full voxel model.Results: The elastic moduli of skeleton models were significantly correlated with those of full voxel models along all three coordinate axes (r2 = 0.38−0.53). The rod‐reconstructed model contained 21.3% of original bone mass and restored 1.5% of elastic moduli, whereas the plate‐reconstructed model contained 90.3% of bone mass and restored 53.2% of elastic moduli. Plate tissue fraction showed a significantly positive correlation (r2 = 0.49) with elastic modulus by a power law, whereas rod tissue fraction showed a significantly negative correlation (r2 = 0.42).Conclusions: These results quantitatively show that the microarchitecture alone affects elastic moduli of trabecular bone and trabecular plates make a far greater contribution than rods to the bone's elastic behavior.
31 October 2006 in NMR in Biomedicine