Distance Regularization Research Articles

Fully automated lesion segmentation and visualization in automated whole breast ultrasound (ABUS) images.

The number of breast cancer patients has increased each year, and the demand for breast cancer detection has become quite large. There are many common breast cancer diagnostic tools. The latest automated whole breast ultrasound (ABUS) technology can obtain a complete breast tissue structure, which improves breast cancer detection technology. However, due to the large amount of ABUS image data, manual interpretation is time-consuming and labor-intensive. If there are lesions in multiple images, there may be some omissions. In addition, if further volume information or the three-dimensional shape of the lesion is needed for therapy, it is necessary to manually segment each lesion, which is inefficient for diagnosis. Therefore, automatic lesion segmentation for ABUS is an important issue for guiding therapy. Due to the amount of speckle noise in an ultrasonic image and the low contrast of the lesion boundary, it is quite difficult to automatically segment the lesion. To address the above challenges, this study proposes an automated lesion segmentation algorithm. The architecture of the proposed algorithm can be divided into four parts: (I) volume of interest selection, (II) preprocessing, (III) segmentation, and (IV) visualization. A volume of interest (VOI) is automatically selected first via a three-dimensional level-set, and then the method uses anisotropic diffusion to address the speckled noise and intensity inhomogeneity correction to eliminate shadowing artifacts before the adaptive distance regularization level set method (DRLSE) conducts segmentation. Finally, the two-dimensional segmented images are reconstructed for visualization in the three-dimensional space. The ground truth is delineated by two radiologists with more than 10 years of experience in breast sonography. In this study, three performance assessments are carried out to evaluate the effectiveness of the proposed algorithm. The first assessment is the similarity measurement. The second assessment is the comparison of the results of the proposed algorithm and the Chan-Vese level set method. The third assessment is the volume estimation of phantom cases. In this study, in the 2D validation of the first assessment, the area Dice similarity coefficients of the real cases named cases A, real cases B and phantoms are 0.84±0.02, 0.86±0.03 and 0.92±0.02, respectively. The overlap fraction (OF) and overlap value (OV) of the real cases A are 0.84±0.06 and 0.78±0.04, real case B are 0.91±0.04 and 0.82±0.05, respectively. The overlap fraction (OF) and overlap value (OV) of the phantoms are 0.95±0.02 and 0.92±0.03, respectively. In the 3D validation, the volume Dice similarity coefficients of the real cases A, real cases B and phantoms are 0.85±0.02, 0.89±0.04 and 0.94±0.02, respectively. The overlap fraction (OF) and overlap value (OV) of the real cases A are 0.82±0.06 and 0.79±0.04, real cases B are 0.92±0.04 and 0.85±0.07, respectively. The overlap fraction (OF) and overlap value (OV) of the phantoms are 0.95±0.01 and 0.93±0.04, respectively. Therefore, the proposed algorithm is highly reliable in most cases. In the second assessment, compared with Chan-Vese level set method, the Dice of the proposed algorithm in real cases A, real cases B and phantoms are 0.84±0.02, 0.86±0.03 and 0.92±0.02, respectively. The Dice of Chan-Vese level set in real cases A, real cases B and phantoms are 0.65±0.23, 0.69±0.14 and 0.76±0.14, respectively. The Dice performance of different methods on segmentation shows a highly significant impact (P<0.01). The results show that the proposed algorithm is more accurate than Chan-Vese level set method. In the third assessment, the Spearman's correlation coefficient between the segmented volumes and the corresponding ground truth volumes is ρ=0.929 (P=0.01). In summary, the proposed method can batch process ABUS images, segment lesions, calculate their volumes and visualize lesions to facilitate observation by radiologists and physicians.

Multitask diffusion affine projection sign algorithm and its sparse variant for distributed estimation

Distributed adaptation over multitask networks has attracted particular attention due to its enhanced modeling capacity compared to that over conventional single-task networks. Most of the existing works derive their multitask adaptive algorithms using mean-square error (MSE) or least squares (LS) criterion, leading to multitask LMS- or LS-type algorithms. These algorithms, however, may suffer from deteriorated convergence rate or even divergence in impulsive noise environments. In order to address this problem, we propose a robust diffusion affine projection sign algorithm for multitask parameter estimation. The algorithm is derived by using the method of data reusing and minimizing the weighted sum of the l1-norms of some intermediate error vectors plus a similarity term subject to constraints on intermediate weight vectors at each agent. The multitask similarity relationship is characterized by the distance regularization among weight vectors. Furthermore, a variant of this algorithm, which is obtained by further regularizing the cost function by the l0-norm of the intermediate weight vector at each agent, is presented to promote convergence rate for jointly sparse parameter vectors estimation.

Nonnegative self-representation with a fixed rank constraint for subspace clustering

A number of approaches to graph-based subspace clustering, which assumes that the clustered data points were drawn from an unknown union of multiple subspaces, have been proposed in recent years. Despite their successes in computer vision and data mining, most neglect to simultaneously consider global and local information, which may improve clustering performance. On the other hand, the number of connected components reflected by the learned affinity matrix is commonly inconsistent with the true number of clusters. To this end, we propose an adaptive affinity matrix learning method, nonnegative self-representation with a fixed rank constraint (NSFRC), in which the nonnegative self-representation and an adaptive distance regularization jointly uncover the intrinsic structure of data. In particular, a fixed rank constraint as a prior is imposed on the Laplacian matrix associated with the data representation coefficients to urge the true number of clusters to exactly equal the number of connected components in the learned affinity matrix. Also, we derive an efficient iterative algorithm based on an augmented Lagrangian multiplier to optimize NSFRC. Extensive experiments conducted on real-world benchmark datasets demonstrate the superior performance of the proposed method over some state-of-the-art approaches.

一种基于最小距离和稀疏图正则约束的非负矩阵解混算法

一种基于最小距离和稀疏图正则约束的非负矩阵解混算法

Fractional Distance Regularized Level Set Evolution With Its Application to Image Segmentation

To avoid the irregularities during the level set evolution, a fractional distance regularized variational model is proposed for image segmentation. We first define a fractional distance regularization term which punishes the deviation of the level set function (LSF) and the signed distance function. Since the fractional derivative of the constant value function outside the starting point is nonzero, the fractional gradient modular of the LSF does not approach infinity where the integer order gradient modular is close to 0. This prevents the sharp reverse diffusion of LSF in flat areas and ensures the smooth evolution of LSF. Then, we use the Grunwald-Letnikov (G-L) fractional derivative to derive the discrete forms of the conjugate of fractional derivatives and fractional divergence. To facilitate the calculation of fractional derivatives and their conjugates, we designed their covering templates. Finally, a numerical solution to the minimization of the energy functional is obtained from these discrete forms and covering templates. Numerical experiments of medical images with different modalities show that the model in this paper can well segment weak images and intensity inhomogeneity images.

Person re-identification by integrating metric learning and support vector machine

Person re-identification (PRID) refers to a technology of matching person across non-overlapping camera views. Metric learning is one of the most commonly used methods in PRID. However, most of them do not explore the label information carried by the labeled training samples, which limits the improvement of recognition performance. To this end, we propose a joint learning method by integrating the discriminative metric learning, the support vector machine (SVM) and the identity discriminator into one model, so as to realize joint construction of metric learning and identity discriminator. In this process, the label information carried by the training samples is fully exploited and the latent identify space of pedestrians is constructed by predicting the person’s identity. To mitigate the appearance ambiguity caused by the variations in camera views, body poses, illumination and occlusion, we develop an extreme distance regularization term and introduce it into the joint learning framework to refine the solution spaces of the metric learning and discriminator. Finally, we present a similarity measure method by combining the advantages of the metric learning and the identity discriminator. Experimental results on several benchmark datasets show that the proposed method significantly outperforms some state-of-the-art methods.

Top distance regularized projection and dictionary learning for person re-identification

Person re-identification (PRID) is integral to many smart surveillance systems. However, owing to the visual ambiguities arising from the variability in viewing angles and illumination, and the presence of occlusions, PRID continues to present many challenges, especially when only a single image per view is available for each person. To overcome this problem, we propose a top distance regularized projection and dictionary learning (DL) model for PRID. The model incorporates both projection and DL to form a unified optimization framework to enhance the effectiveness of both these types of learning. Thus, the dictionary and projection matrix are jointly learned within this framework. In particular, the learned projection maps the coding coefficient into a discriminative space and minimizes the distance between the same persons across non-overlapping views such that the dictionary and projection can be discriminated. Moreover, we exploit listwise distances to capture all pairwise similarities. Based on this design, we derive a top distance regularization term to refine the solution space of the DL model such that the discriminative ability of the learned projection matrix and dictionary are further improved. Experiments on different challenging datasets demonstrate the effectiveness of our method and its superiority over a few current state-of-the-art approaches.

Automated Extraction of Parasite in the Microscopic Images by Distance Regularized Level Set Evolution initialized with Hough Transform

Automated Extraction of Parasite in the Microscopic Images by Distance Regularized Level Set Evolution initialized with Hough Transform

Aging characteristic of effective distance on injecting N2 to displace CH4 in anisotropic coal seam

ABSTRACTGas injection is an effective way to enhance methane drainage in a low permeability coal seam, and the reasonable distance between injection holes and drainage holes depends on effective influence distance of injection gas in a coal seam. For anisotropic permeability of coal seam, the variation of the gas-flow rate from a test hole in the anisotropic coal seam is measured as nitrogen is injected into the injection hole near the test hole. The effective distance regularity is revealed by nitrogen injection in an anisotropic coal seam, and the model of nitrogen injection is established to numerically analyze the aging characteristic of effective influence distance in the anisotropic coal seam. The results show that the permeability in coal seam reduced exponentially with the increase in gas pressure, and it is larger in the direction of parallel to bedding than that in the direction of perpendicular to bedding. The effective distance with nitrogen injection increased exponentially in the direction of parallel/perpendicular to bedding, and the effective distance in the direction of parallel to bedding is bigger than that in the direction of perpendicular to bedding. The optimum distances between injection hole and test hole, with 0.5 MPa of injection gas pressure and 8.36 h of injection gas time, are 2.05 m in parallel to bedding direction, and 1.53 m in perpendicular to bedding direction. This study will have important theoretical and practical significance for a reasonable layout of injection-drainage holes during injecting gas to enhance methane drainage.

Analysis of Tuberculosis in Chest Radiographs for Computerized Diagnosis using Bag of Keypoint Features.

Chest radiography is the most preferred non-invasive imaging technique for early diagnosis of Tuberculosis (TB). However, lack of radiological expertise in TB detection leads to indiscriminate chest radiograph (CXR) screening. A modest classification approach based on the local image description to detect subtle characteristics of TB using CXRs is highly recommended. In this work, an attempt has been made to classify normal and TB CXR images using Bag of Features (BoF) approach with Speeded-Up Robust Feature (SURF) descriptor. The images are obtained from a public database. Lung fields segmentation is performed using Distance Regularized Level Set (DRLS) formulation. The results of segmentation are validated against the ground truth images using similarity, overlap and area correlation measures. BoF approach with SURF keypoint descriptors is implemented to categorize the images using Multilayer Perceptron (MLP) classifier. The obtained results demonstrate that the DRLS method is able to delineate lung fields from CXR images. The BoF with SURF keypoint descriptor is able to characterize local attributes of normal and TB images. The segmentation results are found to be in high correlation with ground truth. MLP classifier is found to provide high Recall, Specificity (Spec), Accuracy, F-score and Area Under the Curve (AUC) values of 87.7%, 85.9%, 87.8%, 87.6% and 94% respectively between normal and abnormal images. The proposed computer aided diagnostic approach is found to perform better as compared to the existing methods. Thus, the study can be of significant assistance to physicians at the point of care in resource constrained regions.

Concurrent optimization of structural topology and infill properties with a CBF-based level set method

In this paper, a parametric level-set-based topology optimization framework is proposed to concurrently optimize the structural topology at the macroscale and the effective infill properties at the micro/meso scale. The concurrent optimization is achieved by a computational framework combining a new parametric level set approach with mathematical programming. Within the proposed framework, both the structural boundary evolution and the effective infill property optimization can be driven by mathematical programming, which is more advantageous compared with the conventional partial differential equation-driven level set approach. Moreover, the proposed approach will be more efficient in handling nonlinear problems with multiple constraints. Instead of using radial basis functions (RBF), in this paper, we propose to construct a new type of cardinal basis functions (CBF) for the level set function parameterization. The proposed CBF parameterization ensures an explicit impose of the lower and upper bounds of the design variables. This overcomes the intrinsic disadvantage of the conventional RBF-based parametric level set method, where the lower and upper bounds of the design variables oftentimes have to be set by trial and error. A variational distance regularization method is utilized in this research to regularize the level set function to be a desired distanceregularized shape. With the distance information embedded in the level set model, the wrapping boundary layer and the interior infill region can be naturally defined. The isotropic infill achieved via the mesoscale topology optimization is conformally fit into the wrapping boundary layer using the shape-preserving conformal mapping method, which leads to a hierarchical physical structure with optimized overall topology and effective infill properties. The proposed method is expected to provide a timely solution to the increasing demand for multiscale and multifunctional structure design.

An Adaptive Stopping Active Contour Model for Image Segmentation

Active contour models (ACMs) are widely used in image segmentation applications. However, the selection of maximum iterations which controls the convergence of the ACMs is still a challenging problem. In this paper, an adaptive method for choosing the optimal number of iterations based on the local and global intensity fitting energy is proposed, which increases the automaticity of the active contour model. Moreover, the adoption of the reaction diffusion (RD) method instead of the distance regularization term can improve the accuracy and speed of segmentation effectively. Experimental results on synthetic and real images show that the proposed model outperforms other representative models in terms of accuracy and efficiency.

A Distance Regularized Level-Set Evolution Model Based MRI Dataset Segmentation of Brain’s Caudate Nucleus

The caudate nucleus of the brain is highly correlated to the emotional decision-making of pessimism, which is an important process for improving the understanding and treatment of depression; and the segmentation of the caudate nucleus is the most basic step in the process of analysis and research concerning this region. In this paper, Level Set Method (LSM) is applied for caudate nucleus segmentation. Firstly, Distance Regularized Level Set Evolution (DRLSE), Region-Scalable Fitting (RSF) and Local Image Fitting (LIF) models are proposed for segmentation of the caudate nucleus of Magnetic Resonance Imaging (MRI) images of the brain, and the segmentation results are compared by using selected evaluation indices. The average Dice Similarity Coefficient (DSC) values of the proposed three methods all exceed 85%, and the average Jaccard Similarity (JS) values are over 77%, respectively. The results indicate that all these three models can have good segmentation results for medical images with intensity inhomogeneity and meet the general segmentation requirements, while the proposed DRLSE model performs better in segmentation.

An Adaptive Fuzzy Level Set Model With Local Spatial Information for Medical Image Segmentation and Bias Correction

Medical image segmentation is still a challenging task due to noise and intensity inhomogeneity. An adaptive fuzzy level set model (AFLSM) with local spatial information is presented in this paper for accurately segmenting medical images and correcting bias field. A weighting scheme that can ensure each pixel in the neighborhood to have anisotropic weight is first introduced to remove noisy pixels and hence improve the robustness to noise. Then, a linear combination of orthogonal basis functions is used to represent bias field to ensure its smoothly and slowly varying property. Besides, to improve the robustness to initialization, this adaptive fuzzy level set model fuses a level set model with the membership function of fuzzy clustering, which can adaptively adjust the evolution of level set function. Finally, the distance regularization term in energy formulation is redefined with a novel double-well potential function to inherently maintain the accuracy and stability of the AFLSM. The AFLSM is first represented in the two-phase case and subsequently extended to the multi-phase formulation. The numerous visual segmentation results and quantitative evaluation can demonstrate the performance of the AFLSM on synthetic and real medical images. Comparison with the state-of-the-art models shows that the AFLSM can achieve better segmentation results with an improvement of 0.2286 ± 0.1477 in Dice coefficient and 0.1350 ± 0.0661 in Jaccard similarity coefficient in terms of robustness and the capability to correct bias field, respectively.

Change Detection of Water Index in Danjiangkou Reservoir Using Mixed Log-Normal Distribution Based Active Contour Model

The use of synthetic aperture radar (SAR) images for water segmentation can accurately extract the boundaries of water areas and is of great significance for studying the temporal and spatial changes of lakes and other environmental elements. In view of the fact that SAR image itself has characteristics such as speckle noise and large volumes, this paper proposes an edge active contour model (ACM) based on the mixed log-normal distribution for SAR image edge extraction and evaluates the parameters of distribution by using the classic expectation-maximization (EM) algorithm. Furthermore, compared with the existing models, the proposed algorithm introduces regional variable coefficients and modifies the evolution rate in the distance regularization term so that the level set can be quickly and accurately stopped at the target edge. For practical application, the proposed ACM is applied to extract the outline of the Danjiangkou Reservoir (DJKR) with time as a sequence. The experimental results show that it is robust to noise, improves the precision of land and water segmentation, and helps to determine the changing trends of indicators, such as water surface area, average water depth, and relative storage capacity.

Automated Parasite’s Detection in Microscopic Images of Stools Using Distance Regularized Level Set Evolution Initialized with Hough Transform

Background and purpose: The analysis of biomedical microscopic images is carried out manually in medical laboratories. The manual analysis of clinical images lets to both repetitive tasks and management of huge amounts of data. This is tedious and times consuming for laboratory technicians. Inevitably, it is also prone to human errors. Our objective in this work is to contribute to the automation of the analysis of microscopic images of stools using Distance Regularized Level Set Evolution automatically initialized by Hough transform. Method: We firstly converted the microscopic images to edge maps using canny algorithm. Next, we located the parasite through circular Hough transform and draw circles around them. Those circles stand as initial contours of DRLSE. The contours evolve until they fit the boundaries of the parasites. The final extraction is performed using a complementary method based on the signed distance character of the level set function. Results: The Distance Regularized Level Set Evolution has been automatically initialized. We applied our method to the detection of intestinal parasites in microscopic images. Experimental results show accurate, efficient and less time consuming of our scheme compared to others recently proposed in the literature. Conclusion: This is a notable contribution to the automation of stools examination in the medical laboratories. In forthcoming works, we plan to include this segmentation process in an expert system of parasitic diseases diagnosis.

An Image Segmentation Method Based on Improved Regularized Level Set Model

When the level set algorithm is used to segment an image, the level set function must be initialized periodically to ensure that it remains a signed distance function (SDF). To avoid this defect, an improved regularized level set method-based image segmentation approach is presented. First, a new potential function is defined and introduced to reconstruct a new distance regularization term to solve this issue of periodically initializing the level set function. Second, by combining the distance regularization term with the internal and external energy terms, a new energy functional is developed. Then, the process of the new energy functional evolution is derived by using the calculus of variations and the steepest descent approach, and a partial differential equation is designed. Finally, an improved regularized level set-based image segmentation (IRLS-IS) method is proposed. Numerical experimental results demonstrate that the IRLS-IS method is not only effective and robust to segment noise and intensity-inhomogeneous images but can also analyze complex medical images well.

Local comparison of cup to disc ratio in right and left eyes based on fusion of color fundus images and OCT B-scans

Symmetry analysis of Optical Coherence Tomography (OCT) images in right and left eyes can lead to introduction of new biomarkers for early detection of eye diseases. In this study, we investigate the symmetry between two eyes by calculating local cup to disc ratio (CDR) from each B-scan based on fusion of fundus images and OCT B-scans. For this purpose, in the first phase, the OCT data of Optic Nerve Head (ONH) in right and left eyes are aligned by finding the equivalent B-scans of both eyes based on the fovea-ONH axes. Since the fovea-ONH axes in OCT data are not available, at first, left and right fundus images are aligned according to their automatically detected fovea-ONH axes. Then, OCT data are registered to corresponding fundus images based on maximum similarity between en-face OCT data and fundus image in each eye. This two-stage alignment procedure depends on 1) the blood vessels, automatically extracted by Hessian analysis of directional curvelet sub-bands, and 2) the disc contour of ONH in fundus images, detected by Distance Regularized Level Set Evolution (DRLSE) algorithm. In the second phase, in order to calculate the local CDRs, the disc and cup boundaries are extracted from the aligned B-scans in left and right eyes. The disc boundary is limited by Bruch-Membrane opening, and the cup boundary is defined by retinal layer border. Therefore, Inner Limiting Membrane (ILM) and Retinal Pigment Epithelium (RPE) layers are extracted using ridgelet transform to find the disc-edge point and cup-edge point in each B-scan. Finally, the ratio of summed areas of cups to summed areas of discs is calculated in corresponding local regions to find the local CDRs. Using this method, we can also introduce a new index called local volumetric CDR (VCDR) by dividing the volume of cup in a specific region to the corresponding volume of disc extracted from OCT images of ONH. Forty healthy OCT datasets of size 650 × 512 × 128 (acquired from Topcon 3D OCT-1000) and corresponding 1536 × 1612 fundus images were used in this study. In addition to point-by-point comparison of CDRs in equivalent B-scans of aligned OCTs, the CDRs in upper, middle and lower regions were calculated and the maximum symmetry is observed in middle region. In addition, using local VCDR, the symmetry of 3D OCTs of both eyes is analyzed in 24 volumetric sectors.

Semi–Automatic Corpus Callosum Segmentation and 3D Visualization Using Active Contour Methods

Accurate 3D computer models of the brain, and also of parts of its structure such as the corpus callosum (CC) are increasingly used in routine clinical diagnostics. This study presents comparative research to assess the utility and performance of three active contour methods (ACMs) for segmenting the CC from magnetic resonance (MR) images of the brain, namely: an edge-based active contour model using an inflation/deflation force with a damping coefficient (EM), the Selective Binary and Gaussian Filtering Regularized Level Set (SBGFRLS) method and the Distance Regularized Level Set Evolution (DRLSE) method. The pre-processing methods applied during research work were to improve the contrast, reduce noise and thus help segment the CC better. In this project, 3D CC models reconstructed based on the segmentations of cross-sections of MR images were also visualised. The results, as measured by quantitative tests of the similarity indice (SI) and overlap value (OV) are the best for the EM model (SI = 92%, OV = 82%) and are comparable to or better than those for other methods taken from a literature review. Furthermore, the properties of the EM model consisting in its ability to both expand and shrink at the same time allow segmentations to be better fitted in subsequent CC slices then in state-of-the art ACMs such as DRLSE or SBGFRLS. The CC contours from previous and subsequent iterations produced by the EM model can be used for initiation in subsequent or previous frames of MR images, which makes the segmentation process easier, particularly as the CC area can increase or decrease in subsequent MR image frames.

MRI Images Segmentation and 3D Reconstruction for Cerebral Cancer Detection

Abstract Cancer is a dense and an abnormal rapid multiplication (proliferation) of cells in the tissues of the human body. The brain tumor is one of the most dangerous and deadly tumors. Fortunately, the evolution of science has allowed us to create very efficient medical imaging techniques in order to discover this type of cancer. Chief among these techniques is magnetic resonance imaging (MRI) which is a very efficient technique compared to ultrasound. In this work, we are interested in the detection of this type of cancer allowing a three-dimensional (3D) reconstruction of MRI images. The segmentation methods used are based primarily on the Fuzzy C-Means algorithm that classifies and isolates parts of the brain tissue, and secondly on Distance Regularized Level Set Evolution technique for tumor detection. The obtained results show the effectiveness of this approach to detect brain tumor. The 3D reconstruction is finally carried out to better visualize the tumor as a whole and to detect its expansion. It is conducted using an indirect volume rendering method, which is the Marching Cubes algorithm.