Vector Field Convolution Research Articles

A novel solution of deep learning for endoscopic ultrasound image segmentation: enhanced computer aided diagnosis of gastrointestinal stromal tumor

Gastrointestinal stromal tumor is one of the critical tumors that doctors do not suggest to get frequent endoscopy, so there is a need for a diagnosis system which can process ultrasound images and figure out the tumor. Many gastrointestinal tumor diagnosis methods were developed, but all of these methods used manual contour rather than automatic segmentation. The research adopts enhanced automatic segmentation to improve the diagnosis of the gastrointestinal stromal tumor with deep convolutional neural networks. This solution’s proposed system is an enhanced automated segmentation methodology using multi-scale Gaussian kernel fuzzy clustering and multi-scale vector field convolution, which segments the ultrasound image automatically into the region of interest (the infected area). Convolutional Neural Network with Class Activation Mapping is done to diagnose an image with the tumor for Four datasets, namely (USS1, SH Hospital, SNUH, BUSI). This proposed system helps to get a clearer tumor image, and the accuracy has increased from 84.275% to 88.4%, and the processing time has reduced from 28.525% to 24.575%. The proposed solution enhanced Automatic Segmentation helped to get clearer tumor image which resulted in increased accuracy and decreased performance time compared to the state-of-the-art. Automatic segmentation overcomes the dependency on the expert for drawing the Region of Interest (ROI).

C3VFC: A Method for Tracing and Quantification of Microglia in 3D Temporal Images

Automatic glia reconstruction is essential for the dynamic analysis of microglia motility and morphology, notably so in research on neurodegenerative diseases. In this paper, we propose an automatic 3D tracing algorithm called C3VFC that uses vector field convolution to find the critical points along the centerline of an object and trace paths that traverse back to the soma of every cell in an image. The solution provides detection and labeling of multiple cells in an image over time, leading to multi-object reconstruction. The reconstruction results can be used to extract bioinformatics from temporal data in different settings. The C3VFC reconstruction results found up to a 53% improvement on the next best performing state-of-the-art tracing method. C3VFC achieved the highest accuracy scores, in relation to the baseline results, in four of the five different measures: Entire structure average, the average bi-directional entire structure average, the different structure average, and the percentage of different structures.

Segmentation of Cotton Leaf Images using Parametric Deformable Model

In this paper, the segmentation of cotton leaves from the complex background has been carried out using deformable model. In order to segment, a database of about 300 cotton leaves image was developed. The collected images were resized to 256x256 size. The resized image has been segmented using Adaptive Diffusion Flow (ADF) model. The ADF model has been obtained by replacing the smoothening energy term of gradient vector flow model with active hyper surface harmonic minimal function used to keep away from weak edges leakage. The infinite Laplace function is used to move the deformable model into narrow concave regions. Further, the developed model has been compared with the gradient vector flow and vector field convolution segmentation methods in terms of number of iterations, time taken for segmentation and various performance parameters namely precision, recall, Manhattan, Jaccard, Dice. From the results, it is concluded that the adaptive diffusion flow method is faster and performance parameters are better than the Gradient Vector Flow (GVF) and Vector Field Convolution (VFC) methods.

An Improved Active Contour Model in Medical Image Segmentation

Aiming at the problem that medical image segmentation has low precision because of restriction of image principle, this paper proposes an improved active contour model based on vector field convolution, and combining both the balloon force and gradient directional information. This method improves the energy function of model, has more chances to approximate the complex boundary. The experimental results show that the new model not only inherits the advantages of the previous models, but also show fast speed, limits the influence of false edge and noise disturbance.

A new effective and powerful medical image segmentation algorithm based on optimum path snakes

Novel segmentation methods based on models of deformable active contours are constantly proposed and validated in different fields of knowledge, with the aim to make the detection of the regions of interest standard. This paper propose a new method called Optimum Path Snakes (OPS), a new adaptive algorithm and free of parameters to define the total energy of a active contour model with automatic initialization and stop criteria. In the experimental assessment, the OPS is compared against some approaches commonly used in the following fields, such as vector field convolution, gradient vector flow, and other specialists methods for lung segmentation using thorax computed tomography images. The segmentation of regions with stroke was carried out with methods based on region growing, watershed and a specialist level set approach. Statistical validations metrics using Dice coefficient (DC) and Hausdorff distance (HD) were also evaluated, as well as the processing time. The results showed that the OPS is a promising tool for image segmentation, presenting satisfactory results for DC and HD, and, many times, superior to the other algorithms it was compared with, including those generated by specialists. Another advantage of the OPS is that it is not restricted to specific types of images, neither applications.

Ultrasound image segmentation using a novel multi-scale Gaussian kernel fuzzy clustering and multi-scale vector field convolution

Ultrasound imaging is most popular technique used for breast cancer screening. Lesion segmentation is challenging step in characterization of breast ultrasound (US) based Computer Aided Diagnosis (CAD) systems due to presence of speckle noise, shadowing effect etc. The aim of this study is to develop an automatic lesion segmentation technique in breast US with high accuracy even in presence of noises, artifacts and multiple lesions. This article presents a novel clustering method called Multi-scale Gaussian Kernel induced Fuzzy C-means (MsGKFCM) for segmentation of lesions in automatically extracted Region of Interest (ROI) in US to delimit the border of the mass. Further, a hybrid approach using MsGKFCM and Multi-scale Vector Field Convolution (MsVFC) is proposed to obtain an accurate lesion margin in breast US images. Initially, the images are filtered using speckle reducing anisotropic diffusion (SRAD) technique. Subsequently, MsGKFCM is applied on filtered images to segment the mass and detect an appropriate cluster center. The detected cluster center is further used by MsVFC to determine the accurate lesion margin. The proposed technique is evaluated on 127 US images using measures such as Jaccard Index, Dice similarity, Shape similarity, Hausdroff difference, Area difference, Accuracy, F-measure and analysis of variance (ANOVA) test. The empirical results suggest that the proposed approach can be used as an expert system to assist medical professionals by providing objective evidences in breast lesion detection. Results obtained are so far looking promising and effective in comparison to state-of-the-art algorithms.

Fully automatic prostate segmentation in MR images using a new hybrid active contour-based approach

Precise prostate segmentation in magnetic resonance (MR) images is mostly utilized for prostate volume estimation, which can help in the determination of prostate-specific antigen density. In this paper, a fully automatic method that contains three successful steps to segment the prostate area in MR images is presented. This method includes a preprocessing stage, an automatic initial point generation step and an active contour-based algorithm with an external force known as vector field convolution (VFC). First, both noise and roughness are approximately removed using Sticks filter and morphology smoothing method. Then, an initial point is automatically generated using multilayer perceptron neural network to initiate the segmentation algorithm. Finally, VFC is employed to extract the prostate region. This system was tested on image data sets to detect the prostate boundaries. Results show that the proposed method can reach a DSC value of $${86~\pm ~6\%}$$ , is faster than existing methods and also more robust as compared to other methods.

Automatic construction of statistical shape models using deformable simplex meshes with vector field convolution energy

Background In the active shape model framework, principal component analysis (PCA) based statistical shape models (SSMs) are widely employed to incorporate high-level a priori shape knowledge of the structure to be segmented to achieve robustness. A crucial component of building SSMs is to establish shape correspondence between all training shapes, which is a very challenging task, especially in three dimensions.Methods We propose a novel mesh-to-volume registration based shape correspondence establishment method to improve the accuracy and reduce the computational cost. Specifically, we present a greedy algorithm based deformable simplex mesh that uses vector field convolution as the external energy. Furthermore, we develop an automatic shape initialization method by using a Gaussian mixture model based registration algorithm, to derive an initial shape that has high overlap with the object of interest, such that the deformable models can then evolve more locally. We apply the proposed deformable surface model to the application of femur statistical shape model construction to illustrate its accuracy and efficiency.ResultsExtensive experiments on ten femur CT scans show that the quality of the constructed femur shape models via the proposed method is much better than that of the classical spherical harmonics (SPHARM) method. Moreover, the proposed method achieves much higher computational efficiency than the SPHARM method.ConclusionsThe experimental results suggest that our method can be employed for effective statistical shape model construction.

Semi-automated myocardial segmentation of bright\xa0blood multi-gradient echo images improves reproducibility of myocardial contours and T2* determination

ObjectivesEarly detection of iron loading is affected by the reproducibility of myocardial contour assessment. A novel semi-automatic myocardial segmentation method is presented on contrast-optimized composite images and compared to the results of manual drawing.Materials and methodsFifty-one short-axis slices at basal, mid-ventricular and apical locations from 17 patients were acquired by bright blood multi-gradient echo MRI. Four observers produced semi-automatic and manual myocardial contours on contrast-optimized composite images. The semi-automatic segmentation method relies on vector field convolution active contours to generate the endocardial contour. After creating radial pixel clusters on the myocardial wall, a combination of pixel-wise coefficient of variance (CoV) assessment and k-means clustering establishes the epicardial contour for each segment.ResultsCompared to manual drawing, semi-automatic myocardial segmentation lowers the variability of T2* quantification within and between observers (CoV of 12.05 vs. 13.86% and 14.43 vs. 16.01%) by improving contour reproducibility (P < 0.001). In the presence of iron loading, semi-automatic segmentation also lowers the T2* variability within and between observers (CoV of 13.14 vs. 15.19% and 15.91 vs. 17.28%).ConclusionApplication of semi-automatic myocardial segmentation on contrast-optimized composite images improves the reproducibility of T2* quantification.

An Efficient Approach for Automated Mass Segmentation and Classification in Mammograms.

Breast cancer is becoming a leading death of women all over the world; clinical experiments demonstrate that early detection and accurate diagnosis can increase the potential of treatment. In order to improve the breast cancer diagnosis precision, this paper presents a novel automated segmentation and classification method for mammograms. We conduct the experiment on both DDSM database and MIAS database, firstly extract the region of interests (ROIs) with chain codes and using the rough set (RS) method to enhance the ROIs, secondly segment the mass region from the location ROIs with an improved vector field convolution (VFC) snake and following extract features from the mass region and its surroundings, and then establish features database with 32 dimensions; finally, these features are used as input to several classification techniques. In our work, the random forest is used and compared with support vector machine (SVM), genetic algorithm support vector machine (GA-SVM), particle swarm optimization support vector machine (PSO-SVM), and decision tree. The effectiveness of our method is evaluated by a comprehensive and objective evaluation system; also, Matthew's correlation coefficient (MCC) indicator is used. Among the state-of-the-art classifiers, our method achieves the best performance with best accuracy of 97.73%, and the MCC value reaches 0.8668 and 0.8652 in unique DDSM database and both two databases, respectively. Experimental results prove that the proposed method outperforms the other methods; it could consider applying in CAD systems to assist the physicians for breast cancer diagnosis.

Convolutional virtual electric field for image segmentation using active contours.

Gradient vector flow (GVF) is an effective external force for active contours; however, it suffers from heavy computation load. The virtual electric field (VEF) model, which can be implemented in real time using fast Fourier transform (FFT), has been proposed later as a remedy for the GVF model. In this work, we present an extension of the VEF model, which is referred to as CONvolutional Virtual Electric Field, CONVEF for short. This proposed CONVEF model takes the VEF model as a convolution operation and employs a modified distance in the convolution kernel. The CONVEF model is also closely related to the vector field convolution (VFC) model. Compared with the GVF, VEF and VFC models, the CONVEF model possesses not only some desirable properties of these models, such as enlarged capture range, u-shape concavity convergence, subject contour convergence and initialization insensitivity, but also some other interesting properties such as G-shape concavity convergence, neighboring objects separation, and noise suppression and simultaneously weak edge preserving. Meanwhile, the CONVEF model can also be implemented in real-time by using FFT. Experimental results illustrate these advantages of the CONVEF model on both synthetic and natural images.

Automatic active contour segmentation approach via vector field convolution

Among the segmentation methods proposed in the literature, the active contour models have been widely used in medical images segmentation. This is due to their efficiency to capture complex shapes. Nevertheless, the adequate set of the initial curves for active contours is needed to lead to good automatic segmentation results. An adequate initialization method should set the initial active contour model close enough, to the final targeted boundary, to avoid local minima and to improve computational efficiency. In this paper, we present a new approach, based on the divergence of vector field and the Dijkstra’s algorithm to automatically initialize and give the B-Snake the ability to change topology in presence of multiple objects very close to each other. The divergence of vector fields informs about the vectors spread. Negative divergence indicates that the vectors converge and positive divergence indicates that the vectors diverge. Thus, we used the negative region of the divergence of the vector field convolution (VFC) to set the initial active contour near the objects and the positive divergence region was used to split the B-Snake via Dijkstra algorithm. To demonstrate the effectiveness of the proposed method, we use computed tomography (CT) images of close bones. This method gives good segmentation results, especially on CT images presenting proximities, compared to results obtained by other automatic segmentation methods from the literature.

Gradient descent with adaptive momentum for active contour models

In active contour models (snakes), various vector force fields replacing the gradient of the original external energy in the equations of motion are a popular way to extract the object boundary. Gradient descent method is usually used to obtain the equations of motion by minimising the energy functional. However, it always suffers from local minimum in extracting complex geometries because of non‐convex functional. Gradient descent method with adaptive momentum term is proposed in this study. First, an acceleration function of evolution is defined. Then, the adaptive momentum term is obtained by calculating the product between the edge stopping function and the defined acceleration function. Finally, adaptive momentum is compatible with the snakes. The edge stopping function is used to decide the influence region of the momentum, whereas the defined acceleration function determines the magnitude of the momentum. It is used to extract the complex geometries (such as deep concavity) when adding the adaptive momentum into some snakes, such as gradient vector field or vector field convolution snakes. On the other hand, the proposed method also accelerates the rate of convergence. It can be applied to extract a single object in real images. The experimental results show that the proposed method is effective and efficient.

Hybrid structural and texture distinctiveness vector field convolution for region segmentation

The segmentation of objects has been an area of interest in numerous fields. The use of texture has been explored to improve convergence in the presence of cluttered backgrounds or objects with distinct textures, where intensity variations are insufficient. Additionally, saliency and feature maps have been applied for contour initialization. However, taking advantage of texture to improve initialization and convergence has not been extensively explored. To address this, we propose a hybrid structural and texture distinctiveness vector field convolution (STVFC) approach, where both the structural characteristics and the concept of texture distinctiveness are incorporated into a multi-functional vector field convolution (VFC) model. In this novel approach, texture distinctiveness is used to enable automatic initialization and is incorporated with intensity variation to improve and accelerate convergence towards the object boundary. Experiments using three datasets, containing natural images and Brodatz textures, demonstrated that STVFC achieved better or comparable segmentation accuracy.

An implicit evolution scheme for active contours and surfaces based on IIR filtering

In this work, we present an approach for implementing an implicit scheme for the numerical solution of the partial differential equation of the evolution of an active contour/surface. The proposed scheme is applicable to any variant of the traditional active contour (AC), irrespectively of the calculation of the image-based force field and it is readily applicable to explicitly parameterized active surfaces (AS). The proposed approach is formulated as an infinite impulse response (IIR) filtering of the coordinates of the contour/surface points. The poles of the filter are determined by the parameters controlling the shape of the active contour/surface. We show that the proposed IIR-based implicit evolution scheme has very low complexity. Furthermore, the proposed scheme is numerically stable, thus it allows the convergence of the AC/AS with significantly fewer iterations than the explicit evolution scheme. It also possesses the separability property along the two parameters of the AS, thus it may be applied to deformable surfaces, without the need to store and invert large sparse matrices. We implemented the proposed IIR-based implicit evolution scheme in the Vector Field Convolution (VFC) AC/AS using synthetic and clinical volumetric data. We compared the segmentation results with those of the explicit AC/AS evolution, in terms of accuracy and efficiency. Results show that the VFC AC/AS with the proposed IIR-based implicit evolution scheme achieves the same segmentation results with the explicit scheme, with considerably less computation time.

A Novel Framework for Segmentation of Secretory Granules in Electron Micrographs

It is still a standard practice for biologists to manually analyze transmission electron microscopy images. This is not only time consuming but also not reproducible and prone to induce subjective bias. For large-scale studies of insulin granules inside beta cells of the islet of Langerhans, an automated method for analysis is essential. Due to the complex structure of the images, standard microscopy segmentation techniques cannot be applied. We present a new approach to segment and measure transmission electron microscopy images of insulin granule cores and membranes from beta cells of rat islets of Langerhans. The algorithm is separated into two broad components, core segmentation and membrane segmentation. Core segmentation proceeds through three steps: pre-segmentation using a novel level-set active contour, morphological cleaning and a refining segmentation on each granule using a novel dual level-set active contour. Membrane segmentation is achieved in four steps: morphological cleaning, membrane sampling and scaling, vector field convolution for gap filling and membrane verification using a novel convergence filter. We show results from our algorithm alongside popular microscopy segmentation methods; the advantages of our method are demonstrated. Our algorithm is validated by comparing automated results to a manually defined ground truth. When the number of granules detected is compared to the number of granules in the ground truth a precision of 91% and recall of 87% is observed. The average granule areas differ by 13.35% and 6.08% for core and membranes respectively, when compared to the average areas of the ground truth. These results compare favorably to previously published data.

A Biased Vector Field Convolution External Force for Snakes

The vector field convolution (VFC) is an effective external force for active contour models. However, it always comes across premature convergence in extracting complex geometries, especially narrow and deep concavity when the initial contour is set outside of the object boundary. In this letter, a biased vector field convolution (BVFC) external force is proposed. In BVFC, an indicator function with respect to the contour and a narrow band are introduced to biasedly utilize the edges gradient information of a concave region. On the other hand, a feature map which better describes the principal curvatures and equally emphasizes both corners and edges is also introduced. Experimental results demonstrate that the BVFC snake improves the performance in extracting object boundary and shows the ability to converge to concavity compared with several state-of-art active contour models.

Extended Topological Active Nets

Topological Active Nets are promising parametric deformable models that integrate features of region-based and boundary-based segmentation techniques. Problems associated with the complexity of the model, however, have limited their utility. This paper introduces an extension of the model, defining a new behavior for changing its topology, as well as a novel external force definition and a new local search optimization procedure. In particular, we propose a new automatic pre-processing phase, a new external energy term based on the Extended Vector Field Convolution, node movement constraints to avoid crossing links, and different procedures to perform link cuts and hole detection. Moreover, the new local search procedure also incorporates heuristics to correct the position of eventually misplaced nodes. The proposal has been tested on 18 synthetic images which present different segmentation difficulties along with 3 real medical images. Its performance has been compared with that of the original Topological Active Net optimization approach along with both state-of-the-art parametric and geometric active contours: two snakes (based on Gradient Vector Flow and Vector Field Convolution), and two level sets (Chan and Vese, and Geodesic Active Contour). Our new method outperforms all the others for the given image sets, in terms of segmentation accuracy measured by using four standard segmentation metrics.

FULLY AUTOMATIC SEGMENTATION OF LEFT VENTRICLE IN A SEQUENCE OF ECHOCARDIOGRAPHY IMAGES OF ONE CARDIAC CYCLE BY DYNAMIC DIRECTIONAL VECTOR FIELD CONVOLUTION (DDVFC) METHOD AND MANIFOLD LEARNING

In this paper, an automatic method for segmentation of the left ventricle in two-dimensional (2D) echocardiography images of one cardiac cycle is proposed. In the first step of this method, using a mean image of a sequence of echocardiography images and its statistical properties the approximate region of left ventricle (LV) is extracted. Then the coordinate of extracted rectangular (ROI) is applied on all frames of sequences automatically. The mean image extracted ROI is used for defining the initial contour by scanning from the center point in polar coordinate. In the next step, all the extracted ROIs from the frames are mapped in a 2D space using the nonlinear dimension reduction manifold learning method. Using the properties of the manifold map end diastole (ED) and end systole (ES) frames are determined. Segmentation of the frames begins from ES frame, utilizing the dynamic directional vector field convolution (DDVFC) level set method with the initial contour as mentioned above. Final contour of each segmented frame is used as the initial contour of the next frame. Maximum range of the active contour motion is limited by a percent of the Euclidean distance between the point corresponds the current frame and the previous one in the resultant manifold. The results obtained from our method are quantitatively evaluated to those obtained by the gold contours drawn by a cardiologist on 489 echocardiographic images of seven volunteers using four distance measures: Hausdorff distance, average distance, area difference and area coverage error. We have also compared our results with the results of applying only DDVFC method. Comparing the implementation of only the DDVFC method, the results show final contours by proposed method are more close to contours drawn by a cardiologist.

Dynamically Constrained Vector Field Convolution for Active Contour Model

Dynamically Constrained Vector Field Convolution for Active Contour Model