A coronary artery CTA segmentation approach based on deep learning

Presence of plaque and coronary artery stenosis are the main causes of coronary heart disease. Detection of plaque and coronary artery segmentation have become the first choice in detecting coronary artery disease. The purpose of this study is to investigate a new method for plaque detection and automatic segmentation and diagnosis of coronary arteries and to test its feasibility of applying to clinical medical image diagnosis. A multi-model fusion coronary CT angiography (CTA) vessel segmentation method is proposed based on deep learning. The method includes three network layer models namely, an original 3-dimensional full convolutional network (3D FCN) and two networks that embed the attention gating (AG) model in the original 3D FCN. Then, the prediction results of the three networks are merged by using the majority voting algorithm and thus the final prediction result of the networks is obtained. In the post-processing stage, the level set function is used to further iteratively optimize the results of network fusion prediction. The JI (Jaccard index) and DSC (Dice similarity coefficient) scores are calculated to evaluate accuracy of blood vessel segmentations. Applying to a CTA dataset of 20 patients, accuracy of coronary blood vessel segmentation using FCN, FCN-AG1, FCN-AG2 network and the fusion method are tested. The average values of JI and DSC of using the first three networks are (0.7962, 0.8843), (0.8154, 0.8966) and (0.8119, 0.8936), respectively. When using new fusion method, average JI and DSC of segmentation results increase to (0.8214, 0.9005), which are better than the best result of using FCN, FCN-AG1 and FCN-AG2 model independently.

Effects of Selection of Inlet Perturbations, Multiphase and Turbulence Equations on Slug Flow Characteristics Using Altair® AcuSolve™

Selection of inlet perturbations, multiphase equations, and the turbulence equation may affect the development of slug flow using computational fluid dynamic simulation tools. The inlet perturbation, such as sinusoidal and random perturbations, play an essential role in inducing slug formation. Multiphase equations such as volume of fluid and level set methods are used to track and capture the gas-liquid immiscible interface. Similarly, turbulence equations such as Spalart Allmaras (SA), Detached Eddy Simulations (DES), k-omega, and k-epsilon can be used to predict the evolution of turbulence within the flow. At present, no direct comparison is available in the literature on the selection of (i) types of inlet perturbations, (ii) the choice of multiphase equations, and (iii) the turbulence equation on the development of slug flow using the Altair computational package. This article aims to compare the effects of the selection of inlet perturbations, multiphase models and turbulence equations on slug flow characteristics using Altair® AcuSolve™. The findings by Altair® simulation were compared to published experimental data and simulation works using ANSYS and STAR-CCM+. The slug flow characteristics of interest include slug morphology, a body length-to-diameter ratio, velocity, frequency, and pressure gradient. It was found that the slug flow could be developed for all combinations of settings. Although level set approach in Altair® can track fluid motion successfully, it has a limitation in modelling the convective transport of the multiphase mixture well, unlike ANSYS and STAR-CCM+. Compared to the standard level set method, the coupling of back-and-forth error compensation and correction with the level set function helps to capture the internal boundary more accurately by reducing errors caused by numerical diffusion in the transport of the level set. It was revealed that the Spalart Allmaras turbulence equation could mimic published experimental result better than DES as it produced the closest slug translational velocity. Since the frequency of the slugs for the developed models showed a good agreement with the published data, the models could be sufficient for the investigation of fluid-structure interaction.

A Microdosimetry Analysis of Reversible Electroporation In Scattered, Overlapping, And Cancerous Cervical Cells

In this paper,a numerical method for studying reversible electroporation on normal and cancerous cervical cells is introduced. This microdosimetry analysis has been done by a unique approach for extracting contours of free and overlapping cervical cells in the cluster from the External Depth field images19. The algorithm used for extracting the contours is a joint optimization of multiple level set function along with the Gaussian mixture model and Maximally Stable Extremal Regions. This contour is then imported a multiphysics domain solver, where variable frequency pulsed electric field is applied. The Trans-Membrane Voltage (TMV) developed across the cell membrane is then calculated using the Maxwell equation coupled with a statistical approach employing the asymptotic Smoluchowski equation, which calculates the generated temporal pore density. The numerical model was validated by successful replication of existing experimental approach that employed low-frequency uni-polar pulses on the overlapping cells to obtain reversible electroporation. Using several overlapping clumps of cervical cells, simulations are performed to match the experimental data. For high-frequency calculation, a combination of normal and cancerous cells is introduced to the computational domain. The cells are assumed to be dispersive and the Debye dispersion equation is a second-order partial derivative equation used for further calculations. The difference in time duration for reaching the threshold value of electroporation is seen between the normal and cancerous cervical cells due to their size and conductivity change. The drug and dye uptake modulation during the high-frequency electric field electroporation is advocated by a mathematical model.

Medical Image Segmentation Using a Combination of Lattice Boltzmann Method and Fuzzy Clustering Based on GPU CUDA Parallel Processing

The rapid development of computer technology has had a significant influence on advances in medical science. This development concerns segmenting medical images that can be used to help doctors diagnose patient diseases. The boundary between objects contained in an image is captured using the level set function. The equation of the level set function is solved numerically by combining the Lattice Boltzmann (LBM) method and fuzzy clustering. Parallel processing using a graphical processing unit (GPU) accelerates the execution of the segmentation process. The results showed that image segmentation with a relatively large size could be done quickly. The use of parallel programming with the GPU can accelerate up to 39.22 times compared to the speed of serial programming with the CPU. In addition, the comparisons with other research and benchmark data show consistent results.

A Cartesian Method with Second-Order Pressure Resolution for Incompressible Flows with Large Density Ratios

An Eulerian method to numerically solve incompressible bifluid problems with high density ratio is presented. This method can be considered as an improvement of the Ghost Fluid method, with the specificity of a sharp second-order numerical scheme for the spatial resolution of the discontinuous elliptic problem for the pressure. The Navier–Stokes equations are integrated in time with a fractional step method based on the Chorin scheme and discretized in space on a Cartesian mesh. The bifluid interface is implicitly represented using a level-set function. The advantage of this method is its simplicity to implement in a standard monofluid Navier–Stokes solver while being more accurate and conservative than other simple classical bifluid methods. The numerical tests highlight the improvements obtained with this sharp method compared to the reference standard first-order methods.

A topology optimization algorithm based on topological derivative and level-set function for designing phononic crystals

PurposeThis work presents a topology optimization methodology for designing microarchitectures of phononic crystals. The objective is to get microstructures having, as a consequence of wave propagation phenomena in these media, bandgaps between two specified bands. An additional target is to enlarge the range of frequencies of these bandgaps.Design/methodology/approachThe resulting optimization problem is solved employing an augmented Lagrangian technique based on the proximal point methods. The main primal variable of the Lagrangian function is the characteristic function determining the spatial geometrical arrangement of different phases within the unit cell of the phononic crystal. This characteristic function is defined in terms of a level-set function. Descent directions of the Lagrangian function are evaluated by using the topological derivatives of the eigenvalues obtained through the dispersion relation of the phononic crystal.FindingsThe description of the optimization algorithm is emphasized, and its intrinsic properties to attain adequate phononic crystal topologies are discussed. Particular attention is addressed to validate the analytical expressions of the topological derivative. Application examples for several cases are presented, and the numerical performance of the optimization algorithm for attaining the corresponding solutions is discussed.Originality/valueThe original contribution results in the description and numerical assessment of a topology optimization algorithm using the joint concepts of the level-set function and topological derivative to design phononic crystals.

Isogeometric analysis based on geometric reconstruction models

In isogeometric analysis (IGA), the boundary representation of computer-aided design (CAD) and the tensor-product non-uniform rational B-spline structure make the analysis of three-dimensional (3D) problems with irregular geometries difficult. In this paper, an IGA method for complex models is presented by reconstructing analysis-suitable models. The CAD model is represented by boundary polygons or point cloud and is embedded into a regular background grid, and a model reconstruction method is proposed to obtain the level set function of the approximate model, which can be directly used in IGA. Three 3D examples are used to test the proposed method, and the results demonstrate that the proposed method can deal with complex engineering parts reconstructed by boundary polygons or point clouds.

Magnetic Resonance Imaging Segmentation via Weighted Level Set Model Based on Local Kernel Metric and Spatial Constraint

Magnetic resonance imaging (MRI) segmentation is a fundamental and significant task since it can guide subsequent clinic diagnosis and treatment. However, images are often corrupted by defects such as low-contrast, noise, intensity inhomogeneity, and so on. Therefore, a weighted level set model (WLSM) is proposed in this study to segment inhomogeneous intensity MRI destroyed by noise and weak boundaries. First, in order to segment the intertwined regions of brain tissue accurately, a weighted neighborhood information measure scheme based on local multi information and kernel function is designed. Then, the membership function of fuzzy c-means clustering is used as the spatial constraint of level set model to overcome the sensitivity of level set to initialization, and the evolution of level set function can be adaptively changed according to different tissue information. Finally, the distance regularization term in level set function is replaced by a double potential function to ensure the stability of the energy function in the evolution process. Both real and synthetic MRI images can show the effectiveness and performance of WLSM. In addition, compared with several state-of-the-art models, segmentation accuracy and Jaccard similarity coefficient obtained by WLSM are increased by 0.0586, 0.0362 and 0.1087, 0.0703, respectively.