Stroke is the third leading cause of death and a major cause of disability in the United States. Extracranial carotid artery disease is a major risk factor for stroke. Local hemodynamic forces are important in the development and progression of atherogenesis with areas of low and oscillatory wall shear stress (WSS) such as those occurring in the carotid bifurcation being more prone to atheroma development. Despite the importance of WSS in atherosclerosis, there is currently no practical means of measuring this variable clinically. Computational fluid dynamics (CFD) simulations of patient-specific models built from imaging data may provide a clinically relevant solution [1]. For CFD results to be clinically applicable, they need to replicate hemodynamic and imaging measurements to provide physiologic WSS values and the simulation and quantification process must be conducted in a time-frame consistent with the short duration needed for plaque and intima-media thickness assessment. LaDisa, Migrino and colleagues recently reported on a rapid and practical means of generating WSS maps associated with carotid atherosclerosis using patient-specific CFD models derived from 2D and Doppler ultrasound for flow information and MRI for 3D structure before and after 6 months of statin treatment [2]. Although these results were achieved after 17±8 hours/patient instead of days or weeks for prior models, model construction, quantification of results and simulation time were the most time consuming portions of the simulation process with CFD model construction being the most user-intensive portion of the process.

Quantitative measurements of wall thickness in human abdominal aortic aneurysms (AAAs) may provide useful information to predict rupture risk. Our procedure for estimating wall thickness in AAAs includes medical image segmentation and wall thickness detection. Image segmentation requires identifying and segmenting the luminal and outer wall boundaries of the blood vessels and wall thickness can be calculated by using intensity histograms and neural networks. The goal of this study is to develop an image-based, semi-automated method to trace the contours of the vessel wall and measure the wall thickness of the abdominal aorta from in-vivo, contrast-enhanced, CT images. An algorithm for the lumen and inner wall segmentations, and wall thickness detection was developed and tested on 10 ruptured and 10 unruptured AAAs. Reproducibility and repeatability of the algorithm were determined by comparing manual tracings made by two observers to contours made automatically by the algorithm itself. There was a high correspondence between automatic and manual area measurements for the lumen (r = 0.96) and between users (r = 0.98). Based on statistical analyses, the algorithm tends to underestimate the lumen area when compared to both observers.

The onset of arthritis is clearly associated with abnormal joint kinematics and contact pressures [1]. Yet our understanding of in vivo joint mechanics is still limited. In order to elucidate the relationship between joint mechanics and arthritis we must increase our knowledge of normal contact pressure distributions that help maintain healthy cartilage and abnormal contact pressure distributions that lead to arthritis. MRI-based modeling is a non-invasive means of determining joint mechanics in vivo , by combining information from MRI scans of joints with and without active functional loading.

When making computational fluid dynamics (CFD) based estimations of wall shear stress (WSS) in the human aorta, medical image converting processes to 3D geometries are important as the result is strongly dependent on the quality of the geometry [1]. The image interpretation process or segmentation can be more or less automated; however in clinical work today the gold standard is to manually interpret the medical image information. This combined magnetic resonance imaging (MRI) and CFD method aims to estimate WSS in human arteries in-vivo as WSS is strongly linked to atherosclerosis [2]. More or less automated segmentation has been used in previous studies but normally based on a stack of 2D individually segmented slices which is combined into a 3D model [3]. The aim of this work is to compare manual 2D and automatic 3D segmentations.

Segmentation of medical image data is often used for the construction of computational models to study the mechanics of diarthrodial joints such as the hip and knee. The analyst must demonstrate that the reconstructed geometry is an accurate representation of the true continuum to ensure model validity. This becomes especially important for computational modeling of joint contact, which requires accurate reconstruction of articular cartilage. Although volumetric computed tomography (CT) is often used to image diarthrodial joints, the lower bounds for detecting articular cartilage thickness and the influence of imaging parameters on the ability to image cartilage have not been reported. The use of contrast agent (CT arthrography) is necessary to visualize the surface of articular cartilage in live patients. Thus, it is of primary interest to quantify the accuracy of CT arthrography to demonstrate the feasibility of patient-specific modeling. The objectives of this study were to assess the accuracy and detection limits of CT for measuring simulated cartilage thickness using a phantom and to quantify changes in accuracy due to alterations in contrast agent concentration, imaging plane direction, spatial resolution and joint spacing.

Advancement in imaging technologies has shortened scan time of computed-tomography (CT). Recently introduced to a clinical market, 64-detector CT is capable of acquiring 64 simultaneous 0.5-mm slices with each 400-ms gantry revolution, yielding precise isotropic imaging of even peripheral airways within a single breath hold.

Human body is constantly under the influence of acceleration loads in environments such as combat flying. This study investigates the effect of body acceleration on human heart function by using finite element analysis. The nonlinear mechanical behavior of myocardium is modeled by Yeoh hyperelastic model. Stress-strain curves of myocardium are determined based on uniaxial compression tests on bovine heart samples. Nonlinear least square curve fitting is conducted in order to obtain material parameters. Heart geometrical modeling in three-dimension is done by segmentation of cardiac MRI images. Obtained material coefficients are assigned to the constructed heart model and appropriate pressure and acceleration loads are considered. By application of finite element method, stress distributions are calculated. Results indicate average stress in the internal wall of left ventricle is increased by 9.4% from 15.9 to 17.4kPa as the acceleration increases from +1G to +6G.

Recently MR image based computational models are being developed to assist targeted drug delivery in the brain by helping determine appropriate catheter position, drug dose among others to achieve the desired drug distribution [1–3]. Such a planning might be important to prevent damaging healthier tissues because many of the drugs (e.g. chemotherapeutic agents) are usually toxic and needs to be concentrated in specific regions of interest (e.g. tumor). However, for the image based model to make accurate predictions, it is important to segment the image and assign appropriate tissue properties such as hydraulic conductivity which are known to vary significantly within the brain. For example, it has been experimentally found that drugs injected into brain parenchyma get preferentially transported along the white matter tracts compared to the gray matter regions [4]. Segmenting MR images is a challenging task since the pixel intensities between different regions often overlap, hence traditional approaches based on thresholds might not provide reliable results. In this study, we used multi-layered perceptron (MLP) neural network to segment rat brain MR images into 3 different regions namely white matter (WM), gray matter (GM) and cerebrospinal fluid (CSF).

Medical image processing has emerged as a powerful way to simulate fluid flows through realistic models of complex patient-specific geometries without relying upon simplifying geometric approximations. However, image-based flow modeling processes traditionally involve several steps (e.g. image segmentation, surface mesh generation, volumetric flow mesh generation, and finally computational simulation) that must often be performed using separate pieces of software. This work presents an alternative methodology in which the entire image-based flow modeling process takes place on a Cartesian domain with the image embedded as an implicit surface, circumventing the need for complex surface meshes and body-fitted flow meshes. The complete framework is demonstrated with flow calculations performed in a computed tomography (CT) image reconstruction of an intracranial aneurysm (ICA). Flow calculations are compared against calculations performed following a standard multi-step route using the Vascular Modeling Toolkit (VMTK) [1, 2] and Fluent™ (Ansys, Inc., Lebanon, NH).

To date patient-specific analysis (both CFD and FEA) is limited by the complex pre-processing procedure needed for generating a patient-specific mesh [1,2]. Following the classical approach, CT / MRI images undergo first reconstruction/segmentation, leading to STL surfaces, and finally a mesh is generated. This multi-step procedure suffers from operator-dependence, is time consuming, does not allow a parametric mesh generation and requires several dedicated softwares. A novel approach is proposed for 3D angiographic coronary data, which allows generating structured parametric patient-specific mesh using a single open-source software, limiting the operator-dependence and the pre-processing time.

Previous efforts to simulate blood flow in patient-specific models either assumed rigid vessel walls or deformable walls with constant mechanical property [1]. We have developed a new workflow to enable blood flow and vessel dynamics simulations using subject-specific geometry and variable wall properties. The geometric model construction is based on 3D segmentation and geometric processing which greatly reduce human labor and increase the objectivity of the model. Variable wall properties are assigned to the model based on combining centerline-based and surface-based methods. This new approach was successfully applied to simulate blood flow and wall dynamics in models with abdominal, thoracic, and cerebral aneurysms.

The link between hemodynamic forces, notably wall shear stress (WSS), and atherogenesis is well established. Patient-specific computational fluid dynamics (CFD) modeling of human vasculature has attracted recent attention because it allows investigators to determine areas at risk for plaque formation and subsequent rupture [1]. Non-invasive in vivo imaging methods such as magnetic resonance (MR) angiography allow acquisition of vascular geometry and cross-sectional velocity such that a CFD model can determine the spatial distribution of WSS. WSS may then be correlated with phenomena such as wall thickening.

Osteoporosis is a major bone disease characterized by low bone mass and microarchitecture deterioration, which affects primarily trabecular sites and leads to increased bone fragility. Trabecular bone mechanical properties have direct relations with bone fragility. High-resolution image based-finite element (FE) models with the detailed 3D microstructure have been widely utilized to assess the mechanical properties of trabecular bone. Voxel-based FE model can be generated by converting individual voxels of high resolution bone images into 8-node brick elements. A number of studies have compared mechanical properties predicted by the voxel model with those by mechanical testing and have demonstrated that the voxel FE model can accurately predict the Young’s modulus and yield strength of human trabecular bone (1). However, the computational expense of the voxel-based technique, in general, limits its clinical applications, especially the nonlinear analysis for whole bone strength. Thus, it is not applicable to apply this technique to clinical use with the respect of current computer capability. There is apparent need for an alternative modeling approach that is more computationally efficient while preserving the accuracy of the predictions.

Annual mortality from ruptured abdominal aortic aneurysm (AAA) in the United States alone is approximately 150,000, which is currently ranked as the 13 th leading cause of death and the 10 th leading cause of death in men over 55 years of age [1]. The vascular surgeon needs to weigh the risk of AAA rupture against the risk of surgical intervention to decide the best course of treatment. Several steps are involved when using computational techniques to evaluate risk of rupture [2], namely medical image segmentation, 3D reconstruction, finite element mesh generation, derivation of boundary conditions, specification of tissue material properties, etc. Currently, computational analysis of AAA biomechanics includes the use of multiple third-party commercial software tools to accomplish each of these steps, which makes its clinical implementation impractical, time-consuming and requiring to interface multiple software tools as this demands an engineering skill set. Additionally, the versatility of general purpose off-the-shelf software comes at the cost of simplifying assumptions regarding geometric modeling, limited user control and boundary conditions. This makes subsequent computational results vulnerable to inaccuracies. In this work, we describe the software tool AAAVASC, built on a MATLAB platform, with an integrated approach for image-based modeling and a novel pipeline that facilitates both geometry quantification and computational analysis of AAA biomechanics.

Thoracic injury ranks second only to head injury in motor vehicle crash injuries in terms of the number of fatalities and serious injuries, the body region most often injured, and the overall economic cost [1, 2]. Skeletal and physiological resilience are known to decline with age, resulting in a decreased ability for the body to withstand traumatic insults [3]. Adults 65 years of age and older constitute more than 12% of the current population and with increases in life expectancy, the elderly population is projected to reach nearly 20% by 2030 [4].

The Modeling and Analysis group in the Division of Engineering at Mayo Clinic, Rochester works in the area of numerical modeling and experimental validation of cerebral aneurysms. In our modeling we use patient-specific geometries from imaging studies provided by Dr. Kallmes. The CFD analysis is performed with commercial packages Mimics (Materialise, Leuven, Belgium), for segmentation and ANSYS (ANSYS Inc. Canonsburg, PA) for meshing, simulation and post-processing. The experiments are conducted using a tomographic Particle Image Velocimetry system from LaVision Inc. The group is led by Dan Dragomir-Daescu, PhD and David F. Kallmes, MD and consists of two postdoctoral fellows and three engineers. For this project, CFD simulations were performed by postdoctoral fellow Simona Hodis, using the finite volume based solver Fluent.

A Matlab-based approach to image segmentation and mesh generation for creating high-quality hexagonal meshes is developed. The successful use of the procedure in patient-specific simulations of blood flow in a carotid artery is demonstrated.

Estimations of pressure-induced tension in the cerebral aneurysm sac wall may serve as metrics that help stratify patients according to rupture risk. Beyond segmentation of patient-specific aneurysm geometries, many modeling choices may be made in order to numerically estimate pressure-induced aneurysm wall tension. Some of these key choices are: 1) the material model and its parameters (this is unknown at a patient-specific level); 2) whether or not to accommodate bending deformation (membrane versus shell); 3) whether and how much of vasculature to include in domain of analysis; and 4) how to address the lack of zero-pressure geometry (forward versus inverse solution schemes). Often, the focus is on what the best of these modeling choices are. In an ideal world, appropriate modeling choices would be the use of an anisotropic finite elastic model [1–2], shell-like behavior [3], a large enough vasculature in the analysis domain to minimize edge-effects at the aneurysm sac, and an inverse solution scheme [3] to address the pre-deformed nature of the sac under in vivo imaging. But are these rigorous modeling choices really that necessary? Would compromising on these choices translate into significant error in the findings? We submit that, to the extent that the goal is to stratify patients according to sac wall tension, modeling choices are really secondary and will not translate into significant error in our interpretations. The objective of this study is to test the importance of three of these modeling choices — material model, truncation of vasculature in analysis domain and solution scheme (forward versus inverse) — in stratifying aneurysms based on wall tension using aneurysms in a study population.

Finite element analysis is a method of continuum mechanics that has been vital to the biomechanics community due to its ability to evaluate structures with complicated geometries. Generation of patient-specific finite element models from medical imaging data requires a series of steps including image segmentation, mesh generation, assignment of material properties, and definition of loading and boundary conditions. Each of these steps can be challenging and time-consuming. Here, we aim to address the challenge of automating the generation of hexahedral meshes.

Recent studies have shown that the maximum transverse diameter of an abdominal aortic aneurysm (AAA) and expansion rate are not entirely reliable indicators of rupture potential. We hypothesize that aneurysm morphology and wall thickness can be quantified in a systematic approach leading to accurate differentiation of the geometric characteristics of aneurysm population subsets. A non-invasive, image-based evaluation of AAA shape was implemented on a retrospective study of sixty-six subjects who underwent elective repair and twenty-eight subjects who suffered AAA rupture within 1 month of their last pre-operative follow-up. The contrast-enhanced computed tomography (CT) scans of these patients were used to generate three-dimensional models from the segmented images. Twenty-eight geometry-based indices were calculated to characterize the size and shape of the AAA sac, and regional variations in wall thickness were estimated based on a novel segmentation algorithm. A multivariate analysis of variance using a maximum AAA diameter of 5.5 cm as a factor was performed for all indices as dependent variables, for the electively repaired group. Box and Whisker plots and ROC curves were generated to determine the indices’ potential as predictors of rupture risk. Listed from highest to lowest area under the ROC curve (AUC), the following six indices were found statistically significant (p < 0.05): volume ( V , p < 0.0001), surface area ( S , p < 0.0001), intraluminal thrombus volume ( V ILT , p < 0.0001), diameter-to-diameter ratio ( DDr , p < 0.0001), diameter-to-height ratio ( DHr , p = 0.015), and centroid distance of the maximum diameter ( d c , p = 0.008). Given that individual AAAs have complex, tortuous and asymmetric shapes with local changes in surface curvature and wall thickness, the assessment of AAA rupture risk should require the accurate characterization of aneurysmal sac shape.