ABSTRACT
BACKGROUND: Time-resolved, three-dimensional phase-contrast magnetic resonance imaging (4D flow MRI) plays an important role in assessing cardiovascular diseases. However, the manual or semi-automatic segmentation of aortic vessel boundaries in 4D flow data introduces variability and limits reproducibility of aortic hemodynamics visualization and quantitative flow-related parameter computation. This paper explores the potential of deep learning to improve 4D flow MRI segmentation by developing models for automatic segmentation and analyzes the impact of the training data on the generalization of the model across different sites, scanner vendors, sequences, and pathologies. METHODS: The study population consists of 260 4D flow MRI datasets, including subjects without known aortic pathology, healthy volunteers, and patients with bicuspid aortic valve (BAV) examined at different hospitals. The dataset was split to train segmentation models on subsets with different representations of characteristics such as pathology, gender, age, scanner model, vendor, and field strength. An enhanced 3D U-net convolutional neural network (CNN) architecture with residual units was trained for 2D+t aortic cross-sectional segmentation. The model performance was evaluated using Dice score, Hausdorff distance, and average symmetric surface distance on test data, datasets with characteristics not represented in the training set (model-specific), and an overall evaluation set. Standard diagnostic flow parameters were computed and compared with manual segmentation results using Bland-Altman analysis and interclass correlation. RESULTS: The representation of technical factors such as scanner vendor and field strength in the training dataset had the strongest influence on the overall segmentation performance. Age had a greater impact than gender. Models solely trained on BAV patients' datasets performed well on datasets of healthy subjects but not vice versa. CONCLUSION: This study highlights the importance of considering a heterogeneous dataset for the training of widely applicable automatic CNN segmentations in 4D flow MRI, with a particular focus on the inclusion of different pathologies and technical aspects of data acquisition.
ABSTRACT
PURPOSE: Numerical phantom methods are widely used in the development of medical imaging methods. They enable quantitative evaluation and direct comparison with controlled and known ground truth information. Cardiac magnetic resonance has the potential for a comprehensive evaluation of the mitral valve (MV). The goal of this work is the development of a numerical simulation framework that supports the investigation of MRI imaging strategies for the mitral valve. METHODS: We present a pipeline for synthetic image generation based on the combination of individual anatomical 3D models with a position-based dynamics simulation of the mitral valve closure. The corresponding images are generated using modality-specific intensity models and spatiotemporal sampling concepts. We test the applicability in the context of MRI imaging strategies for the assessment of the mitral valve. Synthetic images are generated with different strategies regarding image orientation (SAX and rLAX) and spatial sampling density. RESULTS: The suitability of the imaging strategy is evaluated by comparing MV segmentations against ground truth annotations. The generated synthetic images were compared to ones acquired with similar parameters, and the result is promising. The quantitative analysis of annotation results suggests that the rLAX sampling strategy is preferable for MV assessment, reaching accuracy values that are comparable to or even outperform literature values. CONCLUSION: The proposed approach provides a valuable tool for the evaluation and optimization of cardiac valve image acquisition. Its application to the use case identifies the radial image sampling strategy as the most suitable for MV assessment through MRI.
Subject(s)
Mitral Valve Insufficiency , Mitral Valve , Humans , Mitral Valve/diagnostic imaging , Computer Simulation , Mitral Valve Insufficiency/diagnostic imaging , Magnetic Resonance Imaging , Phantoms, ImagingABSTRACT
OBJECTIVES: In aortic valve replacement (AVR), the treatment strategy as well as the model and size of the implanted prosthesis have a major impact on the postoperative hemodynamics and thus on the clinical outcome. Preinterventional prediction of the hemodynamics could support the treatment decision. Therefore, we performed paired virtual treatment with transcatheter AVR (TAVI) and biological surgical AVR (SAVR) and compared hemodynamic outcomes using numerical simulations. METHODS: 10 patients with severe aortic stenosis (AS) undergoing TAVI were virtually treated with both biological SAVR and TAVI to compare post-interventional hemodynamics using numerical simulations of peak-systolic flow. Virtual treatment procedure was done using an in-house developed tool based on position-based dynamics methodology, which was applied to the patient's anatomy including LVOT, aortic root and aorta. Geometries were automatically segmented from dynamic CT-scans and patient-specific flow rates were calculated by volumetric analysis of the left ventricle. Hemodynamics were assessed using the STAR CCM+ software by solving the RANS equations. RESULTS: Virtual treatment with TAVI resulted in realistic hemodynamics comparable to echocardiographic measurements (median difference in transvalvular pressure gradient [TPG]: -0.33 mm Hg). Virtual TAVI and SAVR showed similar hemodynamic functions with a mean TPG with standard deviation of 8.45 ± 4.60 mm Hg in TAVI and 6.66 ± 3.79 mm Hg in SAVR (p = 0.03) while max. Wall shear stress being 12.6 ± 4.59 vs. 10.2 ± 4.42 Pa (p = 0.001). CONCLUSIONS: Using the presented method for virtual treatment of AS, we were able to reliably predict post-interventional hemodynamics. TAVI and SAVR show similar hemodynamics in a pairwise comparison.
Subject(s)
Aortic Valve Stenosis , Heart Valve Prosthesis Implantation , Transcatheter Aortic Valve Replacement , Humans , Aortic Valve/diagnostic imaging , Aortic Valve/surgery , Transcatheter Aortic Valve Replacement/adverse effects , Heart Valve Prosthesis Implantation/adverse effects , Aortic Valve Stenosis/surgery , Treatment Outcome , Hemodynamics , Risk FactorsABSTRACT
Background: Transcatheter edge-to-edge repair (TEER) has developed from innovative technology to an established treatment strategy of mitral regurgitation (MR). The risk of iatrogenic mitral stenosis after TEER is, however, a critical factor in the conflict of interest between maximal reduction of MR and minimal impairment of left ventricular filling. We aim to investigate systematically the impact of device position on the post treatment hemodynamic outcome by involving the patient-specific segmentation of the diseased mitral valve. Materials and methods: Transesophageal echocardiographic image data of ten patients with severe MR (age: 57 ± 8 years, 20% female) were segmented and virtually treated with TEER at three positions by using a position based dynamics approach. Pre- and post-interventional patient geometries were preprocessed for computational fluid dynamics (CFD) and simulated at peak-diastole with patient-specific blood flow boundary conditions. Simulations were performed with boundary conditions mimicking rest and stress. The simulation results were compared with clinical data acquired for a cohort of 21 symptomatic MR patients (age: 79 ± 6 years, 43% female) treated with TEER. Results: Virtual TEER reduces the mitral valve area (MVA) from 7.5 ± 1.6 to 2.6 ± 0.6 cm2. Central device positioning resulted in a 14% smaller MVA than eccentric device positions. Furthermore, residual MVA is better predictable for central than for eccentric device positions (R 2 = 0.81 vs. R 2 = 0.49). The MVA reduction led to significantly higher maximal diastolic velocities (pre: 0.9 ± 0.2 m/s, post: 2.0 ± 0.5 m/s) and pressure gradients (pre: 1.5 ± 0.6 mmHg, post: 16.3 ± 9 mmHg) in spite of a mean flow rate reduction by 23% due to reduced MR after the treatment. On average, velocities were 12% and pressure gradients were 25% higher with devices in central compared to lateral or medial positions. Conclusion: Virtual TEER treatment combined with CFD is a promising tool for predicting individual morphometric and hemodynamic outcomes. Such a tool can potentially be used to support clinical decision making, procedure planning, and risk estimation to prevent post-procedural iatrogenic mitral stenosis.
ABSTRACT
PURPOSE: Decision support systems for mitral valve disease are an important step toward personalized surgery planning. A simulation of the mitral valve apparatus is required for decision support. Building a model of the chordae tendineae is an essential component of a mitral valve simulation. Due to image quality and artifacts, the chordae tendineae cannot be reliably detected in medical imaging. METHODS: Using the position-based dynamics framework, we are able to realistically simulate the opening and closing of the mitral valve. Here, we present a heuristic method for building an initial chordae model needed for a successful simulation. In addition to the heuristic, we present an interactive editor to refine the chordae model and to further improve pathology reproduction as well as geometric approximation of the closed valve. RESULTS: For evaluation, five mitral valves were reconstructed based on image sequences of patients scheduled for mitral valve surgery. We evaluated the approximation of the closed valves using either just the heuristic chordae model or a manually refined model. Using the manually refined models, prolapse was correctly reproduced in four of the five cases compared to two of the five cases when using the heuristic. In addition, using the editor improved the approximation in four cases. CONCLUSIONS: Our approach is suitable to create realistically parameterized mitral valve apparatus reconstructions for the simulation of normally and abnormally closing valves in a decision support system.
Subject(s)
Chordae Tendineae/surgery , Computer Simulation , Mitral Valve Insufficiency/surgery , Mitral Valve/surgery , Models, Anatomic , Chordae Tendineae/pathology , Humans , Mitral Valve/pathology , Mitral Valve Insufficiency/pathologyABSTRACT
PURPOSE: For planning and guidance of minimally invasive mitral valve repair procedures, 3D+t transesophageal echocardiography (TEE) sequences are acquired before and after the intervention. The valve is then visually and quantitatively assessed in selected phases. To enable a quantitative assessment of valve geometry and pathological properties in all heart phases, as well as the changes achieved through surgery, we aim to provide a new 4D segmentation method. METHODS: We propose a tracking-based approach combining gradient vector flow (GVF) and position-based dynamics (PBD). An open-state surface model of the valve is propagated through time to the closed state, attracted by the GVF field of the leaflet area. The PBD method ensures topological consistency during deformation. For evaluation, one expert in cardiac surgery annotated the closed-state leaflets in 10 TEE sequences of patients with normal and abnormal mitral valves, and defined the corresponding open-state models. RESULTS: The average point-to-surface distance between the manual annotations and the final tracked model was [Formula: see text]. Qualitatively, four cases were satisfactory, five passable and one unsatisfactory. Each sequence could be segmented in 2-6 min. CONCLUSION: Our approach enables to segment the mitral valve in 4D TEE image data with normal and pathological valve closing behavior. With this method, in addition to the quantification of the remaining orifice area, shape and dimensions of the coaptation zone can be analyzed and considered for planning and surgical result assessment.
Subject(s)
Cardiac Surgical Procedures/methods , Echocardiography, Four-Dimensional/methods , Echocardiography, Transesophageal/methods , Mitral Valve Insufficiency/diagnosis , Mitral Valve/diagnostic imaging , Surgery, Computer-Assisted/methods , Humans , Mitral Valve Insufficiency/surgeryABSTRACT
PURPOSE: The impact and the development of aneurysms depend to a significant degree on the exchange of liquid between the regular vessel and the pathological extension. A better understanding of this process will lead to improved prediction capabilities. The aim of the current study was to investigate fluid-exchange in aneurysm models of different complexities by combining microscopic magnetic resonance measurements with numerical simulations. In order to evaluate the accuracy and applicability of these methods, the fluid-exchange process between the unaltered vessel lumen and the aneurysm phantoms was analyzed quantitatively using high spatial resolution. METHODS: Magnetic resonance flow imaging was used to visualize fluid-exchange in two different models produced with a 3D printer. One model of an aneurysm was based on histological findings. The flow distribution in the different models was measured on a microscopic scale using time of flight magnetic resonance imaging. The whole experiment was simulated using fast graphics processing unit-based numerical simulations. The obtained simulation results were compared qualitatively and quantitatively with the magnetic resonance imaging measurements, taking into account flow and spin-lattice relaxation. RESULTS: The results of both presented methods compared well for the used aneurysm models and the chosen flow distributions. The results from the fluid-exchange analysis showed comparable characteristics concerning measurement and simulation. Similar symmetry behavior was observed. Based on these results, the amount of fluid-exchange was calculated. Depending on the geometry of the models, 7% to 45% of the liquid was exchanged per second. CONCLUSIONS: The result of the numerical simulations coincides well with the experimentally determined velocity field. The rate of fluid-exchange between vessel and aneurysm was well-predicted. Hence, the results obtained by simulation could be validated by the experiment. The observed deviations can be caused by the noise in the measurement and by the limited resolution of the simulation. The resulting differences are small enough to allow reliable predictions of the flow distribution in vessels with stents and for pulsed blood flow.
Subject(s)
Aneurysm/physiopathology , Blood Circulation , Magnetic Resonance Imaging , Models, BiologicalABSTRACT
Alterations of the blood flow are associated with various cardiovascular diseases. Precise knowledge of the velocity distribution is therefore important for understanding these diseases and predicting the effect of different medical intervention schemes. The goal of this work is to estimate the precision with which the velocity field can be measured and predicted by studying two simple model geometries with NMR micro imaging and computational fluid dynamics. For these initial experiments, we use water as an ideal test medium. The phantoms consist of tubes simulating a straight blood vessel and a step between two tubes of different diameters, which can be seen as a minimal model of the situation behind a stenosis. For both models, we compare the experimental data with the numerical prediction, using the experimental boundary conditions. For the simpler model, we also compare the data to the analytical solution. As an additional validation, we determine the divergence of the velocity field and verify that it vanishes within the experimental uncertainties. We discuss the resulting precision of the simulation and the outlook for extending this approach to the analysis of specific cases of arteriovascular problems.
Subject(s)
Arteries/anatomy & histology , Image Processing, Computer-Assisted/methods , Magnetic Resonance Imaging/methods , Algorithms , Computer Simulation , Electromagnetic Fields , Humans , Phantoms, Imaging , Reproducibility of Results , Signal-To-Noise Ratio , SoftwareABSTRACT
The occlusional performance of sole endoluminal stenting of intracranial aneurysms is controversially discussed in the literature. Simulation of blood flow has been studied to shed light on possible causal attributions. The outcome, however, largely depends on the numerical method and various free parameters. The present study is therefore conducted to find ways to define parameters and efficiently explore the huge parameter space with finite element methods (FEMs) and lattice Boltzmann methods (LBMs). The goal is to identify both the impact of different parameters on the results of computational fluid dynamics (CFD) and their advantages and disadvantages. CFD is applied to assess flow and aneurysmal vorticity in 2D and 3D models. To assess and compare initial simulation results, simplified 2D and 3D models based on key features of real geometries and medical expert knowledge were used. A result obtained from this analysis indicates that a combined use of the different numerical methods, LBM for fast exploration and FEM for a more in-depth look, may result in a better understanding of blood flow and may also lead to more accurate information about factors that influence conditions for stenting of intracranial aneurysms.
Subject(s)
Intracranial Aneurysm/physiopathology , Models, Cardiovascular , Cerebrovascular Circulation , Computational Biology , Finite Element Analysis , Humans , Hydrodynamics , Imaging, Three-Dimensional , Intracranial Aneurysm/therapy , Regional Blood Flow , StentsABSTRACT
OBJECTIVE: As an initial step to a complex reconstruction model for virtual surgery, the present study was carried out to provide data on the prenatal cleft lip and palate uvular muscle in eight specimens. METHOD: Serial sections of viscerocrania of 18 aborted embryos and fetuses were studied microscopically and segmented manually. Registration, three-dimensional reconstruction, and finite element analyses were conducted. RESULTS: Incompletely clefted uvulae showed anterior fusion and dorsal fission of the bilateral uvular muscle bodies. A complete cleft lip and palate specimen evidenced single bilateral uvular muscle bodies lying medially and orally below the cleft shelf, its central longitudinal fibers running beneath the oral-median mucosa. In incompletely clefted uvulae, 10% to 50% of circular peripheral fibers crossed the midline within the central third of the anterioposterior muscle, behind the levator loop. Of the fibers, 30% to 60% crossed to the ipsilateral palatopharyngeus muscle. Fibers inserted into the uvular basal membrane in a 60% nasal and 40% oral distribution at the middle third of the macroscopically clefted uvula. The macroscopic uvula itself consisted of loose connective tissue and salivary glands. Deformation analysis did disclose local stress, suggesting the uvular muscle contributes to velopharyngeal closure in normal anatomy and extends the cleft edges in cleft palate. CONCLUSION: Cleft lip and palate reconstruction should reasonably use the uvular muscle to augment the velar midline bulk. Uvular muscle deformation calculation was successful, permitting functional insight on the basis of microanatomical specimens, so far a bigger complete velar model can be ventured.
Subject(s)
Fetus/embryology , Finite Element Analysis , Palatal Muscles/embryology , Pharyngeal Muscles/embryology , Uvula/embryology , Artifacts , Cadaver , Fascia/embryology , Humans , Image Processing, Computer-Assisted , Imaging, Three-Dimensional , Staining and LabelingABSTRACT
OBJECTIVE: Virtual surgery and virtual patients necessitate quantitative data on the area of interest. The study was conducted to exactly describe the embryonic and fetal uvular muscle (MU), relevant for clinical as well as virtual surgery and virtual patient generation. METHOD: Serially sectioned viscerocrania of 10 aborted embryos and fetuses underwent three-dimensional reconstruction to obtain detailed anatomic data and perform finite element analyses. RESULTS: The MU was paired in 80% of cases, while 20% allowed no clear-cut distinction. The MU merged with the levator muscle beneath the palatal aponeurosis without a hard palate insertion. Superior longitudinal central fibers ran below the nasal mucosa, and few circular peripheral fibers crossed in the central third to the contralateral side. This was seen in 30% of the paired muscles and in all cases when no differentiation was possible; about 40% to 80% MU fibers crossed to the ipsilateral and contralateral palatopharyngeus muscle behind the levator loop. MU fibers inserted 60% nasal and 40% oral to the basal membrane at the middle third of the macroscopic uvula, made of loose connective tissue and salivary glands. The results of the finite element simulation of the uvula showed no distinct patterns or distributions of local stress. CONCLUSIONS: Detailed anatomical study supported the concept of mediocranial MU repositioning during corrective surgery, although the impact is minor to the levator muscle's action. Future mathematical models describing effects of such a maneuver should integrate surrounding structures.
Subject(s)
Fetus/embryology , Palatal Muscles/embryology , Pharyngeal Muscles/embryology , Uvula/embryology , Artifacts , Cadaver , Fascia/embryology , Finite Element Analysis , Humans , Image Processing, Computer-Assisted , Imaging, Three-Dimensional , Staining and LabelingABSTRACT
In this article, we present a new method for the generation of surface meshes of biological soft tissue. The method is based on the deformable surface model technique and is extended to histological data sets. It relies on an iterative adjustment towards polygonal segments describing the histological structures of the soft tissue. The generated surface meshes allow for the construction of volumetric meshes through a standard constrained Delaunay approach and, thus, for the application in finite element methods. The geometric properties of volumetric meshes have an immediate influence on the numerical conditioning and, therewith, on the stability of the finite element method and the convergence of iterative solvers. In this article, the influence of the surface meshes on the quality of the volumetric meshes is analysed in terms of the spectral condition number of the stiffness matrices, which are assembled within Newton's method. The non-linear material behavior of biological soft tissue is modeled by the Mooney-Rivlin material law. The subject is motivated by the requirements of virtual surgery.