Using machine learning to characterize heart failure across the scales.

Heart failure is a progressive chronic condition in which the heart undergoes detrimental changes in structure and function across multiple scales in time and space. Multiscale models of cardiac growth can provide a patient-specific window into the progression of heart failure and guide personalized treatment planning. Yet, the predictive potential of cardiac growth models remains poorly understood. Here, we quantify predictive power of a stretch-driven growth model using a chronic porcine heart failure model, subject-specific multiscale simulation, and machine learning techniques. We combine hierarchical modeling, Bayesian inference, and Gaussian process regression to quantify the uncertainty of our experimental measurements during an 8-week long study of volume overload in six pigs. We then propagate the experimental uncertainties from the organ scale through our computational growth model and quantify the agreement between experimentally measured and computationally predicted alterations on the cellular scale. Our study suggests that stretch is the major stimulus for myocyte lengthening and demonstrates that a stretch-driven growth model alone can explain [Formula: see text] of the observed changes in myocyte morphology. We anticipate that our approach will allow us to design, calibrate, and validate a new generation of multiscale cardiac growth models to explore the interplay of various subcellular-, cellular-, and organ-level contributors to heart failure. Using machine learning in heart failure research has the potential to combine information from different sources, subjects, and scales to provide a more holistic picture of the failing heart and point toward new treatment strategies.

Multidisciplinary decision-making in mitral valve disease: the mitral valve heart team.

BACKGROUND: Although decision-making using the heart-team approach is apparently intuitive and has a class I recommendation in most recent guidelines, supportive data is still lacking. The current study aims to demonstrate the individualised clinical pathway for mitral valve disease patients and to evaluate the outcome of all patients referred to the dedicated mitral valve heart team. METHODS: All patients who were evaluated for mitral valve pathology with or without concomitant cardiac disease between 1 January 2016 and 31 December 2016 were prospectively followed and included. Patients were evaluated, and a treatment strategy was determined by the dedicated mitral valve heart team. RESULTS: One hundred and fifty-eight patients were included; 67 patients were treated surgically (isolated and concomitant surgery), 20 by transcatheter interventions and 71 conservatively. Surgically treated patients had a higher 30-day mortality rate (4.4%), which decreased when specified to a dedicated surgeon (1.7%) and in primary, elective cases (0%). This was also observed for major adverse events within 30 days. Residual mitral regurgitation >grade 2 was more frequent in the catheter-based intervention group (23.5%) compared to the surgical group (4.8%). CONCLUSION: In conclusion, the implementation of a multidisciplinary heart team for mitral valve disease is a valuable approach for the selection of patients for different treatment modalities. Our research group will focus on a future comparative study using historical cohorts to prove the potential superiority of the dedicated multidisciplinary heart-team approach.

Use of a right ventricular continuous flow pump to validate the distensible model of the pulmonary vasculature.

In the pulmonary circulation, resistive and compliant properties overlap in the same vessels. Resistance varies nonlinearly with pressure and flow; this relationship is driven by the elastic properties of the vessels. Linehan et al. correlated the mean pulmonary arterial pressure and mean flow with resistance using an original equation incorporating the distensibility of the pulmonary arteries. The goal of this study was to validate this equation in an in vivo porcine model. In vivo measurements were acquired in 6 pigs. The distensibility coefficient (DC) was measured by placing piezo-electric crystals around the pulmonary artery (PA). In addition to experiments under pulsatile conditions, a right ventricular (RV) bypass system was used to induce a continuous pulmonary flow state. The Linehan et al. equation was then used to predict the pressure from the flow under continuous flow conditions. The diameter-derived DC was 2.4%/mmHg (+/-0.4%), whereas the surface area-based DC was 4.1 %/mmHg (+/-0.1%). An increase in continuous flow was associated with a constant decrease in resistance, which correlated with the diameter-based DC (r=-0.8407, p=0.044) and the surface area-based DC (r=-0.8986, p=0.028). In contrast to the Linehan et al. equation, our results showed constant or even decreasing pressure as flow increased. Using a model of continuous pulmonary flow induced by an RV assist system, pulmonary pressure could not be predicted based on the flow using the Linehan et al. equation. Measurements of distensibility based on the diameter of the PA were inversely correlated with the resistance.

A finite element model to study the effect of tissue anisotropy on ex vivo arterial shear wave elastography measurements.

Shear wave elastography (SWE) is an ultrasound (US) diagnostic method for measuring the stiffness of soft tissues based on generated shear waves (SWs). SWE has been applied to bulk tissues, but in arteries it is still under investigation. Previously performed studies in arteries or arterial phantoms demonstrated the potential of SWE to measure arterial wall stiffness-a relevant marker in prediction of cardiovascular diseases. This study is focused on numerical modelling of SWs in ex vivo equine aortic tissue, yet based on experimental SWE measurements with the tissue dynamically loaded while rotating the US probe to investigate the sensitivity of SWE to the anisotropic structure. A good match with experimental shear wave group speed results was obtained. SWs were sensitive to the orthotropy and nonlinearity of the material. The model also allowed to study the nature of the SWs by performing 2D FFT-based and analytical phase analyses. A good match between numerical group velocities derived using the time-of-flight algorithm and derived from the dispersion curves was found in the cross-sectional and axial arterial views. The complexity of solving analytical equations for nonlinear orthotropic stressed plates was discussed.

Patient-specific CFD models for intraventricular flow analysis from 3D ultrasound imaging: Comparison of three clinical cases.

BACKGROUND: As the intracardiac flow field is affected by changes in shape and motility of the heart, intraventricular flow features can provide diagnostic indications. Ventricular flow patterns differ depending on the cardiac condition and the exploration of different clinical cases can provide insights into how flow fields alter in different pathologies. METHODS: In this study, we applied a patient-specific computational fluid dynamics model of the left ventricle and mitral valve, with prescribed moving boundaries based on transesophageal ultrasound images for three cardiac pathologies, to verify the abnormal flow patterns in impaired hearts. One case (P1) had normal ejection fraction but low stroke volume and cardiac output, P2 showed low stroke volume and reduced ejection fraction, P3 had a dilated ventricle and reduced ejection fraction. RESULTS: The shape of the ventricle and mitral valve, together with the pathology influence the flow field in the left ventricle, leading to distinct flow features. Of particular interest is the pattern of the vortex formation and evolution, influenced by the valvular orifice and the ventricular shape. The base-to-apex pressure difference of maximum 2mmHg is consistent with reported data. CONCLUSION: We used a CFD model with prescribed boundary motion to describe the intraventricular flow field in three patients with impaired diastolic function. The calculated intraventricular flow dynamics are consistent with the diagnostic patient records and highlight the differences between the different cases. The integration of clinical images and computational techniques, therefore, allows for a deeper investigation intraventricular hemodynamics in patho-physiology.

Patient-specific CFD simulation of intraventricular haemodynamics based on 3D ultrasound imaging.

BACKGROUND: The goal of this paper is to present a computational fluid dynamic (CFD) model with moving boundaries to study the intraventricular flows in a patient-specific framework. Starting from the segmentation of real-time transesophageal echocardiographic images, a CFD model including the complete left ventricle and the moving 3D mitral valve was realized. Their motion, known as a function of time from the segmented ultrasound images, was imposed as a boundary condition in an Arbitrary Lagrangian-Eulerian framework. RESULTS: The model allowed for a realistic description of the displacement of the structures of interest and for an effective analysis of the intraventricular flows throughout the cardiac cycle. The model provides detailed intraventricular flow features, and highlights the importance of the 3D valve apparatus for the vortex dynamics and apical flow. CONCLUSIONS: The proposed method could describe the haemodynamics of the left ventricle during the cardiac cycle. The methodology might therefore be of particular importance in patient treatment planning to assess the impact of mitral valve treatment on intraventricular flow dynamics.

Unstructured hexahedral mesh generation of complex vascular trees using a multi-block grid-based approach.

The trend towards realistic numerical models of (pathologic) patient-specific vascular structures brings along larger computational domains and more complex geometries, increasing both the computation time and the operator time. Hexahedral grids effectively lower the computational run time and the required computational infrastructure, but at high cost in terms of operator time and minimal cell quality, especially when the computational analyses are targeting complex geometries such as aneurysm necks, severe stenoses and bifurcations. Moreover, such grids generally do not allow local refinements. As an attempt to overcome these limitations, a novel approach to hexahedral meshing is proposed in this paper, which combines the automated generation of multi-block structures with a grid-based method. The robustness of the novel approach is tested on common complex geometries, such as tree-like structures (including trifurcations), stenoses, and aneurysms. Additionally, the performance of the generated grid is assessed using two numerical examples. In the first example, a grid sensitivity analysis is performed for blood flow simulated in an abdominal mouse aorta and compared to tetrahedral grids with a prismatic boundary layer. In the second example, the fluid-structure interaction in a model of an aorta with aortic coarctation is simulated and the effect of local grid refinement is analyzed.

Non-invasive, energy-based assessment of patient-specific material properties of arterial tissue.

The mechanical properties of human biological tissue vary greatly. The determination of arterial material properties should be based on experimental data, i.e. diameter, length, intramural pressure, axial force and stress-free geometry. Currently, clinical data provide only non-invasively measured pressure-diameter data for superficial arteries (e.g. common carotid and femoral artery). The lack of information forces us to take into account certain assumptions regarding the in situ configuration to estimate material properties in vivo. This paper proposes a new, non-invasive, energy-based approach for arterial material property estimation. This approach is compared with an approach proposed in the literature. For this purpose, a simplified finite element model of an artery was used as a mock experimental situation. This method enables exact knowledge of the actual material properties, thereby allowing a quantitative evaluation of material property estimation approaches. The results show that imposing conditions on strain energy can provide a good estimation of the material properties from the non-invasively measured pressure and diameter data.

Influence of valve size, orientation and downstream geometry of an aortic BMHV on leaflet motion and clinically used valve performance parameters.

The aim of this study was to reconcile some of our own previous work and the work of others to generate a physiologically realistic numerical simulation environment that allows to virtually assess the performance of BMHVs. The model incorporates: (i) a left ventricular deformable model to generate a physiological inflow to the aortic valve; (ii) a patient-specific aortic geometry (root, arch and descending aorta); (iii) physiological pressure and flow boundary conditions. We particularly studied the influence of downstream geometry, valve size and orientation on leaflet kinematics and functional indices used in clinical routine. Compared to the straight tube geometry, the patient-specific aorta leads to a significant asynchronous movement of the valve, especially during the closing of the valve. The anterior leaflet starts to close first, impacts the casing at the closed position and remains in this position. At the same time, the posterior leaflet impacts the pivoting mechanisms at the fully open position. At the end of systole, this leaflet subsequently accelerates to the closed position, impacting the casing with an angular velocity of approximately -477 rad/s. The valve size greatly influences the transvalvular pressure gradient (TPG), but does not change the overall leaflet kinematics. This is in contrast to changes in valve orientation, where changing valve orientation induces large differences in leaflet kinematics, but the TPG remains approximately the same.

Engineering point of view on liver transplantation strategies: multi-level modeling of hepatic perfusion.

BACKGROUND: Hepatic perfusion plays a crucial role in liver transplantation strategies, for example, when preserving procured organs with the use of machine perfusion preservation (MP) and in the case of living donor liver transplantation (LDLT). Liver hemodynamics are not yet fully understood because of insufficient knowledge on the hepatic vascular morphology and its perfusion characteristics, hampering the optimization of liver transplantation procedures. To this end, we developed computer models to simulate the complex blood circulation through the liver from the macro-scale down to the terminal micro-scale level. METHODS: A combination of state-of-the-art techniques (vascular corrosion casting, micro-CT scanning up to a 2.6-µm resolution, and image processing) led to 3D visualizations and detailed geometrical analyses of the complex architecture of the liver's 3 vascular trees, ranging from the largest vessels (macrocirculation) down to the sinusoids (microcirculation). RESULTS: On the basis of these data, we developed various computational models (electrical analog models and 3D computational fluid dynamics models) to study the blood flow-induced forces acting on the hepatic blood vessels. The latter was done for physiological blood flow through the liver as well as for livers undergoing MP or LDLT procedures. Hereby, several scenarios were simulated to study the behavior of livers in different hemodynamic circumstances. CONCLUSIONS: A novel, multi-level modeling framework was developed to simulate hepatic perfusion in support of liver transplantation strategies. We obtained unique anatomical data on the vascular architecture of both human and rat livers. These data formed the building blocks of electrical analog models of hepatic perfusion and numerical models of the liver microcirculation. The results revealed novel insights into the hemodynamic impact of liver MP and LDLT procedures as well as into the microcirculatory perfusion characteristics. The presented methodology is also applicable to other tree-like structures (eg, the biliary tree) or organs (eg, kidneys, lungs).

What if you stretch the IFU? A mechanical insight into stent graft Instructions For Use in angulated proximal aneurysm necks.

Endovascular treatment for patients with a proximal neck anatomy outside instructions for use is an ongoing topic of debate in endovascular aneurysm repair. This paper employs the finite element method to offer insight into possible adverse effects of deploying a stent graft into an angulated geometry. The effect of angulation, straight neck length and device oversize was investigated in a full factorial parametric analysis. Stent apposition, area reduction of the graft, asymmetry of contact forces and the ability to find a good seal were investigated. Most adverse effects are expected for combinations of high angulation and short straight landing zones. Higher oversize has a beneficiary effect, but not enough to compensate the adverse effects of (very) short and angulated angles. Our analysis shows that for an angle between the suprarenal aorta and proximal neck above 60°, proximal kinking of the device can occur. The method used offers a engineering view on the morphological limits of EVAR for a clinically used device.

Validation of a numerical FSI simulation of an aortic BMHV by in vitro PIV experiments.

In this paper, a validation of a recently developed fluid-structure interaction (FSI) coupling algorithm to simulate numerically the dynamics of an aortic bileaflet mechanical heart valve (BMHV) is performed. This validation is done by comparing the numerical simulation results with in vitro experiments. For the in vitro experiments, the leaflet kinematics and flow fields are obtained via the particle image velocimetry (PIV) technique. Subsequently, the same case is numerically simulated by the coupling algorithm and the resulting leaflet kinematics and flow fields are obtained. Finally, the results are compared, revealing great similarity in leaflet motion and flow fields between the numerical simulation and the experimental test. Therefore, it is concluded that the developed algorithm is able to capture very accurately all the major leaflet kinematics and dynamics and can be used to study and optimize the design of BMHVs.

The influence of vascular anatomy on carotid artery stenting: a parametric study for damage assessment.

Carotid artery stenting is emerging as an alternative technique to surgery for the treatment of symptomatic severe carotid stenosis. Clinical and experimental evidence demonstrates that both plaque morphology and biomechanical changes due to the device implantation can be possible causes of an unsuccessful treatment. In order to gain further insights of the endovascular intervention, a virtual environment based on structural finite element simulations was built to emulate the stenting procedure on generalized atherosclerotic carotid geometries which included a damage model to quantify the injury of the vessel. Five possible lesion scenarios were simulated by changing both material properties and vascular geometrical features to cover both presumed vulnerable and stable plaques. The results were analyzed with respect to lumen gain and wall stresses which are potentially related to the failure of the procedure according to previous studies. Our findings show that an elliptic lumen shape and a thinner fibrous cap with an underlying lipid pool result in higher stenosis reduction, while large calcifications and fibrotic tissue are more prone to recoil. The shielding effect of a thicker fibrous cap helps to reduce local compressive stresses in the soft plaque. The presence of a soft plaque reduces the damage in the healthy vascular structures. Contrarily, the presence of hard plaque promotes less damage volume in the fibrous cap and reduces stress peaks in this region, but they seem to increase stresses in the media-intima layer. Finally the reliability of the achieved results was put into clinical perspective.

Filling the void: a coalescent numerical and experimental technique to determine aortic stent graft mechanics.

The presented study details a combined experimental and computational method to assess and compare the mechanical behavior of the main body of 4 different stent graft designs. The mechanical response to a flat plate compression and radial crimping of the devices is derived and related to geometrical and material features of different stent designs. The finite element modeling procedure is used to complement the experimental results and conduct a solution sensitivity study. Finite element evaluations of the mechanical behavior match well with experimental findings and are used as a quantitative basis to discuss design characteristics of the different devices.

Haemodynamic impact of stent-vessel (mal)apposition following carotid artery stenting: mind the gaps!

Carotid artery stenting (CAS) has emerged as a minimally invasive alternative to endarterectomy but its use in clinical treatment is limited due to the post-stenting complications. Haemodynamic actors, related to blood flow in the stented vessel, have been suggested to play a role in the endothelium response to stenting, including adverse reactions such as in-stent restenosis and late thrombosis. Accessing the flow-related shear forces acting on the endothelium in vivo requires space and time resolutions which are currently not achievable with non-invasive clinical imaging techniques but can be obtained from image-based computational analysis. In this study, we present a framework for accurate determination of the wall shear stress (WSS) in a mildly stenosed carotid artery after the implantation of a stent, resembling the commercially available Acculink (Abbott Laboratories, Abbott Park, Illinois, USA). Starting from angiographic CT images of the vessel lumen and a micro-CT scan of the stent, a finite element analysis is carried out in order to deploy the stent in the vessel, reproducing CAS in silico. Then, based on the post-stenting anatomy, the vessel is perfused using a set of boundary conditions: total pressure is applied at the inlet, and impedances that are assumed to be insensitive to the presence of the stent are imposed at the outlets. Evaluation of the CAS outcome from a geometrical and haemodynamic perspective shows the presence of atheroprone regions (low time-average WSS, high relative residence time) colocalised with stent malapposition and stent strut interconnections. Stent struts remain unapposed in the ostium of the external carotid artery disturbing the flow and generating abnormal shear forces, which could trigger thromboembolic events.

A computational exploration of helical arterio-venous graft designs.

Although arterio-venous grafts (AVGs) are the second best option as long-term vascular access for hemodialysis, they suffer from complications caused by intimal hyperplasia, mainly located in vessel regions of low and oscillating wall shear stress. However, certain flow patterns in the bulk may reduce these unfavorable hemodynamic conditions. We therefore studied, with computational fluid dynamics (CFD), the impact of a helical AVG design on the occurrence of (un)favorable hemodynamic conditions at the venous anastomosis. Six CFD-models of an AVG in closed-loop configuration were constructed: one conventional straight graft, and five helical designed grafts with a pitch of 105 mm down to 35 mm. At the venous anastomosis, disturbed shear was assessed by quantifying the area with unfavorable conditions, and by analyzing averaged values in a case-specific patch. The bulk hemodynamics were assessed by analyzing the kinetic helicity in and the pressure drop over the graft. The most helical design scores best, being instrumental to suppress disturbed shear in the venous segment. There is, however, no trivial relationship between the number of helix turns of the graft and disturbed shear in the venous segment, when a realistic closed-loop AVG model is investigated. Bulk flow investigation showed a marked increase of helicity intensity in, and a moderate pressure drop over the AVG by introducing a lower pitch. At the venous anastomosis, unfavorable hemodynamic conditions can be reduced by introducing a helical design. However, due to the complex flow conditions, the optimal helical design for an AVG cannot be derived without studying case by case.

Virtual evaluation of stent graft deployment: a validated modeling and simulation study.

The presented study details the virtual deployment of a bifurcated stent graft (Medtronic Talent) in an Abdominal Aortic Aneurysm model, using the finite element method. The entire deployment procedure is modeled, with the stent graft being crimped and bent according to the vessel geometry, and subsequently released. The finite element results are validated in vitro with placement of the device in a silicone mock aneurysm, using high resolution CT scans to evaluate the result. The presented work confirms the capability of finite element computer simulations to predict the deformed configuration after endovascular aneurysm repair (EVAR). These simulations can be used to quantify mechanical parameters, such as neck dilations, radial forces and stresses in the device, that are difficult or impossible to obtain from medical imaging.

Our capricious vessels: The influence of stent design and vessel geometry on the mechanics of intracranial aneurysm stent deployment.

There is a growing interest in virtual tools to assist clinicians in evaluating different procedures and devices for endovascular treatment. In the present study we use finite element analysis to investigate the influence of stent design and vessel geometry for stent assisted coiling of intracranial aneurysms. Nine virtual stenting procedures were performed: three nitinol stent designs ((i) an open cell stent resembling the Neuroform, (ii) a generic stiff and (iii) a more flexible closed cell design), were deployed in three patient-specific cerebral aneurysmatic vessels. We investigated the percentage of strut area covering the aneurysm neck, the straightening induced on the cerebrovasculature by the stent placement (quantified by the reduction in tortuosity), and stent apposition to the wall (quantified as the percentage of struts within 0.2mm of the vessel). The results suggest that the open cell design better covers the aneurysm neck (11.0±1.1%) compared to both the stiff (7.8±1.6%) and flexible (8.7±1.6%) closed cell stents, and induces less straightening of the vessel (-5.1±1.6% vs. -42.9±9.8% and -26.9±11.9% ). The open cell design has, however, less struts apposing well to the vessel wall (56.0±6.4%) compared to the flexible (73.4±4.6%) and stiff (70.4±5.1%) closed cell design. With the presented study, we hope to contribute to and improve aneurysm treatment, using a novel patient specific environment as a possible pre-operative tool to evaluate mechanical stent behavior in different vascular geometries.

Experimental validation of a pulse wave propagation model for predicting hemodynamics after vascular access surgery.

Hemodialysis patients require a vascular access that is, preferably, surgically created by connecting an artery and vein in the arm, i.e. an arteriovenous fistula (AVF). The site for AVF creation is chosen by the surgeon based on preoperative diagnostics, but AVFs are still compromised by flow-associated complications. Previously, it was shown that a computational 1D-model is able to describe pressure and flow after AVF surgery. However, predicted flows differed from measurements in 4/10 patients. Differences can be attributed to inaccuracies in Doppler measurements and input data, to neglecting physiological mechanisms or to an incomplete physical description of the pulse wave propagation after AVF surgery. The physical description can be checked by validating against an experimental setup consisting of silicone tubes mimicking the aorta and arm vasculature both before and after AVF surgery, which is the aim of the current study. In such an analysis, the output uncertainty resulting from measurement uncertainty in model input should be quantified. The computational model was fed by geometrical and mechanical properties collected from the setup. Pressure and flow waveforms were simulated and compared with experimental waveforms. The precision of the simulations was determined by performing a Monte Carlo study. It was concluded that the computational model was able to simulate mean pressures and flows accurately, whereas simulated waveforms were less attenuated than experimental ones, likely resulting from neglecting viscoelasticity. Furthermore, it was found that in the analysis output uncertainties, resulting from input uncertainties, cannot be neglected and should thus be considered.