Creation and application of virtual patient cohorts of heart models.

Patient-specific cardiac models are now being used to guide therapies. The increased use of patient-specific cardiac simulations in clinical care will give rise to the development of virtual cohorts of cardiac models. These cohorts will allow cardiac simulations to capture and quantify inter-patient variability. However, the development of virtual cohorts of cardiac models will require the transformation of cardiac modelling from small numbers of bespoke models to robust and rapid workflows that can create large numbers of models. In this review, we describe the state of the art in virtual cohorts of cardiac models, the process of creating virtual cohorts of cardiac models, and how to generate the individual cohort member models, followed by a discussion of the potential and future applications of virtual cohorts of cardiac models. This article is part of the theme issue 'Uncertainty quantification in cardiac and cardiovascular modelling and simulation'.

A Metamodeling Approach for Instant Severity Assessment and Uncertainty Quantification of Iliac Artery Stenoses.

Two-dimensional (2D) or three-dimensional (3D) models of blood flow in stenosed arteries can be used to patient-specifically predict outcome metrics, thereby supporting the physicians in decision making processes. However, these models are time consuming which limits the feasibility of output uncertainty quantification (UQ). Accurate surrogates (metamodels) might be the solution. In this study, we aim to demonstrate the feasibility of a generalized polynomial chaos expansion-based metamodel to predict a clinically relevant output metric and to quantify the output uncertainty. As an example, a metamodel was constructed from a recently developed 2D model that was shown to be able to estimate translesional pressure drops in iliac artery stenoses (-0.9 ± 12.7 mmHg, R2 = 0.81). The metamodel was constructed from a virtual database using the adaptive generalized polynomial chaos expansion (agPCE) method. The constructed metamodel was then applied to 25 stenosed iliac arteries to predict the patient-specific pressure drop and to perform UQ. Comparing predicted pressure drops of the metamodel and in vivo measured pressure drops, the mean bias (-0.2 ± 13.7 mmHg) and the coefficient of determination (R2 = 0.80) were as good as of the original 2D computational fluid dynamics (CFD) model. UQ results of the 2D and metamodel were comparable. Estimation of the uncertainty interval using the original 2D model took 14 days, whereas the result of the metamodel was instantly available. In conclusion, it is feasible to quantify the uncertainty of the output metric and perform sensitivity analysis (SA) instantly using a metamodel. Future studies should investigate the possibility to construct a metamodel of more complex problems.

The brisk-standing-test for long QT syndrome in prepubertal school children: defining normal.

Aims: Long QT syndrome (LQTS) is associated with malignant arrhythmias and sudden death from birth to advanced age. Prolongation of the QT-interval, may however be concealed on standard electrocardiograms (ECG). The brisk-standing-test (BST) was developed to guide LQTS-diagnosis and treatment in adults. We hypothesized that the BST may be used in prepubertal children to identify LQTS subjects. Accordingly, reference values for the BST should be available to prevent incorrect diagnosis and treatment of LQTS. In this study, we aim to present reference values for prepubertal children. Methods and results: Healthy, prepubertal children, aged 7-13 years underwent a standard supine resting ECG and during continuous ECG recording performed a BST. The QT-interval and heart rate corrected QTc were measured during the different BST stages. Fifty-seven children, 29 boys (10.2 ± 1.1 years) and 28 girls (9.9 ± 1.1 years) were included. Baseline characteristics and response to standing were not statistically different for boys and girls: mean supine pre-standing heart rate 74 ± 9 vs. 77 ± 9 bpm, supine pre-standing QTc 406 ± 27 vs. 407 ± 17 ms, maximal heart rate upon standing 109 ± 11 vs. 112 ± 11 bpm, and QTc at maximal heart rate 484 ± 29 vs. 487 ± 35 ms. The QT interval corrected for heart rate-prolongation at maximal tachycardia after standing was 79 ± 26 (19-144) ms, which is significantly longer than previously published values in adults (50± 30 ms). Conclusions: The QT interval corrected for heart rate prolongation after brisk standing in healthy prepubertal children is more pronounced than in healthy adults. This finding advocates distinct prepubertal cut-off values because using adult values for prepubertal children would yield false positive results with the risk of incorrect LQTS-diagnosis and overtreatment.

New insights from a computational model on the relation between pacing site and CRT response.

AIMS: Cardiac resynchronization therapy (CRT) produces clinical benefits in chronic heart failure patients with left bundle-branch block (LBBB). The position of the pacing site on the left ventricle (LV) is considered an important determinant of CRT response, but the mechanism how the LV pacing site determines CRT response is not completely understood. The objective of this study is to investigate the relation between LV pacing site during biventricular (BiV) pacing and cardiac function. METHODS AND RESULTS: We used a finite element model of BiV electromechanics. Cardiac function, assessed as LV dp/dtmax and stroke work, was evaluated during normal electrical activation, typical LBBB, fascicular blocks and BiV pacing with different LV pacing sites. The model replicated clinical observations such as increase of LV dp/dtmax and stroke work, and the disappearance of a septal flash during BiV pacing. The largest hemodynamic response was achieved when BiV pacing led to best resynchronization of LV electrical activation but this did not coincide with reduction in total BiV activation time (∼ QRS duration). Maximum response was achieved when pacing the mid-basal lateral wall and this was close to the latest activated region during intrinsic activation in the typical LBBB, but not in the fascicular block simulations. CONCLUSIONS: In these model simulations, the best cardiac function was obtained when pacing the mid-basal LV lateral wall, because of fastest recruitment of LV activation. This study illustrates how computer modeling can shed new light on optimizing pacing therapies for CRT. The results from this study may help to design new clinical studies to further investigate the importance of the pacing site for CRT response.

Variability of microcirculatory measurements in healthy volunteers.

Reliable assessment of the microcirculation is important to investigate microcirculatory properties in various disease states. The GlycoCheck system automatically analyzes sublingual sidestream dark field images to determine the perfused boundary region (PBR; a measure of glycocalyx thickness), red blood cell filling percentage, and microvascular vessel density. Although GlycoCheck has been used to study the microcirculation in patients, little is known about the reproducibility of measurements in healthy volunteers. We assessed intra- and interobserver agreement by having two experienced observers perform three consecutive microcirculation measurements with the GlycoCheck system in 49 healthy volunteers. Intraobserver agreement of single measurements were poor (intraclass correlation coefficients (ICCs) < 0.4) for PBR, red blood cell filling percentage and microvascular vessel density. ICCs increased to values > 0.6 (indicating good reproducibility) for all parameters when performing and averaging three consecutive measurements. No systematic differences were observed between observers for any parameter. Interobserver variability was fair for PBR (ICC = 0.53) and red blood cell filling percentage (ICC = 0.58) and poor for perfused vessel density (ICC = 0.20). In conclusion, GlycoCheck software can be used with acceptable reliability and reproducibility for microcirculation measurements on a population level when averaging three consecutive measurements. Repeated measurements are preferably performed by the same observer.

Multimodal measurement of glycocalyx degradation during coronary artery bypass grafting.

Background: Glycocalyx shedding and subsequent endothelial dysfunction occur in many conditions, such as in sepsis, in critical illness, and during major surgery such as in coronary artery bypass grafting (CABG) where it has been shown to associate with organ dysfunction. Hitherto, there is no consensus about the golden standard in measuring glycocalyx properties in humans. The objective of this study was to compare different indices of glycocalyx shedding and dysfunction. To this end, we studied patients undergoing elective CABG surgery, which is a known cause of glycocalyx shedding. Materials and methods: Sublingual glycocalyx thickness was measured in 23 patients by: 1) determining the perfused boundary region (PBR)-an inverse measure of glycocalyx thickness-by means of sidestream dark field imaging technique. This is stated double, 2) measuring plasma levels of the glycocalyx shedding products syndecan-1, hyaluronan, and heparan sulfate and 3) measuring plasma markers of impaired glycocalyx function and endothelial activation (Ang-2, Tie-2, E-selectin, and thrombomodulin). Measurements were performed directly after induction, directly after onset of cardiopulmonary bypass (CPB), and directly after cessation of CPB. We assessed changes over time as well as correlations between the various markers. Results: The PBR increased from 1.81 ± 0.21 µm after induction of anesthesia to 2.27 ± 0.25 µm (p < 0.0001) directly after CPB was initiated and did not change further during CPB. A similar pattern was seen for syndecan-1, hyaluronan, heparan sulfate, Ang-2, Tie-2, and thrombomodulin. E-selectin levels also increased between induction and the start of CPB and increased further during CPB. The PBR correlated moderately with heparan sulfate, E-selectin, and thrombomodulin and weakly with Syndecan-1, hyaluronan, and Tie-2. Shedding markers syndecan-1 and hyaluronan correlated with all functional markers. Shedding marker heparan sulfate only correlated with Tie-2, thrombomodulin, and E-selectin. Thrombomodulin correlated with all shedding markers. Conclusion: Our results show that glycocalyx thinning, illustrated by increased sublingual PBR and increased levels of shedding markers, is paralleled with impaired glycocalyx function and increased endothelial activation in CABG surgery with CPB. As correlations between different markers were limited, no single marker could be identified to represent the glycocalyx in its full complexity.

Cardiac resynchronization: insight from experimental and computational models.

Cardiac resynchronization therapy (CRT) is a promising therapy for heart failure patients with a conduction disturbance, such as left bundle branch block. The aim of CRT is to resynchronize contraction between and within ventricles. However, about 30% of patients do not respond to this therapy. Therefore, a better understanding is needed for the relation between electrical and mechanical activation. In this paper, we focus on to what extent animal experiments and mathematical models can help in order to understand the pathophysiology of asynchrony to further improve CRT.

A geometry-based model for non-invasive estimation of pressure gradients over iliac artery stenoses.

The aim of this study was to develop and verify a model that provides an accurate estimation of the trans-lesion hyperemic pressure gradient in iliac artery stenoses in seconds by only using patient-specific geometric properties obtained from 3-dimensional rotational angiography (3DRA). Twenty-one patients with symptomatic peripheral arterial disease (PAD), iliac artery stenoses and an ultrasound based peak systolic velocity ratio between 2.5 and 5.0 underwent 3DRA and intra-arterial pressure measurements under hyperemic conditions. For each lesion, geometric properties were extracted from the 3DRA images using quantitative vascular analysis software. Hyperemic blood flow was estimated based on stenosis geometry using an empirical relation. The geometrical properties and hyperemic flow were used to estimate the pressure gradient by means of the geometry-based model. The predicted pressure gradients were compared with in vivo measured intra-arterial pressure measurements performed under hyperemic conditions. The developed geometry-based model showed good agreement with the measured hyperemic pressure gradients resulting in a concordance correlation coefficient of 0.86. The mean bias ±â€¯2SD between the geometry-based model and in vivo measurements was comparable to results found by evaluating the actual computational fluid dynamics model (-1.0 ±â€¯14.7 mmHg vs -0.9 ±â€¯12.7 mmHg). The developed model estimates the trans-lesional pressure gradient in seconds without the need for an additional computational fluid dynamics software package. The results justify further study to assess the potential use of a geometry-based model approach to estimate pressure gradient on non-invasive CTA or MRA, thereby reducing the need for diagnostic angiography in patients suffering from PAD.

Towards model-based analysis of cardiac MR tagging data: relation between left ventricular shear strain and myofiber orientation.

Many cardiac pathologies are reflected in abnormal myocardial deformation, accessible through magnetic resonance tagging (MRT). Interpretation of the MRT data is difficult, since the relation between pathology and deformation is not straightforward. Mathematical models of cardiac mechanics could be used to translate measured abnormalities into the underlying pathology, but, so far, they even fail to correctly simulate myocardial deformation in the healthy heart. In this study we investigated to what extent (1) our previously published three-dimensional finite element model of cardiac mechanics [Kerckhoffs, R.C.P., Bovendeerd, P.H.M., Kotte, J.C.S., Prinzen, F.W., Smits, K., Arts, T., 2003. Homogeneity of cardiac contraction despite physiological asynchrony of depolarization: a model study. Ann. Biomed. Eng. 31, 536-547] can simulate measured cardiac deformation, and (2) discrepancies between strains in model and experiment are related to the choice of the myofiber orientation in the model. To this end, we measured midwall circumferential strain E(cc) and circumferential-radial shear strain E(cr) in three healthy subjects using MRT. E(cc) as computed in the model agreed well with measured E(cc). Computed E(cr) differed significantly from measured E(cr). The time course of E(cr) was found to be very sensitive to the choice of the myofiber orientation, in particular to the choice of the transverse angle. Discrepancies between circumferential-radial shear strain in model and experiment were reduced strongly by increasing the transverse angle in the original model by 25%.

Estimates of regional work in the canine left ventricle.

Assessment of the magnitude of regional myocardial work requires knowledge of regional fiber stress and fiber shortening. The theoretical development and experimental validation of a method is presented which used values of estimated active and passive fiber stress according to a fluid-fiber model, and measured fiber strain values. This enables the construction of regional stress-strain diagrams, a regional analog of the pressure-volume area model by Suga and co-investigators, which can be linked to regional oxygen consumption. In the left ventricle, either normally or asynchronously activated, the method yields reliable data on strain and active and passive fiber stress. The relation between estimated regional work and myocardial oxygen demand is in quantitative agreement with previously reported relations between global oxygen demand and measured pressure-volume area. During coronary artery occlusion, however, these values were less reliable, which might be due to inaqdequate knowledge of the (passive) material properties of the myocardium.

Personalization of models with many model parameters: an efficient sensitivity analysis approach.

Uncertainty quantification and global sensitivity analysis are indispensable for patient-specific applications of models that enhance diagnosis or aid decision-making. Variance-based sensitivity analysis methods, which apportion each fraction of the output uncertainty (variance) to the effects of individual input parameters or their interactions, are considered the gold standard. The variance portions are called the Sobol sensitivity indices and can be estimated by a Monte Carlo (MC) approach (e.g., Saltelli's method [1]) or by employing a metamodel (e.g., the (generalized) polynomial chaos expansion (gPCE) [2, 3]). All these methods require a large number of model evaluations when estimating the Sobol sensitivity indices for models with many parameters [4]. To reduce the computational cost, we introduce a two-step approach. In the first step, a subset of important parameters is identified for each output of interest using the screening method of Morris [5]. In the second step, a quantitative variance-based sensitivity analysis is performed using gPCE. Efficient sampling strategies are introduced to minimize the number of model runs required to obtain the sensitivity indices for models considering multiple outputs. The approach is tested using a model that was developed for predicting post-operative flows after creation of a vascular access for renal failure patients. We compare the sensitivity indices obtained with the novel two-step approach with those obtained from a reference analysis that applies Saltelli's MC method. The two-step approach was found to yield accurate estimates of the sensitivity indices at two orders of magnitude lower computational cost.

Respiration and measurement of cardiac output by thermodilution and central or peripheral dye dilution.

Cardiac output as measured by indicator dilution methods during artificial ventilation shows differences up to +/- 35%. We studied the influence of spontaneous breathing on measurement of cardiac output by thermodilution (TD) and central (CDD) and peripheral dye dilution (PDD) in seven anesthetized dogs. Injection of indicator was timed at one of five chosen moments in a respiratory cycle. The indicator for TD was also used as solvent for indocyanine green. Results were normalized by the value obtained with injection at inspiratory onset. Results of the central dilution methods showed a slight but not significant difference between values measured with injection at 25 and 75% of the respiratory cycle: 105.7 and 98.0%, respectively, (TD) and 102.3 and 97.2% (CDD). Mean cardiac output determined by TD, CDD, or PDD was not significantly different. We conclude that 1) a reasonable estimate of cardiac output may be obtained by means of a single indicator-dilution curve and 2) the choice of the dilution method may be determined by practical considerations.

Modeling the relation between cardiac pump function and myofiber mechanics.

Complexity of the geometry and structure of the heart hampers easy modeling of cardiac mechanics. The modeling can however be simplified considerably when using the hypothesis that in the normal heart myofiber structure and geometry adapt, until load is evenly distributed. A simple and realistic relationship is found between the hemodynamic variables cavity pressure and volume, and myofiber load parameters stress and strain. The most important geometric parameter in the latter relation is the ratio of cavity volume to wall volume, while actual geometry appears practically irrelevant. Applying the found relationship, a realistic maximum is set to left ventricular pressure after chronic pressure load. Pressures exceeding this level are likely to cause decompensation and heart failure. Furthermore, model is presented to simulate left and right ventricular pump function with left-right interaction.

Asymmetrical changes in ventricular wall mass by asynchronous electrical activation of the heart.

Ventricular pacing causes asynchronous electrical activation of the ventricular wall, because impulse conduction occurs via muscle fibers rather than via the Purkinje system. Chronic (up to 3 months) ventricular pacing caused about 30% decrease of wall mass in early activated regions but did not change wall mass in late activated regions. These are the first data indicating that chronic asynchronous activation induces asymmetrical structural adaptations. This asymmetry is likely to be evoked by regional differences in contractile work, as demonstrated in previous experiments from our laboratory. The nature of the structural adaptation as well as its clinical implications deserve more detailed investigation.

[Hypertrophic cardiomyopathy: from diagnosis to prevention]. / Hypertrofische cardiomyopathie: van diagnostiek naar preventie.

In a 55-year-old man, his two sons aged 32 and 30 years and a 3-year-old grandchild, a gene mutation related to an increased chance of hypertrophic cardiomyopathy was found. The adults had complaints of cardiac arrhythmias varying from dizziness to cardiac arrest and received medication; a defibrillator was also implanted in the case of the sons. Familial hypertrophic cardiomyopathy can be diagnosed in an early stage with DNA screening methods. At that time 'patients' (carriers) often do not exhibit symptoms. This offers the possibility of taking therapeutic measures to prevent or slow down the disease process. However, the scientific basis for preventive therapy is inadequate, so that medical-ethical dilemmas arise which can affect the decision-making process with respect to genetic testing. Additional research must lead to satisfactory preventive therapies, so that in the future the genetic diagnosis 'hypertrophic cardiomyopathy' can be effectively translated into disease prevention.

Applicability of the polynomial chaos expansion method for personalization of a cardiovascular pulse wave propagation model.

Patient-specific modeling requires model personalization, which can be achieved in an efficient manner by parameter fixing and parameter prioritization. An efficient variance-based method is using generalized polynomial chaos expansion (gPCE), but it has not been applied in the context of model personalization, nor has it ever been compared with standard variance-based methods for models with many parameters. In this work, we apply the gPCE method to a previously reported pulse wave propagation model and compare the conclusions for model personalization with that of a reference analysis performed with Saltelli's efficient Monte Carlo method. We furthermore differentiate two approaches for obtaining the expansion coefficients: one based on spectral projection (gPCE-P) and one based on least squares regression (gPCE-R). It was found that in general the gPCE yields similar conclusions as the reference analysis but at much lower cost, as long as the polynomial metamodel does not contain unnecessary high order terms. Furthermore, the gPCE-R approach generally yielded better results than gPCE-P. The weak performance of the gPCE-P can be attributed to the assessment of the expansion coefficients using the Smolyak algorithm, which might be hampered by the high number of model parameters and/or by possible non-smoothness in the output space.

Mapping displacement and deformation of the heart with local sine-wave modeling.

The new SinMod method extracts motion from magnetic resonance imaging (MRI)-tagged (MRIT) image sequences. Image intensity in the environment of each pixel is modeled as a moving sine wavefront. Displacement is estimated at subpixel accuracy. Performance is compared with the harmonic-phase analysis (HARP) method, which is currently the most common method used to detect motion in MRIT images. SinMod can handle line tags, as well as speckle patterns. In artificial images (tag distance six pixels), SinMod detects displacements accurately (error < 0.02 pixels). Effects of noise are suppressed effectively. Sharp transitions in motion at the boundary of an object are smeared out over a width of 0.6 tag distance. For MRIT images of the heart, SinMod appears less sensitive to artifacts, especially later in the cardiac cycle when image quality deteriorates. For each pixel, the quality of the sine-wave model in describing local image intensity is quantified objectively. If local quality is low, artifacts are avoided by averaging motion over a larger environment. Summarizing, SinMod is just as fast as HARP, but it performs better with respect to accuracy of displacement detection, noise reduction, and avoidance of artifacts.