Patient-specific requirements and clinical validation of MRI-based pressure mapping: A two-center study in patients with aortic coarctation

Deep learning based assessment of hemodynamics in the coarctation of the aorta: comparison of bidirectional recurrent and convolutional neural networks

The utilization of numerical methods, such as computational fluid dynamics (CFD), has been widely established for modeling patient-specific hemodynamics based on medical imaging data. Hemodynamics assessment plays a crucial role in treatment decisions for the coarctation of the aorta (CoA), a congenital heart disease, with the pressure drop (PD) being a crucial biomarker for CoA treatment decisions. However, implementing CFD methods in the clinical environment remains challenging due to their computational cost and the requirement for expert knowledge. This study proposes a deep learning approach to mitigate the computational need and produce fast results. Building upon a previous proof-of-concept study, we compared the effects of two different artificial neural network (ANN) architectures trained on data with different dimensionalities, both capable of predicting hemodynamic parameters in CoA patients: a one-dimensional bidirectional recurrent neural network (1D BRNN) and a three-dimensional convolutional neural network (3D CNN). The performance was evaluated by median point-wise root mean square error (RMSE) for pressures along the centerline in 18 test cases, which were not included in a training cohort. We found that the 3D CNN (median RMSE of 3.23 mmHg) outperforms the 1D BRNN (median RMSE of 4.25 mmHg). In contrast, the 1D BRNN is more precise in PD prediction, with a lower standard deviation of the error (±7.03 mmHg) compared to the 3D CNN (±8.91 mmHg). The differences between both ANNs are not statistically significant, suggesting that compressing the 3D aorta hemodynamics into a 1D centerline representation does not result in the loss of valuable information when training ANN models. Additionally, we evaluated the utility of the synthetic geometries of the aortas with CoA generated by using a statistical shape model (SSM), as well as the impact of aortic arch geometry (gothic arch shape) on the model's training. The results show that incorporating a synthetic cohort obtained through the SSM of the clinical cohort does not significantly increase the model's accuracy, indicating that the synthetic cohort generation might be oversimplified. Furthermore, our study reveals that selecting training cases based on aortic arch shape (gothic versus non-gothic) does not improve ANN performance for test cases sharing the same shape.

Intra-cardiac pressure drop and flow distribution of bicuspid aortic valve disease in preserved ejection fraction

Background

Bicuspid aortic valve (BAV) is more than a congenital defect since it is accompanied by several secondary complications that intensify induced impairments. Hence, BAV patients need lifelong evaluations to prevent severe clinical sequelae. We applied 4D-flow magnetic resonance imaging (MRI) for in detail visualization and quantification of in vivo blood flow to verify the reliability of the left ventricular (LV) flow components and pressure drops in the silent BAV subjects with mild regurgitation and preserved ejection fraction (pEF).

Materials and methods

A total of 51 BAV patients with mild regurgitation and 24 healthy controls were recruited to undergo routine cardiac MRI followed by 4D-flow MRI using 3T MRI scanners. A dedicated 4D-flow module was utilized to pre-process and then analyze the LV flow components (direct flow, retained inflow, delayed ejection, and residual volume) and left-sided [left atrium (LA) and LV] local pressure drop. To elucidate significant diastolic dysfunction in our population, transmitral early and late diastolic 4D flow peak velocity (E-wave and A-wave, respectively), as well as E/A ratio variable, were acquired.

Results

The significant means differences of each LV flow component (global measurement) were not observed between the two groups (p > 0.05). In terms of pressure analysis (local measurement), maximum and mean as well as pressure at E-wave and A-wave timepoints at the mitral valve (MV) plane were significantly different between BAV and control groups (p: 0.005, p: 0.02, and p: 0.04 and p: <0.001; respectively). Furthermore, maximum pressure and pressure difference at the A-wave timepoint at left ventricle mid and left ventricle apex planes were significant. Although we could not find any correlation between LV diastolic function and flow components, Low but statistically significant correlations were observed with local pressure at LA mid, MV and LV apex planes at E-wave timepoint (R: −0.324, p: 0.005, R: −0.327, p: 0.004, and R: −0.306, p: 0.008, respectively).

Conclusion

In BAV patients with pEF, flow components analysis is not sensitive to differentiate BAV patients with mild regurgitation and healthy control because flow components and EF are global parameters. Inversely, pressure (local measurement) can be a more reliable biomarker to reveal the early stage of diastolic dysfunction.

The effect of turbulence modelling on the assessment of platelet activation

Pathological platelet activation by abnormal shear stresses is regarded as a main clinical complication in recipients of cardiovascular mechanical devices. In order to improve their performance computational fluid dynamics (CFD) are used to evaluate flow fields and related shear stresses. CFD models are coupled with mathematical models that describe the relation between fluid dynamics variables, and in particular shear stresses, and the platelet activation state (PAS). These models typically use a Lagrangian approach to compute the shear stresses along possible platelet trajectories. However, in the case of turbulent flow, the choice of the proper turbulence closure is still debated for both concerning its effect on shear stress calculation and Lagrangian statistics. In this study different numerical simulations of the flow through a mechanical heart valve were performed and then compared in terms of Eulerian and Lagrangian quantities: a direct numerical simulation (DNS), a large eddy simulation (LES), two Reynolds-averaged Navier-Stokes (RANS) simulations (SST k-ω and RSM) and a "laminar" (no turbulence modelling) simulation. Results exhibit a large variability in the PAS assessment depending on the turbulence model adopted. "Laminar" and RSM estimates of platelet activation are about 60% below DNS, while LES is 16% less. Surprisingly, PAS estimated from the SST k- ω velocity field is only 8% less than from DNS data. This appears more artificial than physical as can be inferred after comparing frequency distributions of PAS and of the different Lagrangian variables of the mechano-biological model of platelet activation. Our study indicates how much turbulence closures may affect platelet activation estimates, in comparison to an accurate DNS, when assessing blood damage in blood contacting devices.

CTA-Based Non-invasive Estimation of Pressure Gradients Across a CoA: a Validation Against Cardiac Catheterisation

Non-invasive estimation of pressure gradients across a coarctation of the aorta (CoA) can reduce the need for diagnostic cardiac catheterisation. We aimed to validate two novel computational strategies-target-value approaching (TVA) and target-value fixing (TVF)-together with unrefined Doppler estimates, and to compare their diagnostic performance in identifying critical pressure drops for 40 patients. Compared to catheterisation, no statistically significant difference was demonstrated with TVA (P = 0.086), in contrast to TVF (P = 0.005) and unrefined Doppler echocardiography (P < 0.001). TVA manifested the strongest correlation with catheterisation (r = 0.93), compared to TVF (r = 0.83) and echocardiography (r = 0.67) (all P < 0.001). In discriminating pressure gradients greater than 20 mmHg, TVA, TVF, and echocardiography had respective sensitivities of 0.92, 0.88, and 0.80; specificities of 0.93, 0.80, and 0.73; and AUCs of 0.96, 0.89, and 0.80. The TVA strategy may serve as an effective and easily implemented approach to be used in clinical management of patients with CoA. Graphical Abstract Central illustration. Pressure gradients estimated using Doppler echocardiography and two novel computational strategies (TVA and TVF) were compared with cardiac catheterisation for 40 patients. TVA and TVF utilised the CTA images to obtain the CoA anatomy and Doppler echocardiography velocimetry to obtain velocity data for the assignment of CFD boundary conditions.

Validation of 4D Flow based relative pressure maps in aortic flows

While the clinical gold standard for pressure difference measurements is invasive catheterization, 4D Flow MRI is a promising tool for enabling a non-invasive quantification, by linking highly spatially resolved velocity measurements with pressure differences via the incompressible Navier-Stokes equations. In this work we provide a validation and comparison with phantom and clinical patient data of pressure difference maps estimators. We compare the classical Pressure Poisson Estimator (PPE) and the new Stokes Estimator (STE) against catheter pressure measurements under a variety of stenosis severities and flow intensities. Specifically, we use several 4D Flow data sets of realistic aortic phantoms with different anatomic and hemodynamic severities and two patients with aortic coarctation. The phantom data sets are enriched by subsampling to lower resolutions, modification of the segmentation and addition of synthetic noise, in order to study the sensitivity of the pressure difference estimators to these factors. Overall, the STE method yields more accurate results than the PPE method compared to catheterization data. The superiority of the STE becomes more evident at increasing Reynolds numbers with a better capacity of capturing pressure gradients in strongly convective flow regimes. The results indicate an improved robustness of the STE method with respect to variation in lumen segmentation. However, with heuristic removal of the wall-voxels, the PPE can reach a comparable accuracy for lower Reynolds' numbers.

Image-Based Computational Model Predicts Dobutamine-Induced Hemodynamic Changes in Patients With Aortic Coarctation

BACKGROUND

Pharmacological stress testing can help to uncover pathological hemodynamic conditions and is, therefore, used in the clinical routine to assess patients with structural heart diseases such as aortic coarctation with borderline indication for treatment. The aim of this study was to develop and test a reduced-order model predicting dobutamine stress induced pressure gradients across the coarctation.

METHODS

The reduced-order model was developed based on n=21 imaging data sets of patients with aortic coarctation and a meta-analysis of subjects undergoing dobutamine stress testing. Within an independent test cohort of n=21 patients with aortic coarctation, the results of the model were compared with dobutamine stress testing during catheterization.

RESULTS

In n=19 patients responding to dobutamine stress testing, pressure gradients across the coarctation during dobutamine stress increased from 15.7±5.1 to 33.6±10.3 mm Hg (paired t test, P<0.001). The model-predicted pressure gradients agreed with catheter measurements with a mean difference of -2.2 mm Hg and a limit of agreement of ±11.16 mm Hg according to Bland-Altman analysis. Significant equivalence between catheter-measured and simulated pressure gradients during stress was found within the study cohort (two 1-sided tests of equivalence with a noninferiority margin of 5.0 mm Hg, 33.6±10.33 versus 31.5±11.15 mm Hg, P=0.021).

CONCLUSIONS

The developed reduced-order model can instantly predict dobutamine-induced hemodynamic changes with accuracy equivalent to heart catheterization in patients with aortic coarctation. The method is easy to use, available as a web-based calculator, and provides a promising alternative to conventional stress testing in the clinical routine. Registration: URL: https://www.clinicaltrials.gov. Unique identifier: NCT02591940.

Assessment of hemodynamic responses to exercise in aortic coarctation using MRI-ergometry in combination with computational fluid dynamics

In patients with aortic coarctation it would be desirable to assess pressure gradients as well as information about blood flow profiles at rest and during exercise. We aimed to assess the hemodynamic responses to physical exercise by combining MRI-ergometry with computational fluid dynamics (CFD). MRI was performed on 20 patients with aortic coarctation (13 men, 7 women, mean age 21.5 ± 13.7 years) at rest and during ergometry. Peak systolic pressure gradients, wall shear stress (WSS), secondary flow degree (SFD) and normalized flow displacement (NFD) were calculated using CFD. Stroke volume was determined based on MRI. On average, the pressure gradient was 18.0 ± 16.6 mmHg at rest and increased to 28.5 ± 22.6 mmHg (p < 0.001) during exercise. A significant increase in cardiac index was observed (p < 0.001), which was mainly driven by an increase in heart rate (p < 0.001). WSS significantly increased during exercise (p = 0.006), whereas SFD and NFD remained unchanged. The combination of MRI-ergometry with CFD allows assessing pressure gradients as well as flow profiles during physical exercise. This concept has the potential to serve as an alternative to cardiac catheterization with pharmacological stress testing and provides hemodynamic information valuable for studying the pathophysiology of aortic coarctation.

4D flow MRI applications in congenital heart disease

Advances in the diagnosis and management of congenital heart disease (CHD) have resulted in a growing population of patients surviving well into adulthood and requiring lifelong follow-up. Flow quantification is a central component in the assessment of patients with CHD. 4D flow magnetic resonance imaging (MRI) has emerged as a tool that enables comprehensive study of flow. It involves the acquisition of a three-dimensional time-resolved volume with velocity encoding in all three spatial directions along the cardiac cycle. This allows flow quantification and visualization of blood flow patterns as well as the study of advanced hemodynamic parameters as kinetic energy and wall shear stress. 4D flow MRI-based study of flow has given insight into the altered hemodynamics in CHD particularly in bicuspid aortic valve disease and Fontan circulation. The aim of this review is to discuss the expanding clinical and research applications of 4D flow MRI in CHD as well its limitations.Key Points• Three-dimensional velocity encoding allows not only flow quantification but also the visualization of multidirectional flow patterns and the study of advanced hemodynamic parameters.• 4D flow MRI has added insight into the abnormal hemodynamics involved in congenital heart disease in particular in bicuspid aortic valve and Fontan circulation.• The main limitation of 4D flow MRI in congenital heart disease is the relatively long scan duration required for the complete coverage of the heart and great vessels with adequate spatiotemporal resolution.

A Patient-Specific CFD Pipeline Using Doppler Echocardiography for Application in Coarctation of the Aorta in a Limited Resource Clinical Context

Congenital heart disease (CHD) is the most common birth defect globally and coarctation of the aorta (CoA) is one of the commoner CHD conditions, affecting around 1/1800 live births. CoA is considered a CHD of critical severity. Unfortunately, the prognosis for a child born in a low and lower-middle income country (LLMICs) with CoA is far worse than in a high-income country. Reduced diagnostic and interventional capacities of specialists in these regions lead to delayed diagnosis and treatment, which in turn lead to more cases presenting at an advanced stage. Computational fluid dynamics (CFD) is an important tool in this context since it can provide additional diagnostic data in the form of hemodynamic parameters. It also provides an in silico framework, both to test potential procedures and to assess the risk of further complications arising post-repair. Although this concept is already in practice in high income countries, the clinical infrastructure in LLMICs can be sparse, and access to advanced imaging modalities such as phase contrast magnetic resonance imaging (PC-MRI) is limited, if not impossible. In this study, a pipeline was developed in conjunction with clinicians at the Red Cross War Memorial Children’s Hospital, Cape Town and was applied to perform a patient-specific CFD study of CoA. The pipeline uses data acquired from CT angiography and Doppler transthoracic echocardiography (both much more clinically available than MRI in LLMICs), while segmentation is conducted via SimVascular and simulation is realized using OpenFOAM. The reduction in cost through use of open-source software and the use of broadly available imaging modalities makes the methodology clinically feasible and repeatable within resource-constrained environments. The project identifies the key role of Doppler echocardiography, despite its disadvantages, as an intrinsic component of the pipeline if it is to be used routinely in LLMICs.

Impact of predictive medicine on therapeutic decision making: a randomized controlled trial in congenital heart disease

Computational modelling has made significant progress towards clinical application in recent years. In addition to providing detailed diagnostic data, these methods have the potential to simulate patient-specific interventions and to predict their outcome. Our objective was to evaluate to which extent patient-specific modelling influences treatment decisions in coarctation of the aorta (CoA), a common congenital heart disease. We selected three cases with CoA, two of which had borderline indications for intervention according to current clinical guidelines. The third case was not indicated for intervention according to guidelines. For each case, we generated two separate datasets. First dataset included conventional diagnostic parameters (echocardiography and magnetic resonance imaging). In the second, we added modelled parameters (pressure fields). For the two cases with borderline indications for intervention, the second dataset also included pressure fields after virtual stenting simulations. All parameters were computed by modelling methods that were previously validated. In an online-administered, invitation-only survey, we randomized 178 paediatric cardiologists to view either conventional (control) or add-on modelling (experimental) datasets. Primary endpoint was the proportion of participants recommending different therapeutic options: (1) surgery or catheter lab (collectively, "intervention") or (2) no intervention (follow-up with or without medication). Availability of data from computational predictive modelling influenced therapeutic decision making in two of three cases. There was a statistically significant association between group assignment and the recommendation of an intervention for one borderline case and one non-borderline case: 94.3% vs. 72.2% (RR: 1.31, 95% CI: 1.14-1.50, p = 0.00) and 18.8% vs. 5.1% (RR: 3.09, 95% CI: 1.17-8.18, p = 0.01) of participants in the experimental and control groups respectively recommended an intervention. For the remaining case, there was no difference between the experimental and control group and the majority of participants recommended intervention. In sub-group analyses, findings were not affected by the experience level of participating cardiologists. Despite existing clinical guidelines, the therapy recommendations of the participating physicians were heterogeneous. Validated patient-specific computational modelling has the potential to influence treatment decisions. Future studies in broader areas are needed to evaluate whether differences in decisions result in improved outcomes (Trial Registration: NCT02700737).

Evaluation of 4D flow MRI-based non-invasive pressure assessment in aortic coarctations

Severity of aortic coarctation (CoA) is currently assessed by estimating trans-coarctation pressure drops through cardiac catheterization or echocardiography. In principle, more detailed information could be obtained non-invasively based on space- and time-resolved magnetic resonance imaging (4D flow) data. Yet the limitations of this imaging technique require testing the accuracy of 4D flow-derived hemodynamic quantities against other methodologies. With the objective of assessing the feasibility and accuracy of this non-invasive method to support the clinical diagnosis of CoA, we developed an algorithm (4DF-FEPPE) to obtain relative pressure distributions from 4D flow data by solving the Poisson pressure equation. 4DF-FEPPE was tested against results from a patient-specific fluid-structure interaction (FSI) simulation, whose patient-specific boundary conditions were prescribed based on 4D flow data. Since numerical simulations provide noise-free pressure fields on fine spatial and temporal scales, our analysis allowed to assess the uncertainties related to 4D flow noise and limited resolution. 4DF-FEPPE and FSI results were compared on a series of cross-sections along the aorta. Bland-Altman analysis revealed very good agreement between the two methodologies in terms of instantaneous data at peak systole, end-diastole and time-averaged values: biases (means of differences) were +0.4 mmHg, -1.1 mmHg and +0.6 mmHg, respectively. Limits of agreement (2 SD) were ±0.978 mmHg, ±1.06 mmHg and ±1.97 mmHg, respectively. Peak-to-peak and maximum trans-coarctation pressure drops obtained with 4DF-FEPPE differed from FSI results by 0.75 mmHg and -1.34 mmHg respectively. The present study considers important validation aspects of non-invasive pressure difference estimation based on 4D flow MRI, showing the potential of this technology to be more broadly applied to the clinical practice.