On the Role and Effects of Uncertainties in Cardiovascular in silico Analyses

Fabrication of deformable patient-specific AAA models by material casting techniques

Background

Abdominal Aortic Aneurysm (AAA) is a balloon-like dilatation that can be life-threatening if not treated. Fabricating patient-specific AAA models can be beneficial for in-vitro investigations of hemodynamics, as well as for pre-surgical planning and training, testing the effectiveness of different interventions, or developing new surgical procedures. The current direct additive manufacturing techniques cannot simultaneously ensure the flexibility and transparency of models required by some applications. Therefore, casting techniques are presented to overcome these limitations and make the manufactured models suitable for in-vitro hemodynamic investigations, such as particle image velocimetry (PIV) measurements or medical imaging.

Methods

Two complex patient-specific AAA geometries were considered, and the related 3D models were fabricated through material casting. In particular, two casting approaches, i.e. lost molds and lost core casting, were investigated and tested to manufacture the deformable AAA models. The manufactured models were acquired by magnetic resonance, computed tomography (CT), ultrasound imaging, and PIV. In particular, CT scans were segmented to generate a volumetric reconstruction for each manufactured model that was compared to a reference model to assess the accuracy of the manufacturing process.

Results

Both lost molds and lost core casting techniques were successful in the manufacturing of the models. The lost molds casting allowed a high-level surface finish in the final 3D model. In this first case, the average signed distance between the manufactured model and the reference was (- 0.2 ± 0.2 ) mm. However, this approach was more expensive and time-consuming. On the other hand, the lost core casting was more affordable and allowed the reuse of the external molds to fabricate multiple copies of the same AAA model. In this second case, the average signed distance between the manufactured model and the reference was (0.1 ± 0.6 ) mm. However, the final model's surface finish quality was poorer compared to the model obtained by lost molds casting as the sealing of the outer molds was not as firm as the other casting technique.

Conclusions

Both lost molds and lost core casting techniques can be used for manufacturing patient-specific deformable AAA models suitable for hemodynamic investigations, including medical imaging and PIV.

Calibration of patient-specific boundary conditions for coupled CFD models of the aorta derived from 4D Flow-MRI

Introduction: Patient-specific computational fluid dynamics (CFD) models permit analysis of complex intra-aortic hemodynamics in patients with aortic dissection (AD), where vessel morphology and disease severity are highly individualized. The simulated blood flow regime within these models is sensitive to the prescribed boundary conditions (BCs), so accurate BC selection is fundamental to achieve clinically relevant results. Methods: This study presents a novel reduced-order computational framework for the iterative flow-based calibration of 3-Element Windkessel Model (3EWM) parameters to generate patient-specific BCs. These parameters were calibrated using time-resolved flow information derived from retrospective four-dimensional flow magnetic resonance imaging (4D Flow-MRI). For a healthy and dissected case, blood flow was then investigated numerically in a fully coupled zero dimensional-three dimensional (0D-3D) numerical framework, where the vessel geometries were reconstructed from medical images. Calibration of the 3EWM parameters was automated and required ~3.5 min per branch. Results: With prescription of the calibrated BCs, the computed near-wall hemodynamics (time-averaged wall shear stress, oscillatory shear index) and perfusion distribution were consistent with clinical measurements and previous literature, yielding physiologically relevant results. BC calibration was particularly important in the AD case, where the complex flow regime was captured only after BC calibration. Discussion: This calibration methodology can therefore be applied in clinical cases where branch flow rates are known, for example, via 4D Flow-MRI or ultrasound, to generate patient-specific BCs for CFD models. It is then possible to elucidate, on a case-by-case basis, the highly individualized hemodynamics which occur due to geometric variations in aortic pathology high spatiotemporal resolution through CFD.

Uncertainty Quantification in the In Vivo Image-Based Estimation of Local Elastic Properties of Vascular Walls

Introduction: Patient-specific computational models are a powerful tool for planning cardiovascular interventions. However, the in vivo patient-specific mechanical properties of vessels represent a major source of uncertainty. In this study, we investigated the effect of uncertainty in the elastic module (E) on a Fluid–Structure Interaction (FSI) model of a patient-specific aorta. Methods: The image-based χ-method was used to compute the initial E value of the vascular wall. The uncertainty quantification was carried out using the generalized Polynomial Chaos (gPC) expansion technique. The stochastic analysis was based on four deterministic simulations considering four quadrature points. A deviation of about ±20% on the estimation of the E value was assumed. Results: The influence of the uncertain E parameter was evaluated along the cardiac cycle on area and flow variations extracted from five cross-sections of the aortic FSI model. Results of stochastic analysis showed the impact of E in the ascending aorta while an insignificant effect was observed in the descending tract. Conclusions: This study demonstrated the importance of the image-based methodology for inferring E, highlighting the feasibility of retrieving useful additional data and enhancing the reliability of in silico models in clinical practice.

Learning Whole Heart Mesh Generation From Patient Images for Computational Simulations

Patient-specific cardiac modeling combines geometries of the heart derived from medical images and biophysical simulations to predict various aspects of cardiac function. However, generating simulation-suitable models of the heart from patient image data often requires complicated procedures and significant human effort. We present a fast and automated deep-learning method to construct simulation-suitable models of the heart from medical images. The approach constructs meshes from 3D patient images by learning to deform a small set of deformation handles on a whole heart template. For both 3D CT and MR data, this method achieves promising accuracy for whole heart reconstruction, consistently outperforming prior methods in constructing simulation-suitable meshes of the heart. When evaluated on time-series CT data, this method produced more anatomically and temporally consistent geometries than prior methods, and was able to produce geometries that better satisfy modeling requirements for cardiac flow simulations. Our source code and pretrained networks are available at https://github.com/fkong7/HeartDeformNets.

An image-based approach for the estimation of arterial local stiffness in vivo

The analysis of mechanobiology of arterial tissues remains an important topic of research for cardiovascular pathologies evaluation. In the current state of the art, the gold standard to characterize the tissue mechanical behavior is represented by experimental tests, requiring the harvesting of ex-vivo specimens. In recent years though, image-based techniques for the in vivo estimation of arterial tissue stiffness were presented. The aim of this study is to define a new approach to provide local distribution of arterial stiffness, estimated as the linearized Young's Modulus, based on the knowledge of in vivo patient-specific imaging data. In particular, the strain and stress are estimated with sectional contour length ratios and a Laplace hypothesis/inverse engineering approach, respectively, and then used to calculate the Young's Modulus. After describing the method, this was validated by using a set of Finite Element simulations as input. In particular, idealized cylinder and elbow shapes plus a single patient-specific geometry were simulated. Different stiffness distributions were tested for the simulated patient-specific case. After the validation from Finite Element data, the method was then applied to patient-specific ECG-gated Computed Tomography data by also introducing a mesh morphing approach to map the aortic surface along the cardiac phases. The validation process revealed satisfactory results. In the simulated patient-specific case, root mean square percentage errors below 10% for the homogeneous distribution and below 20% for proximal/distal distribution of stiffness. The method was then successfully used on the three ECG-gated patient-specific cases. The resulting distributions of stiffness exhibited significant heterogeneity, nevertheless the resulting Young's moduli were always contained within the 1-3 MPa range, which is in line with literature.

CT-Based Simulation of Left Ventricular Hemodynamics: A Pilot Study in Mitral Regurgitation and Left Ventricle Aneurysm Patients

Background

Cardiac CT (CCT) is well suited for a detailed analysis of heart structures due to its high spatial resolution, but in contrast to MRI and echocardiography, CCT does not allow an assessment of intracardiac flow. Computational fluid dynamics (CFD) can complement this shortcoming. It enables the computation of hemodynamics at a high spatio-temporal resolution based on medical images. The aim of this proposed study is to establish a CCT-based CFD methodology for the analysis of left ventricle (LV) hemodynamics and to assess the usability of the computational framework for clinical practice.

Materials and Methods

The methodology is demonstrated by means of four cases selected from a cohort of 125 multiphase CCT examinations of heart failure patients. These cases represent subcohorts of patients with and without LV aneurysm and with severe and no mitral regurgitation (MR). All selected LVs are dilated and characterized by a reduced ejection fraction (EF). End-diastolic and end-systolic image data was used to reconstruct LV geometries with 2D valves as well as the ventricular movement. The intraventricular hemodynamics were computed with a prescribed-motion CFD approach and evaluated in terms of large-scale flow patterns, energetic behavior, and intraventricular washout.

Results

In the MR patients, a disrupted E-wave jet, a fragmentary diastolic vortex formation and an increased specific energy dissipation in systole are observed. In all cases, regions with an impaired washout are visible. The results furthermore indicate that considering several cycles might provide a more detailed view of the washout process. The pre-processing times and computational expenses are in reach of clinical feasibility.

Conclusion

The proposed CCT-based CFD method allows to compute patient-specific intraventricular hemodynamics and thus complements the informative value of CCT. The method can be applied to any CCT data of common quality and represents a fair balance between model accuracy and overall expenses. With further model enhancements, the computational framework has the potential to be embedded in clinical routine workflows, to support clinical decision making and treatment planning.