Shape-driven deep neural networks for fast acquisition of aortic 3D pressure and velocity flow fields.

Computational fluid dynamics (CFD) can be used to simulate vascular haemodynamics and analyse potential treatment options. CFD has shown to be beneficial in improving patient outcomes. However, the implementation of CFD for routine clinical use is yet to be realised. Barriers for CFD include high computational resources, specialist experience needed for designing simulation set-ups, and long processing times. The aim of this study was to explore the use of machine learning (ML) to replicate conventional aortic CFD with automatic and fast regression models. Data used to train/test the model consisted of 3,000 CFD simulations performed on synthetically generated 3D aortic shapes. These subjects were generated from a statistical shape model (SSM) built on real patient-specific aortas (N = 67). Inference performed on 200 test shapes resulted in average errors of 6.01% ±3.12 SD and 3.99% ±0.93 SD for pressure and velocity, respectively. Our ML-based models performed CFD in ∼0.075 seconds (4,000x faster than the solver). This proof-of-concept study shows that results from conventional vascular CFD can be reproduced using ML at a much faster rate, in an automatic process, and with reasonable accuracy.

Automatic segmentation of the great arteries for computational hemodynamic assessment.

BACKGROUND: Computational fluid dynamics (CFD) is increasingly used for the assessment of blood flow conditions in patients with congenital heart disease (CHD). This requires patient-specific anatomy, typically obtained from segmented 3D cardiovascular magnetic resonance (CMR) images. However, segmentation is time-consuming and requires expert input. This study aims to develop and validate a machine learning (ML) method for segmentation of the aorta and pulmonary arteries for CFD studies. METHODS: 90 CHD patients were retrospectively selected for this study. 3D CMR images were manually segmented to obtain ground-truth (GT) background, aorta and pulmonary artery labels. These were used to train and optimize a U-Net model, using a 70-10-10 train-validation-test split. Segmentation performance was primarily evaluated using Dice score. CFD simulations were set up from GT and ML segmentations using a semi-automatic meshing and simulation pipeline. Mean pressure and velocity fields across 99 planes along the vessel centrelines were extracted, and a mean average percentage error (MAPE) was calculated for each vessel pair (ML vs GT). A second observer (SO) segmented the test dataset for assessment of inter-observer variability. Friedman tests were used to compare ML vs GT, SO vs GT and ML vs SO metrics, and pressure/velocity field errors. RESULTS: The network's Dice score (ML vs GT) was 0.945 (interquartile range: 0.929-0.955) for the aorta and 0.885 (0.851-0.899) for the pulmonary arteries. Differences with the inter-observer Dice score (SO vs GT) and ML vs SO Dice scores were not statistically significant for either aorta or pulmonary arteries (p = 0.741, p = 0.061). The ML vs GT MAPEs for pressure and velocity in the aorta were 10.1% (8.5-15.7%) and 4.1% (3.1-6.9%), respectively, and for the pulmonary arteries 14.6% (11.5-23.2%) and 6.3% (4.3-7.9%), respectively. Inter-observer (SO vs GT) and ML vs SO pressure and velocity MAPEs were of a similar magnitude to ML vs GT (p > 0.2). CONCLUSIONS: ML can successfully segment the great vessels for CFD, with errors similar to inter-observer variability. This fast, automatic method reduces the time and effort needed for CFD analysis, making it more attractive for routine clinical use.

Computational investigation of the haemodynamics shows criticalities of central venous lines used for chronic haemodialysis in children.

Background: Haemodialysis is a life-saving treatment for children with kidney failure. The majority of children have haemodialysis through central venous lines (CVLs). The use of CVLs in pediatric patients is often associated to complications which can lead to their replacement. The aim of this study is to investigate haemodynamics of pediatric CVLs to highlight the criticalities of different line designs. Methods: Four models of CVLs for pediatric use were included in this study. The selected devices varied in terms of design and sizes (from 6.5 Fr to 14 Fr). Accurate 3D models of CVLs were reconstructed from high-resolution images including venous and arterial lumens, tips and side holes. Computational fluid dynamics (CFD) analyses were carried out to simulate pediatric working conditions of CVLs in ideal and anatomically relevant conditions. Results: The arterial lumens of all tested CVLs showed the most critical conditions with the majority of blood flowing through the side-holes. A zone of low flow was identified at the lines' tip. The highest shear stresses distribution (>10 Pa) was found in the 8 Fr line while the highest platelet lysis index in the 10 Fr model. The analysis on the anatomical geometry showed an increase in wall shear stress measured in the 10 F model compared to the idealised configuration. Similarly, in anatomical models an increased disturbance and velocity of the flow was found inside the vein after line placement. Conclusion: This study provided a numerical characterization of fluid dynamics in pediatric CVLs highlighting performance criticalities (i.e. high shear stresses and areas of stagnation) associated to specific sizes (8 Fr and 10 Fr) and conditions (i.e. anatomical test).

Characterization of Flow Dynamics in the Pulmonary Bifurcation of Patients With Repaired Tetralogy of Fallot: A Computational Approach.

The hemodynamic environment of the pulmonary bifurcation is of great importance for adult patients with repaired tetralogy of Fallot (rTOF) due to possible complications in the pulmonary valve and narrowing of the left pulmonary artery (LPA). The aim of this study was to computationally investigate the effect of geometrical variability and flow split on blood flow characteristics in the pulmonary trunk of patient-specific models. Data from a cohort of seven patients was used retrospectively and the pulmonary hemodynamics was investigated using averaged and MRI-derived patient-specific boundary conditions on the individualized models, as well as a statistical mean geometry. Geometrical analysis showed that curvature and tortuosity are higher in the LPA branch, compared to the right pulmonary artery (RPA), resulting in complex flow patterns in the LPA. The computational analysis also demonstrated high time-averaged wall shear stress (TAWSS) at the outer wall of the LPA and the wall of the RPA proximal to the junction. Similar TAWSS patterns were observed for averaged boundary conditions, except for a significantly modified flow split assigned at the outlets. Overall, this study enhances our understanding about the flow development in the pulmonary bifurcation of rTOF patients and associates some morphological characteristics with hemodynamic parameters, highlighting the importance of patient-specificity in the models. To confirm these findings, further studies are required with a bigger cohort of patients.

Effects of Uncertainty of Outlet Boundary Conditions in a Patient-Specific Case of Aortic Coarctation.

Computational Fluid Dynamics (CFD) simulations of blood flow are widely used to compute a variety of hemodynamic indicators such as velocity, time-varying wall shear stress, pressure drop, and energy losses. One of the major advances of this approach is that it is non-invasive. The accuracy of the cardiovascular simulations depends directly on the level of certainty on input parameters due to the modelling assumptions or computational settings. Physiologically suitable boundary conditions at the inlet and outlet of the computational domain are needed to perform a patient-specific CFD analysis. These conditions are often affected by uncertainties, whose impact can be quantified through a stochastic approach. A methodology based on a full propagation of the uncertainty from clinical data to model results is proposed here. It was possible to estimate the confidence associated with model predictions, differently than by deterministic simulations. We evaluated the effect of using three-element Windkessel models as the outflow boundary conditions of a patient-specific aortic coarctation model. A parameter was introduced to calibrate the resistances of the Windkessel model at the outlets. The generalized Polynomial Chaos method was adopted to perform the stochastic analysis, starting from a few deterministic simulations. Our results show that the uncertainty of the input parameter gave a remarkable variability on the volume flow rate waveform at the systolic peak simulating the conditions before the treatment. The same uncertain parameter had a slighter effect on other quantities of interest, such as the pressure gradient. Furthermore, the results highlight that the fine-tuning of Windkessel resistances is not necessary to simulate the post-stenting scenario.

A workflow for patient-specific fluid-structure interaction analysis of the mitral valve: A proof of concept on a mitral regurgitation case.

The mechanics of the mitral valve (MV) are the result of the interaction of different anatomical structures complexly arranged within the left heart (LH), with the blood flow. MV structure abnormalities might cause valve regurgitation which in turn can lead to heart failure. Patient-specific computational models of the MV could provide a personalised understanding of MV mechanics, dysfunctions and possible interventions. In this study, we propose a semi-automatic pipeline for MV modelling based on the integration of state-of-the-art medical imaging, i.e. cardiac magnetic resonance (CMR) and 3D transoesophageal-echocardiogram (TOE) with fluid-structure interaction (FSI) simulations. An FSI model of a patient with MV regurgitation was implemented using the finite element (FE) method and smoothed particle hydrodynamics (SPH). Our study showed the feasibility of combining image information and computer simulations to reproduce patient-specific MV mechanics as seen on medical images, and the potential for efficient in-silico studies of MV disease, personalised treatments and device design.

Patient-specific simulations for planning treatment in congenital heart disease.

Patient-specific computational models have been extensively developed over the last decades and applied to investigate a wide range of cardiovascular problems. However, translation of these technologies into clinical applications, such as planning of medical procedures, has been limited to a few single case reports. Hence, the use of patient-specific models is still far from becoming a standard of care in clinical practice. The aim of this study is to describe our experience with a modelling framework that allows patient-specific simulations to be used for prediction of clinical outcomes. A cohort of 12 patients with congenital heart disease who were referred for percutaneous pulmonary valve implantation, stenting of aortic coarctation and surgical repair of double-outlet right ventricle was included in this study. Image data routinely acquired for clinical assessment were post-processed to set up patient-specific models and test device implantation and surgery. Finite-element and computational fluid dynamics analyses were run to assess feasibility of each intervention and provide some guidance. Results showed good agreement between simulations and clinical decision including feasibility, device choice and fluid-dynamic parameters. The promising results of this pilot study support translation of computer simulations as tools for personalization of cardiovascular treatments.