Summary of CFD modelling applications in cardiovascular medicine
| Area | Clinical applications | Data and evidence | Potential clinical impact | Limitations and challenges | References |
|---|---|---|---|---|---|
| Coronary artery disease and physiology | Models based upon coronary angiography (CT /invasive) to compute physiological coronary lesion significance less invasively | Multiple trials demonstrating broadly good agreement between standard and CFD-derived FFR (vFFR). Lesion significance established in ∼80–90% | Widened access to the benefits of physiological lesion assessment; vFFR lacks the practical limitations that restrict use of the invasive technique. Virtual stenting enables planning and selection of optimal treatment strategy | Accurate vessel reconstruction and patient-specific tuning of the model boundary conditions (especially those of myocardial resistance) | 5 6 7 8 |
| Valve prostheses | Evaluation and optimisation of prosthetic valve design from a haemodynamic perspective | Included in the design dossier given to RA for approval before use in humans. Third party, comparative studies are in engineering literature | CFD modelling enables the best design, yielding the optimal haemodynamics and lowest achievable risk of design-related thrombosis and thromboembolism | Dependence upon validity of models to interpret fluid stresses in terms of thrombogenic/haemolytic potential. Primarily relates to mechanical valves. Tissue valve leaflets remain challenging to model | 9 10 11 |
| Native valve haemodynamics in health and disease | Non-invasive computation and quantification of trans-valvular pressure drop and regurgitant fraction from CT imaging | Accurate 3D simulations in patient-specific models with valves in open and closed states to predict transvalvular dynamics in diseased states | Improved objective assessment and surveillance of valve disease from non-invasive imaging data | Requires high quality 3D images of valve orifice—not routinely generated. Balancing the requirement for complex dynamic simulation (FSI) vs simpler models (valve open/closed) | 12 13 |
| Aortic aneurysm | Provides quantitative haemodynamic data for non-invasive imaging to emphasise the significance of findings. Virtual therapy simulation/predictions | No published outcome trials, only single centre experiences and small cohorts using different boundary condition and computational methods | To better predict aneurysm progression and risk of rupture. Prediction of putative therapeutic effects. Individualised care and reduction in costs for unnecessary follow-up imaging and visits | Impact of low image contrast structures of aortic aneurysm (eg, wall, thrombus) as well as wall motion needs to be further assessed. CFD alone is probably limited and needs to be complemented by, for example, FSI | 3 14 15 16 |
| Aortic dissection | Pathophysiological conditions in true and false lumen computed from non-invasive boundary conditions (CT and MRI+PC). Effects of virtual therapy. | No published outcome trials, only single centre experiences and small cohorts using different boundary conditions and computational methods | Computed pressure and flow conditions used to guide (semi-) invasive therapeutic procedure decisions. Physiological effects of therapies can be simulated and better predicted | Significant early and late re-modelling of the dissected wall. Entry, re-entry and communication channels create a complex computational scenario. CFD alone might be limited. A potential role for FSI | 17 18 19 |
| Stent design | Prediction of WSS and related metrics that influence endothelial function and NH due to stent-induced haemodynamic disturbance | Turbulent or disturbed laminar flow reduces WSS stimulating adverse vessel remodelling. NH preferentially accumulates in these regions | Not possible to measure arterial WSS in vivo, especially in the vicinity of stent struts post-PCI. Modelling provides detailed analysis of flow, and the influence of stent design through patient-specific reconstructions, enabling the optimal stent design to be achieved | High resolution imaging, vessel reconstruction and boundary conditions are challenging. CFD simulations demand fine computational meshes and time-resolved pulsatility. Run-times are long, even with high performance computing | 20 21 22 23 24 25 26 27 |
| Cerebral aneurysm | Prediction of intra-aneurysmal flow, stasis, jet impingement and WSS from MRI and CT cerebral angiography data | Published data on association between WSS, aneurysm initiation, growth, and potentially rupture | Detailed, individualised haemodynamic analysis with potential for risk prediction. Impact of putative treatments on local haemodynamics evaluated in silico | Difficulty interpreting complex and detailed WSS results. Understanding how results translate to rupture risk. Validation of rupture predictions—a rare event | 4 28 29 |
| Pulmonary hypertension (PH) | Greater insights into complex PH physiology. Increasing interest in non-invasive diagnosis and monitoring of response to treatment | Models based on MR flows demonstrated to differentiate between healthy volunteers and to stratify PH subcategories | Imaging-based modelling of pulmonary haemodynamics can reduce the requirement for right heart catheterisation. Models show association between reduced WSS and invasive PH metrics. PH subtype characteristics simulated to understand the structural changes contributing to increased PAP | Spatial resolution of imaging and segmentation protocols. The use of a pressure surrogate measure. The presence of many outlets requiring many measurements to tune the outflow boundary conditions | 30 31 32 33 |
| Arterial wall shear stress (WSS) | WSS mapping, cross-referenced with vascular disease phenotype, is contributing to the understanding of cellular biology | An abnormal WSS pattern has been correlated with vascular diseases, including atherosclerosis, aneurysm and post-stent NH | Ultimate understanding of the development and progression of atherosclerosis. WSS map combined with multi-scale modelling may inform clinical practice, such as the site of rupture in aneurysm, and severity of in-stent restenosis. | A detailed vascular geometry is essential for an accurate WSS map. Acquisition of patient specific boundary conditions remains clinically challenging. | 20 21 22 25 |
| Heart failure | Models based upon CT and MR help compute haemodynamics and the spatio-temporal distributions of pressure and myocardial stress/strain | CFD/FSI models replicate realistic pathophysiology in models of health and disease (eg, HFREF, HFPEF, HCM, DCM, and RWMA post-MI) | Additional haemodynamic data potentially enables early diagnosis and stratifies disease phenotypes and severities. Characterising complex vortex flows identifies areas of flow stagnation and thrombus risk | Resolution of imaging and reconstruction (representing trabeculae and papillary muscles). Tuning with realistic boundary conditions. Requirement for FSI in many models | 2 34 35 36 37 38 39 40 41 |
| CRT | Coupled electro-mechanical models of the ventricle incorporating CFD (multi-physics models) used to investigate heart function | Published reports of accurate patient-specific haemodynamic simulations with sufficient detail to optimise CRT before surgical intervention | Improved selection of CRT responders. Simulation and selection of optimal tuning of device settings and lead placement on an individual case basis | Uncertainties and assumptions regarding boundary conditions and the range of clinical measurements required for parameterisation. Mesh generation, prolonged computation times | |
| VADs | Generic optimisation of pump design. Patient-specific models can aid implantation strategy and tuning of output according to patient physiology | Published models describing haemodynamic influences of catheter placement and minimisation of adverse haemodynamic effects | Pump tuning to ensure periodic opening and closing of AV, preventing leaflet fusion. Personalised catheter placement planning (prediction and avoidance stasis and thrombus formation) | Post-implantation imaging artefact limits modelling. Optimising performance requires the balance of multiple competing factors. As for all cardiac electromechanical models, selection of appropriate patient specific parameters is difficult due to sparsity of data | |
| Congenital heart disease | CFD simulates haemodynamics which are complex and hard to predict in the context of a diverse and heterogeneous range of disease phenotypes | Range of models described, including reduced order, 3D CFD, FSI and multiscale, particularly in the context of univentricular circulation, aortic and pulmonary malformations | Modelling enables greater understanding of systemic and regional haemodynamics and the prediction of response to putative surgical or device-based treatments which often involve significant modifications to the circulatory tree | Acquisition and application of model parameters and boundary conditions from patient and literature data. The ultimate personalisation challenge | 42 43 44 |
AV, aortic valve; CFD, computational fluid dynamics; CRT, cardiac resynchronisation therapy; CT (A), CT (angiography); DCM, dilated cardiomyopathy; FSI, fluid solid interaction; HCM, hypertrophic cardiomyopathy; HFPEF, heart failure with preserved EF; HFREF, heart failure with reduced EF; MI, myocardial infarction; NH, neointimal hyperplasia; PAP, pulmonary artery pressure; PC, phase-contrast; PCI, percutaneous coronary intervention; RA, regulatory authority; RWMA, regional wall motion abnormality; (v)FFR; (virtual) fractional flow reserve; VAD, ventricular assist device; WSS, wall shear stress.