Population‐specific modelling of between/within‐subject flow variability in the carotid arteries of the elderly

Abstract

Computational fluid dynamics models are increasingly proposed for assisting the diagnosis and management of vascular diseases. Ideally, patient-specific flow measurements are used to impose flow boundary conditions. When patient-specific flow measurements are unavailable, mean values of flow measurements across small cohorts are used as normative values. In reality, both the between-subjects and within-subject flow variabilities are large. Consequently, neither one-shot flow measurements nor mean values across a cohort are truly indicative of the flow regime in a given person. We develop models for both the between-subjects and within-subject variability of internal carotid flow. A log-linear mixed effects model is combined with a Gaussian process to model the between-subjects flow variability, while a lumped parameter model of cerebral autoregulation is used to model the within-subject flow variability in response to heart rate and blood pressure changes. The model parameters are identified from carotid ultrasound measurements in a cohort of 103 elderly volunteers. We use the models to study intracranial aneurysm flow in 54 subjects under rest and exercise and conclude that OSI, a common wall shear-stress derived quantity in vascular CFD studies, may be too sensitive to flow fluctuations to be a reliable biomarker.KEYWORDS cerebrovascular disease, computational fluid dynamics, Gaussian process models, patient-specific models, uncertainty quantification Numerous computational fluid dynamics (CFD) models of human cardiovascular and cerebrovascular physiology are published every year, but few of them make any impact in clinical practice. This inconvenient truth has been blamed on the improper use of CFD solvers, 1 insufficient validation of biomechanics models, 2 and lack of understanding of the clinical decision-making process by the biomedical engineers building the models. 3 One additional explanation is that many "patient-specific" CFD models fail to consider the physiological variability of vascular flow, confounding interpretation of model results and producing overly confident predictions of flow quantities.Patient-specific modelling of vascular flow requires an accurate description of the lumen plus the definition of boundary conditions. The latter is done by measuring patient-specific flow waveforms using either phase contrast magnetic resonance imaging (pcMRI) or ultrasound-based flow measurement techniques. When patient-specific flow measurements are not available, cohort-averaged values of flow from the literature are often used. For example, the small-scale studies 4-6 Int J Numer Meth Biomed Engng. 2020;36:e3271.wileyonlinelibrary.com/journal/cnm

In-silico trial of intracranial flow diverters replicates and expands insights from conventional clinical trials

“…Virtual PED flow diverters are implanted using the algorithm from 38 , followed by lumen mesh generation. Carotid flow boundary conditions are generated using the model from 54 , representing patients with both normotensive and hypertensive conditions (N = 164). The 3-D Navier-Stokes equations are solved and the results post-processed to measure the aneurysm mean velocity reduction (AMVR).…”

“…Normotensive ICA flow waveforms were taken as the mean of a virtual population of waveforms generated by a Gaussian process model (GPM) trained on patient-specific phase-contrast magnetic resonance imaging measurements of ICA flow in 17 healthy young adults (age = 28 ± 7 years). Details on this GPM and its validation are reported in 35,54 . To maintain a physiologically realistic flow specific to each virtual patient, Poiseuille's law was used to scale the GPM-generated waveforms such that the time-averaged wall shear stress was the physiological value of 1.5 Pa at the ICA inlet.…”

Turbulence Descriptors in Arterial Flows : Patient-Specific Computational Hemodynamics

“…[224,106,104,23]). With this in mind, the hemodynamic descriptors outlined in this thesis should preferably be analyzed against, for example, the most sensitive inputs, which have be proposed to be foremost associated with the inflow conditions [132] and geometry uncertainties [147]. Also, UQ related to the LES subgrid modeling alone [159] would be interesting to examine for cardiovascular flow predictions.…”

In-silico trial of intracranial flow diverters confirms and expands insights from conventional clinical trials

“…Normotensive ICA flow waveforms were taken as the mean of a virtual population of waveforms generated by a Gaussian process model (GPM) trained on patient-specific phase-contrast magnetic resonance imaging (MRI) measurements of ICA flow in 17 healthy young adults (age = 28±7 yr). Details on this GPM and its validation are reported in [34,43]. To maintain a physiologically realistic flow specific to each virtual patient, Poiseuille's law was used to scale the GPM-generated waveforms such that the time-averaged wall shear stress (WSS) was the physiological value of 1.5 Pa at the ICA inlet.…”

“…However, the mean age of the population we used for modelling normotensive flow conditions was 28±7. When producing hypertensive flow conditions for each virtual anatomy, the dependency on age was modelled using the methods detailed in [34]. However, the effect of age on flow diversion outcomes has previously been hypothesised to be related to inadequate endothelialisation in elderly individuals [19,36], which was beyond the scope of the FD-PASS trial.…”