G Model
ARTICLE IN PRESS
CMIG-1305; No. of Pages 13
Computerized Medical Imaging and Graphics xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Computerized Medical Imaging and Graphics journal homepage: www.elsevier.com/locate/compmedimag
Computational fluid dynamics in coronary artery disease Zhonghua Sun a , Lei Xu b,∗ a b
Discipline of Medical Imaging, Department of Imaging and Applied Physics, Curtin University, Perth, Western Australia 6845, Australia Department of Radiology, Beijing Anzhen Hospital, Capital Medical University, Beijing, China
a r t i c l e
i n f o
Article history: Received 5 July 2014 Received in revised form 22 August 2014 Accepted 3 September 2014 Keywords: Atherosclerosis Coronary artery disease Computational fluid dynamics Haemodynamics Plaque
a b s t r a c t Computational fluid dynamics (CFD) is a widely used method in mechanical engineering to solve complex problems by analysing fluid flow, heat transfer, and associated phenomena by using computer simulations. In recent years, CFD has been increasingly used in biomedical research of coronary artery disease because of its high performance hardware and software. CFD techniques have been applied to study cardiovascular haemodynamics through simulation tools to predict the behaviour of circulatory blood flow in the human body. CFD simulation based on 3D luminal reconstructions can be used to analyse the local flow fields and flow profiling due to changes of coronary artery geometry, thus, identifying risk factors for development and progression of coronary artery disease. This review aims to provide an overview of the CFD applications in coronary artery disease, including biomechanics of atherosclerotic plaques, plaque progression and rupture; regional haemodynamics relative to plaque location and composition. A critical appraisal is given to a more recently developed application, fractional flow reserve based on CFD computation with regard to its diagnostic accuracy in the detection of haemodynamically significant coronary artery disease. © 2014 Elsevier Ltd. All rights reserved.
1. Introduction Coronary artery disease is the leading cause of morbidity and mortality in advanced countries and its prevalence is increasing in developing countries. Coronary artery disease is responsible for 7.3 million deaths and 58 million disability-adjusted life years lost worldwide [1]. Despite its large socio-economic impact, the underlying mechanisms of coronary artery disease are only partially understood. It is generally accepted that coronary artery disease is an inflammatory disease with lipid deposition in the arterial wall as an initial stage of atherosclerosis [2–5]. Although the risk factors for atherosclerotic coronary plaque formation, including high cholesterol, diabetes, and hypertension are systemic in nature, plaques are located at specific sites in the coronary artery where disturbed flow and low endothelial shear stress occur (Fig. 1) [6–9]. In recent years, blood flow/shear stress has gained a lot of interest as complementary explanation for plaque formation [10,11]. The role of blood flow in the development of atherosclerosis is based on the observation that vascular inflammation and plaques are distributed near side branches or arterials stenosis, where blood
∗ Corresponding author at: Department of Radiology, Beijing Anzhen Hospital, Capital Medical University, No. 2, Anzhen Road, Chaoyang District, 100029 Beijing, China. Tel.: +86 10 64456071; fax: +86 10 64456962. E-mail addresses: [email protected] (Z. Sun), [email protected] (L. Xu).
flow is non-uniform, and at the lesser curvature of bends where blood flow rate is relatively low [12,13]. The effect of blood flow on the vessel wall is through shear stress, which influences the behaviour of endothelial cells including morphological adaptations and physiological changes of endothelial cells. Shear stress induces a shearing deformation of the endothelial cells which affects the phenotype of the endothelial cells and therefore the inflammatory component and plaque progression [14]. Low and oscillatory shear stress represent major features of the haemodynamic environment of regions opposite to arterial flow dividers that are predisposed to the formation of atherosclerosis (Fig. 2) [15–19]. It has been shown in vitro and in vivo studies that disturbed or oscillatory flows near arterial bifurcation, branch ostia and curvatures are associated with atheroma formation and intimal wall thickening [20–25]. The methods for in vivo estimation of wall shear stress can be performed by acquiring three-dimensional (3D) reconstruction of vessel volume using either invasive modalities such as intravascular ultrasound and invasive coronary angiography, or less-invasive techniques including coronary computed tomography angiography and cardiac magnetic resonance angiography with application of numerical methods to calculate flow within the reconstructed arterial volume for solving fluid dynamics [26,27]. The numerical methods are known as computational fluid dynamics (CFD) techniques. 3D reconstruction of coronary artery tree with subsequent numerical simulation using CFD techniques based on individual
http://dx.doi.org/10.1016/j.compmedimag.2014.09.002 0895-6111/© 2014 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Sun Z, Xu L. Computational fluid dynamics in coronary artery disease. Comput Med Imaging Graph (2014), http://dx.doi.org/10.1016/j.compmedimag.2014.09.002
G Model CMIG-1305; No. of Pages 13
ARTICLE IN PRESS Z. Sun, L. Xu / Computerized Medical Imaging and Graphics xxx (2014) xxx–xxx
2
Fig. 1. Molecular and cellular mechanisms through which low ESS promotes atherosclerosis. Data are derived from in vitro and animal in vivo experiments by applying a great variety of ESS patterns, both in direction and in magnitude. ESS, endothelial shear stress; BMP, bone morphogenic protein; eNOS, endothelial nitric oxide syntase; ET, endothelin; ICAM-1, intercellular adhesion molecule 1; IFN-g, interferon-g; IL, interleukin; LDL, low-density lipoprotein cholesterol; MCP, monocyte chemoattractant protein; MMP, matrix metalloproteinase; NF kB, nuclear factor-kB; NO, nitric oxide; PDGF, platelet-derived growth factor; ROS, reactive oxygen species; SREBP, sterol regulatory elements binding protein; TF, transcription factor; TGF-b, transforming growth factor b; TNF-a, tumour necrosis factor-a; t-PA, tissue plasminogen activator; VCAM, vascular cell adhesion molecule; VEGF, vascular endothelial growth factor; VSMC, vascular smooth muscle cell. Reprint with permission from Ref. [9].
patient data is being increasingly used to study haemodynamics and in predicting the behaviour of circulatory blood flow inside the coronary arteries [28–30]. This review article provides an overview of the applications of CFD in the diagnosis of coronary artery disease. The main focus of this review is to present an evidence-based review of the CFD in the biomechanics of atherosclerotic plaques and in the detection of high-risk coronary plaques and plaque progression with the aim of identifying high-risk patients so as to achieve the goal of reducing cardiac mortality. 2. Computational fluid dynamics CFD is a general term of all numerical techniques that are used to describe and analyse the flow of fluid elements at each location in certain geometry. The merit of CFD is developing new and improved devices and system designs, and optimisation is conducted on existing equipment through computational simulations resulting in enhanced efficiency and lower operating costs [31]. The governing equations of fluid dynamics can be computed to obtain coronary flow and pressure. These equations are called the
Navier–Stokes equations and have been known for more than 150 years. In order to simulate realistic coronary blood flow, a domain of interest must be defined, and boundary conditions specified. The isolation and generation of boundary condition is one of the greatest challenges in the integration of CFD for the assessment of the physiologic significance of coronary artery disease [32]. The coronary arterial wall is constantly exposed to both flowinduced wall shear stress (WSS) and arterial strain (relative displacement of wall within coronary artery) by blood pressure, myocardial contraction and local biological environment. Haemodynamic parameters of WSS, such as average wall shear stress (AWSS), average wall shear stress gradient (AWSSG), oscillatory shear index (OSI) and relative residence time (RRT) are possible indicators for atherosclerotic plaque prone sites [33–36]. Of these parameters, low WSS or OSI is a well-described mechanical stimulus that promotes the inflammatory process by inducing an oxidative response in endothelial vascular cells [37]. Lowering WSS has been reported to induce structural responses, according to several studies performed on cultured endothelial cells [38–40]. Thus, vascular remodelling is a response to alterations in WSS and
Fig. 2. Differential distribution of shear stress in a straight arterial segment proximal to a lumen-protruding atherosclerotic plaque. Reprint with permission from Ref. [16].
Please cite this article in press as: Sun Z, Xu L. Computational fluid dynamics in coronary artery disease. Comput Med Imaging Graph (2014), http://dx.doi.org/10.1016/j.compmedimag.2014.09.002
G Model CMIG-1305; No. of Pages 13
ARTICLE IN PRESS Z. Sun, L. Xu / Computerized Medical Imaging and Graphics xxx (2014) xxx–xxx
3
Fig. 3. Contour maps of maximal principal stresses. Comparison between a positive morphology model and the negative morphology models with concentric and eccentric plaques. The three cases presented have the same general dimensions: 75% lipid core, 46.6% stenosis ratio and 0.03 mm of fibrous cap thickness. Reprint with permission from Ref. [65].
other mechanical factors. This has led to the increasing applications of CFD in the investigation of pathogenesis of coronary artery disease, in particular, the identification and detection of vulnerable coronary plaques.
3. Definition of vulnerable or high-risk atherosclerotic plaque The vulnerable atherosclerotic plaque is called a “high-risk” or “thrombosis-prone” plaque. Major criteria to characterise such plaques include the presence of active inflammation (monocyte, macrophage or T-cell infiltration), a thin inflamed fibrous cap (40% of the total volume of the plaque), the presence of endothelial denudation with superficial platelet aggregation and the presence of haemodynamically significant stenosis (>90%) [41]. Minor criteria include the presence of spotty calcification, intraplaque haemorrhage, endothelial dysfunction and expansive remodelling [42,43]. A more clinical relevant definition of a vulnerable plaque is a lesion that places a patient at risk for developing future major adverse cardiac events, including death, myocardial infarction, or progressive angina. The identification of such plaques before they become symptomatic would enable prognostic stratification and facilitate primary prevention (e.g., aspirin, statins, and risk factor modification). Large lipid core and calcified areas (defined as >10% of the plaque area) and thin-cap fibroatheroma have been found to be associated with positive vascular remodelling [44–46]. Clinical and biomechanical studies have explored the risk factors associated with plaque vulnerability and have identified plaque composition and morphology as key determining factors for plaque vulnerability and likelihood of rupture [47–53]. Plaque characteristics can be determined by numerical simulation for evaluation of the vulnerability. WSS is one of the important physical factors in the process of atherosclerosis. Ex vivo studies have shown the interactions between shear stress distribution and changes in vascular anatomy and have demonstrated that high shear stress may be related to plaque rupture [54]. In vivo studies based on realistic patient’s data have confirmed these findings [52,53,55]. Abnormal WSS in geometrically susceptible coronary segments is thought to promote the development of atherosclerosis. Early plaque progression within a low WSS segment could result in gradual luminal narrowing that eventually leads to an increase in WSS into the physiologic range. As a plaque continues to encroach into the lumen, WSS is thought to increase into the high range [56]. The resultant high WSS has been demonstrated to induce apoptosis of smooth-muscle cells, increase matrix metalloproteinases which induce thinning of the plaque fibrous cap making it more vulnerable to fissuring or rupture [41,57–59]. Clinical studies support the
evidence that high WSS is associated with plaque vulnerability, plaque rupture and expansive vascular remodelling [60,61]. Vascular remodelling was first described in human autopsy studies as a compensatory change in vessel size as a result of increasing plaque burden and changes in local WSS in order to preserve the cross-sectional luminal area. It was reported that the development of luminal stenosis was delayed until the coronary plaque began to occupy 40% of the internal elastic lamina area [62]. This was confirmed by an in vivo study comparing intravascular ultrasound with clinical presentations with results showing that positive remodelling and larger plaque areas were associated with an unstable clinical presentation, while negative remodelling with a stable presentation [63]. Ohayon et al. studied the role of remodelling index on plaque stresses by considering positive and negative vascular remodelling as a parameter to affect plaque stability [64]. The authors investigated the biomechanical interaction between vessel and plaque geometry and the risk of plaque rupture. Their results showed that in the early stages of positive remodelling, lesions were more prone to rupture. This is supported by a recent study by Cilla et al. who examined the effect of positive and negative modelling on plaque stability by taking into account the arterial plaque morphologies [65]. Different stages and variations in atherosclerotic plaque remodelling were simulated with comparison of important determining parameters on plaque stresses including the fibrous cap thickness, the stenosis ratio, and lipid core dimensions and the atheroma plaque distribution. Their results support the findings by Ohayon et al. that coronary arteries with positive remodelling are more prone to rupture than those with negative remodelling for comparable geometrical conditions (Fig. 3).
4. CFD applications in coronary artery disease 4.1. Parameters used in CFD simulation For generation of coronary artery model, both fluid and structural domains are meshed with hexahedral cells so as to minimise numerical diffusion and lower the number of elements. Reliable blood flow data in the normal coronary arteries are required for conducting a fluid dynamics study, and this can be obtained with various imaging techniques, such as Doppler ultrasound or phasecontrast magnetic resonance imaging (MRI) [60,66]. The coronary artery flow pattern is significantly different from that in the aorta or peripheral arterial circulation, since blood supply to the heart occurs during diastole (Fig. 4). Flow in the coronary artery is unsteady as blood flow varies with the cardiac cycle. Most of the studies apply the cardiac cycle at the main aorta to provide both the inflow and outflow conditions to observe haemodynamic changes in the arterial system including coronary artery [67–72].
Please cite this article in press as: Sun Z, Xu L. Computational fluid dynamics in coronary artery disease. Comput Med Imaging Graph (2014), http://dx.doi.org/10.1016/j.compmedimag.2014.09.002
G Model CMIG-1305; No. of Pages 13 4
ARTICLE IN PRESS Z. Sun, L. Xu / Computerized Medical Imaging and Graphics xxx (2014) xxx–xxx
Fig. 4. Pulsatile blood flow waves used in the computational fluid dynamics simulation in the left coronary artery. T stands for one cardiac cycle; LCA, left coronary artery.
The inlet and outlet boundary conditions are based on a physiological pulsatile flow rate and pressure at the aorta [73]. As coronary artery disease most commonly occurs in the left coronary artery due to formation of bifurcation angle by two main coronary branches, left anterior descending (LAD) and left circumflex (LCx), a periodical pulsatile volume flow profile is applied as a velocity inlet boundary condition in the left coronary artery. Based on the average physiological human coronary flow data that are widely available in the literature [74], the average blood flow into the left coronary artery is reported to be 57 ml/min reaching a maximum value of 105 ml/min during the diastolic phases [75]. The blood flow distribution in the left bifurcation is calculated as 71% directed through the LAD and 29% through LCx according to previous reports [75,76]. Pulsatile aortic pressure is applied as an inlet boundary condition at the entrance of left main coronary artery, while pulsatile velocity conditions are imposed on both the LAD and LCx outlet boundaries (Fig. 4). A flow simulation is conducted over a time-span of several cardiac cycles, which are represented by time steps. The time steps can be divided into a number of coupling iterations, with each time step converged to a residual target of less than 1 × 10−4 by approximately 100 iterations [13,17,77]. A total of 600 time steps are required to achieve satisfactory convergence for fluid simulation when all velocity component changes from iteration to iteration are less than 10−6 [77,78]. The fluid–structure interaction (FSI) is an approach that can simultaneously model blood flow and arterial wall deformations, and it has received increasing interest in recent years due to its potential impact in the medical field [18,79–82]. Studies have shown the effect of vessel compliance on flow patterns and differences in WSS produced by FSI and rigid coronary models [83,84]. Dong et al. in their recent study based on idealised and realistic coronary models have demonstrated significant differences between FSI and rigid models, which are represented by noticeable qualitative discrepancies in WSS distributions over the bifurcation apex and the moderate narrow lumen site downstream to the LAD branch, and apparent quantitative differences in WSS profiles at the bifurcation apex over the diastole phase, respectively [76]. Furthermore, specific geometric features of coronary artery tree such as tortuosity or kinking would lead to flow alteration, thus, affecting the distribution of WSS [85–87]. These results further confirm the anatomical regions in the coronary artery which are prone to develop atherosclerotic lesions. 4.2. Clinical applications of CFD in coronary artery disease Mechanical forces and intravascular haemodynamics can affect and regulate blood vessels structure which induces a chronic
inflammatory response in the arterial walls resulting in atherosclerosis [6,88]. Applying CFD to study the blood flow in coronary artery disease is challenging as the coronary arteries are highly curved, movable and deformable during the heart beating. However, in recent years considerable insight has been gained on CFD studies in coronary artery disease, which is represented in two areas: development of CFD techniques and tremendous progress in applying CFD method to elucidate the role of haemodynamics in coronary artery disease development and progression [89]. Early CFD-based haemodynamic studies were conducted to represent in vitro conditions within restrictive assumptions [90–93]. Later reports demonstrated that CFD methods have been coupled with medical imaging techniques to provide detailed haemodynamic information which cannot be obtained from cardiac imaging alone. These studies have shown that CFD simulations have the potential to enhance the data obtained from in vivo methods (computed tomography [CT] or MRI) by providing a complete characterisation of haemodynamic conditions under precisely controlled conditions [94–98]. Knight et al. performed an analysis of the haemodynamic parameters including AWSSG, WSS and OSI obtained through a CFD study on the right coronary arteries of 30 patients undergoing coronary CT angiography [95]. These parameters were correlated to each patient’s specific plaque profile with the aim of predicting the particular plaque location. Their results showed a statistically significant difference between AWSSG and OSI in sensitivity and positive predictive value for the identification of atherosclerotic plaque sites in the right coronary artery. These findings further strengthen the theory that low WSS is a contributor to the initiation of atherosclerosis. In contrast to the right coronary artery, there is considerable variability in left main/LAD and LCx geometries among patients. A larger value of the angle between LAD and LCx would increase the disparity in WSS between medial and lateral walls. In other words, a larger bifurcation angle, with no change in the amount of branch flow, increases the likelihood of low and oscillating WSS at the lateral walls [96]. This has been confirmed by Chaichana and colleagues who showed a direct relationship between angulations of the left coronary artery and corresponding haemodynamic changes, based on simulated and realistic coronary models [13]. Low WSS and WSSG was found in the larger bifurcation angles ranging from 75◦ to 120◦ compared to the smaller bifurcation models ranging from 15◦ to 60◦ (Fig. 5). In addition to the CFD analysis of main coronary arteries, impact of side branches on local wall shear stress should not be neglected. Wellnhofer et al. studied the impact of side branches on WSS calculation in 17 patients and they concluded that side branches showed significant impact on coronary flow and WSS profile in the right coronary artery (Fig. 6) [96]. Chaichana et al. investigated the influence of realistic coronary plaques on the left coronary side branches, based on coronary CT angiography images in a patient with coronary artery stenosis at the left coronary bifurcation. Coronary plaques were found to be closely related to the subsequent WSS and wall pressure stress gradient changes in the coronary side branches [97]. These research findings improve the understanding of the development of atherosclerosis by exploring the haemodynamic effect of coronary plaques using CFD technique, although further studies based on a large cohort are required to verify these results. In summary, promising results have been achieved with use of CFD in patient-specific models for diagnosis of coronary artery disease. With the rapid developments of CFD methods and image processing techniques, the role of haemodynamic forces in the development of coronary artery disease will be elucidated with patient-specific CFD applications becoming more popular
Please cite this article in press as: Sun Z, Xu L. Computational fluid dynamics in coronary artery disease. Comput Med Imaging Graph (2014), http://dx.doi.org/10.1016/j.compmedimag.2014.09.002
G Model CMIG-1305; No. of Pages 13
ARTICLE IN PRESS Z. Sun, L. Xu / Computerized Medical Imaging and Graphics xxx (2014) xxx–xxx
5
4.3. Haemodynamic analysis of coronary plaques-biomechanics of atherosclerotic plaques
Fig. 5. Wall shear stress gradient observed with different angles of the realistic left coronary artery models generated from coronary computed tomography angiography at peak systolic phase of 0.4 s. The arrows indicate the wall shear stress gradient distributions, with a large region of the low magnitude present at a 120◦ model (A) and a small region at a 73◦ model (B). Reprint with permission from Ref. [13].
[94–102]. In the following sections, a review of the CFD applications in coronary plaques is provided with regard to the plaque biomechanics, plaque development, progression and rupture. A critical appraisal is given to the clinical value of a recently developed CFD-computed fractional flow reserve in the diagnosis of ischaemia-causing lesions.
It is generally accepted that in the presence of systematic risk factors, WSS plays a key role in the initiation of early atherosclerosis [103]. Low WSS is a well-recognised mechanical stimulus that promotes the inflammatory process by inducing an oxidative response in endothelial vascular cells and leads to endothelial dysfunction inducing a pro-atherogenic environment [37]. Thus, shear stress is regarded as an important factor leading to the focal distribution of atherosclerotic plaques, with extensive studies being conducted to investigate the relationship between shear stress and early phase of atherosclerosis [16,35,104]. Early studies were performed to understand associations between these elements with vessel pathology using CFD by modelling the formation and progression of atherosclerotic plaque [105,106]. In particular, the analysis of blood velocity and WSS, associated with artery geometry, has been recognised to provide early biomarkers of the atherosclerotic plaque formation and growth [107–109]. Later reports confirmed these findings regarding the role of wall shear stress in atherosclerosis development. Wahle et al. studied the correlation between WSS and plaque distribution in a set of 48 in vivo vessel segments based on a 3D reconstruction of intravascular ultrasound images [109]. The inverse relationship between local WSS and plaque thickness was significantly more pronounced in vessel cross sections demonstrating compensatory enlargement (positive remodelling) without lumen narrowing than when the full spectrum of disease severity was considered. Their findings confirmed that relatively lower WSS is associated with early plaque development. Similar studies have been performed using other imaging modalities to further verify the correlation between WSS and plaque development, such as extraction of vessel centreline from coronary CT angiography [110], investigation of relationship between WSS, plaque thickness and plaque burden by means of MRI images [111], reconstruction of side branch with coronary CT angiography data for investigation of 3D shear stress distribution in coronary bifurcations [112], and generation of 4D coronary arterial tree reconstructed from angiographic
Fig. 6. Three representative cases of transient simulations: oscillatory shear index (OSI) and wall shear stress gradient (WSSG) distributions. Reprint with permission from Ref. [96].
Please cite this article in press as: Sun Z, Xu L. Computational fluid dynamics in coronary artery disease. Comput Med Imaging Graph (2014), http://dx.doi.org/10.1016/j.compmedimag.2014.09.002
G Model CMIG-1305; No. of Pages 13 6
ARTICLE IN PRESS Z. Sun, L. Xu / Computerized Medical Imaging and Graphics xxx (2014) xxx–xxx
images [113]. These techniques may require further optimisation in order to be incorporated into an automatic processing system. In summary, these studies confirmed the role of biomechanics, in particular, WSS in the initiation of atherosclerotic plaques. Some early atherosclerotic plaques may progress either to non-stenotic high-risk rupture–prone plaques, which develop in areas characterised by low shear stress with excessive expansive remodelling, or to stenotic high-risk plaques. Thus, it is of paramount importance to explore the haemodynamics of coronary plaques. 4.4. Haemodynamic analysis of coronary plaques-prediction of plaque progression and vulnerability One of the main purposes of simulation studies is to identify the mechanical factors responsible for the plaque progression and/or rupture. Stone et al. reported data on plaque progression in six human coronary arteries over a 6-month period using wall thickness data obtained from intravascular ultrasound [114]. They found plaque progression and outward remodelling in the low shear stress patches, and in these patches shear stress increased after 6 months. The most recent study on this topic conducted by Samady et al. extended the framework proposed by other researchers to introduce plaque type analysis [60]. Authors studied the relation between WSS and the plaque progression in terms of type and amount of plaque in 20 patients who were monitored both on baseline and follow-up exams by intravascular ultrasound (virtual histology-intravascular ultrasound, VH-IVUS) and CFD modelling for WSS analysis. Over a period of 6-month follow-up, coronary segments with low WSS showed an increase in plaque area (0.12 ± 0.78 mm2 ) compared with those segments with intermediate WSS (−0.09 ± 1.16 mm2 ; p = 0.027) and high WSS (−0.16 ± 1.17 mm2 ; p = 0.002), both of which demonstrated plaque area regression. VH-IVUS-derived plaque composition indicated that coronary segments with low and high WSS were associated with significant increase in necrotic core area compared with segments with intermediate WSS (0.15 ± 0.26 and 0.09 ± 0.51 versus −0.03 ± 0.44 mm2 ; p < 0.001). VH-IVUS-derived plaque composition showed significant regression of fibrous area (−0.21 ± 0.83 versus 0.14 ± 0.56 mm2 ; p < 0.001) and fibro-fatty area (−0.14 ± 0.44 versus 0.01 ± 0.09 mm2 ; p < 0.001) and progression of dense calcium area (0.08 ± 0.34 versus 0.03 ± 0.09 mm2 ; p = 0.03) in the coronary segments with high WSS compared to those with low WSS. Furthermore, coronary segments with low WSS were found more likely to be associated with constrictive remodelling compared with those with intermediate (73% versus 42%; p = 0.15) and high WSS (73% versus 30%; p = 0.06), whereas high-WSS segments more likely underwent excessive expansive remodelling when compared with low-WSS segments (42% versus 15%; p = 0.16). These findings observed in low- and high-WSS segments suggest transformation of coronary plaque to a more vulnerable type. By far the most extensive follow-up study on shear stress in human coronary arteries was published by Stone et al. the Prediction of Progression of Coronary Artery Disease and Clinical Outcomes Using Vascular Profiling of Endothelial Shear Stress and Arterial Plaque Characteristics (PREDICTION) study [115]. The PREDICTION study is the largest and most comprehensive serial anatomic natural history study of coronary atherosclerosis with use of innovative methodologies to investigate potential pathophysiological mechanisms responsible for coronary artery disease progression. This study focused on the change in plaque thickness over a period of 6–10 months in 329 patients with high-risk coronary artery disease consisting of 824 coronary arteries. They observed that low shear stress was associated with a decrease in lumen area, while high shear stress segments showed an increase in lumen area. The positive predictive value of using baseline
vascular characteristics to identify a lesion that progresses clinically at follow-up (such as symptoms worsening, development of acute coronary syndrome due to lumen narrowing) increased from 22% to 41%, if only large plaque burden was presented when compared to the presence of both large plaque burden and low wall shear stress. The negative predictive value to identify a high-risk lesion was 92% if both large plaque burden and low shear stress were absent. This implies that plaque burden, defined as the ratio of plaque area and vessel area, and low local shear stress provide independent and additive prediction to identify plaques that develop progressive enlargement and lumen narrowing. Despite the PREDICTION study having both intravascular ultrasound and WSS data, no plaque phenotype data was included in their analysis. This limitation has been addressed by Corban et al. in their recent study who explored the incremental value of pathological intimal thickening, large plaque burden and high WSS for prediction of plaque progression and vulnerability [116]. Authors studied 20 patients with coronary artery disease who underwent baseline and 6-month follow-up for coronary virtual histologyintravascular ultrasound and CFD modelling for calculation of WSS. Among the three plaque phenotypes, coronary segments with pathological intimal thickening was found to lead to the greatest increase in necrotic core area (+0.31 ± 0.45 mm2 , p = 0.06) and greatest decrease in fibro-fatty area (−0.37 ± 0.48 mm2 , p < 0.0001). Their results showed that coronary segments with a combination of pathological intimal thickening, large plaque burden (>60%) and high WSS at baseline is associated with a higher rate of transformation to vulnerable plaque at follow-up when compared to the coronary segments with one or two of these variables. In summary, these studies emphasise the incremental value of regional WSS in combination with anatomic and composition characteristics of coronary plaques in predicting plaque progression and vulnerability. These findings warrant further investigation in large prospective trials, in particular, analysis of symptomatic clinical events between baseline vascular characteristics and symptomatic clinical outcome events as current studies lack such statistical inferences [60,115,116]. 4.5. Haemodynamic analysis of coronary plaques-correlation with plaque composition and location The distribution of ruptured or prone-to-rupture plaques is known to be non-uniform throughout the coronary tree [117,118]. According to several studies based on angiographic or intravascular ultrasound analyses, plaque rupture rarely occurs in the left main stem and the distal part of the coronary arteries, whereas it is far more common in the proximal part of the coronary arteries, especially in the left anterior descending artery [119–122]. Why vulnerable or ruptured plaques tend to spare the left main stem and distal segments of the left coronary arteries remain poorly understood. Plaque composition, favouring propensity to vulnerability, might also be non-uniformally distributed along the coronary arteries [123,124]. Nakazawa et al. studied histological data from 26 human coronary bifurcations and they observed the largest necrotic core in the proximal part of the plaque at the hips of coronary bifurcations, while the size of the necrotic core decreased gradually when going from proximal to distal [125]. Maehara et al. analysed the location of 300 plaque ruptures in culprit and non-culprit coronary arteries and they reported that the rupture location was upstream of the minimum lumen area [126]. Similarly, Fujii et al. confirmed these results in their study of plaque ruptures in 74 patients by showing that 80% of the ruptures were found upstream of the minimum lumen area in patients suffering from acute myocardial infarction [127]. The study by Chaichana et al. based on idealised and realistic coronary artery models showed a direct correlation between the
Please cite this article in press as: Sun Z, Xu L. Computational fluid dynamics in coronary artery disease. Comput Med Imaging Graph (2014), http://dx.doi.org/10.1016/j.compmedimag.2014.09.002
G Model CMIG-1305; No. of Pages 13
ARTICLE IN PRESS Z. Sun, L. Xu / Computerized Medical Imaging and Graphics xxx (2014) xxx–xxx
7
Fig. 7. The pressure gradient patterns inside left bifurcation at effective plaque locations with eight types of bifurcation plaques and normal condition during the diastolic phase (0.7 s). As shown in figure, the largest distribution of high pressure gradient is the type D which involves multiple coronary branches with significant lumen stenosis. Reprint with permission from Ref. [128].
low shear stress and left coronary bifurcation regions [13]. These studies confirm that at bifurcation level lower values of shear stress are present in regions opposite to flow divider, and these regions correspond to the sites mainly predisposed to the development of atherosclerosis [78]. Coronary plaque generally originates in the bifurcation region due to the angulations. The angulations cause a region of low wall shear stress, as confirmed by previous reports [13,99–101,121,122]. CFD provides an opportunity to predict the haemodynamic behaviour. Thus, the characterisation of haemodynamic variations due to the various types of bifurcation plaque in the configurations can be further explored with flow visualisations; this exceeds the traditional anatomical analysis of coronary stenosis or occlusion. Chaichana et al. studied various types of plaques in different anatomical positions of the left coronary artery to represent realistic distribution of coronary plaques [128]. The WSS, velocity and wall pressure gradient were computed and compared using the CFD method. These findings indicate that extra plaques located in the left coronary artery may increase the risk of plaque rupture (Fig. 7), although further studies are needed to analyse the realistic plaque at the coronary artery based on different configurations (concentric versus eccentric plaques) and compositions (calcified versus non-calcified or mixed plaques).
In summary, clinical and biomechanical studies have identified plaque composition, location and morphology as key predictors of plaque formation, vulnerability and likelihood of rupture [129,130]. However, identifying lesions vulnerable to rupture and characterising them remains a major issue for the prevention of acute cardiac events. A multimodality approach comprising coronary image postprocessing, computer modelling and simulation, and clinical indices of patient outcomes will improve risk assessment of coronary plaques, in particular, accurate identification of vulnerable plaques. 5. CFD-computed fractional flow reserve The fractional flow reserve (FFR) technique uses a pressure sensitive catheter to assess the rate of maximal myocardial blood flow through a stenotic artery relative to the flow through a normal portion of the aorta in a hyperemic state [131–133]. The FFR is a lesion-specific technique and is widely considered the gold standard physiologic test for assessment of the functional significance of a coronary stenosis. The FFR is considered diagnostic of ischaemia at values
ARTICLE IN PRESS
CMIG-1305; No. of Pages 13
Computerized Medical Imaging and Graphics xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Computerized Medical Imaging and Graphics journal homepage: www.elsevier.com/locate/compmedimag
Computational fluid dynamics in coronary artery disease Zhonghua Sun a , Lei Xu b,∗ a b
Discipline of Medical Imaging, Department of Imaging and Applied Physics, Curtin University, Perth, Western Australia 6845, Australia Department of Radiology, Beijing Anzhen Hospital, Capital Medical University, Beijing, China
a r t i c l e
i n f o
Article history: Received 5 July 2014 Received in revised form 22 August 2014 Accepted 3 September 2014 Keywords: Atherosclerosis Coronary artery disease Computational fluid dynamics Haemodynamics Plaque
a b s t r a c t Computational fluid dynamics (CFD) is a widely used method in mechanical engineering to solve complex problems by analysing fluid flow, heat transfer, and associated phenomena by using computer simulations. In recent years, CFD has been increasingly used in biomedical research of coronary artery disease because of its high performance hardware and software. CFD techniques have been applied to study cardiovascular haemodynamics through simulation tools to predict the behaviour of circulatory blood flow in the human body. CFD simulation based on 3D luminal reconstructions can be used to analyse the local flow fields and flow profiling due to changes of coronary artery geometry, thus, identifying risk factors for development and progression of coronary artery disease. This review aims to provide an overview of the CFD applications in coronary artery disease, including biomechanics of atherosclerotic plaques, plaque progression and rupture; regional haemodynamics relative to plaque location and composition. A critical appraisal is given to a more recently developed application, fractional flow reserve based on CFD computation with regard to its diagnostic accuracy in the detection of haemodynamically significant coronary artery disease. © 2014 Elsevier Ltd. All rights reserved.
1. Introduction Coronary artery disease is the leading cause of morbidity and mortality in advanced countries and its prevalence is increasing in developing countries. Coronary artery disease is responsible for 7.3 million deaths and 58 million disability-adjusted life years lost worldwide [1]. Despite its large socio-economic impact, the underlying mechanisms of coronary artery disease are only partially understood. It is generally accepted that coronary artery disease is an inflammatory disease with lipid deposition in the arterial wall as an initial stage of atherosclerosis [2–5]. Although the risk factors for atherosclerotic coronary plaque formation, including high cholesterol, diabetes, and hypertension are systemic in nature, plaques are located at specific sites in the coronary artery where disturbed flow and low endothelial shear stress occur (Fig. 1) [6–9]. In recent years, blood flow/shear stress has gained a lot of interest as complementary explanation for plaque formation [10,11]. The role of blood flow in the development of atherosclerosis is based on the observation that vascular inflammation and plaques are distributed near side branches or arterials stenosis, where blood
∗ Corresponding author at: Department of Radiology, Beijing Anzhen Hospital, Capital Medical University, No. 2, Anzhen Road, Chaoyang District, 100029 Beijing, China. Tel.: +86 10 64456071; fax: +86 10 64456962. E-mail addresses: [email protected] (Z. Sun), [email protected] (L. Xu).
flow is non-uniform, and at the lesser curvature of bends where blood flow rate is relatively low [12,13]. The effect of blood flow on the vessel wall is through shear stress, which influences the behaviour of endothelial cells including morphological adaptations and physiological changes of endothelial cells. Shear stress induces a shearing deformation of the endothelial cells which affects the phenotype of the endothelial cells and therefore the inflammatory component and plaque progression [14]. Low and oscillatory shear stress represent major features of the haemodynamic environment of regions opposite to arterial flow dividers that are predisposed to the formation of atherosclerosis (Fig. 2) [15–19]. It has been shown in vitro and in vivo studies that disturbed or oscillatory flows near arterial bifurcation, branch ostia and curvatures are associated with atheroma formation and intimal wall thickening [20–25]. The methods for in vivo estimation of wall shear stress can be performed by acquiring three-dimensional (3D) reconstruction of vessel volume using either invasive modalities such as intravascular ultrasound and invasive coronary angiography, or less-invasive techniques including coronary computed tomography angiography and cardiac magnetic resonance angiography with application of numerical methods to calculate flow within the reconstructed arterial volume for solving fluid dynamics [26,27]. The numerical methods are known as computational fluid dynamics (CFD) techniques. 3D reconstruction of coronary artery tree with subsequent numerical simulation using CFD techniques based on individual
http://dx.doi.org/10.1016/j.compmedimag.2014.09.002 0895-6111/© 2014 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Sun Z, Xu L. Computational fluid dynamics in coronary artery disease. Comput Med Imaging Graph (2014), http://dx.doi.org/10.1016/j.compmedimag.2014.09.002
G Model CMIG-1305; No. of Pages 13
ARTICLE IN PRESS Z. Sun, L. Xu / Computerized Medical Imaging and Graphics xxx (2014) xxx–xxx
2
Fig. 1. Molecular and cellular mechanisms through which low ESS promotes atherosclerosis. Data are derived from in vitro and animal in vivo experiments by applying a great variety of ESS patterns, both in direction and in magnitude. ESS, endothelial shear stress; BMP, bone morphogenic protein; eNOS, endothelial nitric oxide syntase; ET, endothelin; ICAM-1, intercellular adhesion molecule 1; IFN-g, interferon-g; IL, interleukin; LDL, low-density lipoprotein cholesterol; MCP, monocyte chemoattractant protein; MMP, matrix metalloproteinase; NF kB, nuclear factor-kB; NO, nitric oxide; PDGF, platelet-derived growth factor; ROS, reactive oxygen species; SREBP, sterol regulatory elements binding protein; TF, transcription factor; TGF-b, transforming growth factor b; TNF-a, tumour necrosis factor-a; t-PA, tissue plasminogen activator; VCAM, vascular cell adhesion molecule; VEGF, vascular endothelial growth factor; VSMC, vascular smooth muscle cell. Reprint with permission from Ref. [9].
patient data is being increasingly used to study haemodynamics and in predicting the behaviour of circulatory blood flow inside the coronary arteries [28–30]. This review article provides an overview of the applications of CFD in the diagnosis of coronary artery disease. The main focus of this review is to present an evidence-based review of the CFD in the biomechanics of atherosclerotic plaques and in the detection of high-risk coronary plaques and plaque progression with the aim of identifying high-risk patients so as to achieve the goal of reducing cardiac mortality. 2. Computational fluid dynamics CFD is a general term of all numerical techniques that are used to describe and analyse the flow of fluid elements at each location in certain geometry. The merit of CFD is developing new and improved devices and system designs, and optimisation is conducted on existing equipment through computational simulations resulting in enhanced efficiency and lower operating costs [31]. The governing equations of fluid dynamics can be computed to obtain coronary flow and pressure. These equations are called the
Navier–Stokes equations and have been known for more than 150 years. In order to simulate realistic coronary blood flow, a domain of interest must be defined, and boundary conditions specified. The isolation and generation of boundary condition is one of the greatest challenges in the integration of CFD for the assessment of the physiologic significance of coronary artery disease [32]. The coronary arterial wall is constantly exposed to both flowinduced wall shear stress (WSS) and arterial strain (relative displacement of wall within coronary artery) by blood pressure, myocardial contraction and local biological environment. Haemodynamic parameters of WSS, such as average wall shear stress (AWSS), average wall shear stress gradient (AWSSG), oscillatory shear index (OSI) and relative residence time (RRT) are possible indicators for atherosclerotic plaque prone sites [33–36]. Of these parameters, low WSS or OSI is a well-described mechanical stimulus that promotes the inflammatory process by inducing an oxidative response in endothelial vascular cells [37]. Lowering WSS has been reported to induce structural responses, according to several studies performed on cultured endothelial cells [38–40]. Thus, vascular remodelling is a response to alterations in WSS and
Fig. 2. Differential distribution of shear stress in a straight arterial segment proximal to a lumen-protruding atherosclerotic plaque. Reprint with permission from Ref. [16].
Please cite this article in press as: Sun Z, Xu L. Computational fluid dynamics in coronary artery disease. Comput Med Imaging Graph (2014), http://dx.doi.org/10.1016/j.compmedimag.2014.09.002
G Model CMIG-1305; No. of Pages 13
ARTICLE IN PRESS Z. Sun, L. Xu / Computerized Medical Imaging and Graphics xxx (2014) xxx–xxx
3
Fig. 3. Contour maps of maximal principal stresses. Comparison between a positive morphology model and the negative morphology models with concentric and eccentric plaques. The three cases presented have the same general dimensions: 75% lipid core, 46.6% stenosis ratio and 0.03 mm of fibrous cap thickness. Reprint with permission from Ref. [65].
other mechanical factors. This has led to the increasing applications of CFD in the investigation of pathogenesis of coronary artery disease, in particular, the identification and detection of vulnerable coronary plaques.
3. Definition of vulnerable or high-risk atherosclerotic plaque The vulnerable atherosclerotic plaque is called a “high-risk” or “thrombosis-prone” plaque. Major criteria to characterise such plaques include the presence of active inflammation (monocyte, macrophage or T-cell infiltration), a thin inflamed fibrous cap (40% of the total volume of the plaque), the presence of endothelial denudation with superficial platelet aggregation and the presence of haemodynamically significant stenosis (>90%) [41]. Minor criteria include the presence of spotty calcification, intraplaque haemorrhage, endothelial dysfunction and expansive remodelling [42,43]. A more clinical relevant definition of a vulnerable plaque is a lesion that places a patient at risk for developing future major adverse cardiac events, including death, myocardial infarction, or progressive angina. The identification of such plaques before they become symptomatic would enable prognostic stratification and facilitate primary prevention (e.g., aspirin, statins, and risk factor modification). Large lipid core and calcified areas (defined as >10% of the plaque area) and thin-cap fibroatheroma have been found to be associated with positive vascular remodelling [44–46]. Clinical and biomechanical studies have explored the risk factors associated with plaque vulnerability and have identified plaque composition and morphology as key determining factors for plaque vulnerability and likelihood of rupture [47–53]. Plaque characteristics can be determined by numerical simulation for evaluation of the vulnerability. WSS is one of the important physical factors in the process of atherosclerosis. Ex vivo studies have shown the interactions between shear stress distribution and changes in vascular anatomy and have demonstrated that high shear stress may be related to plaque rupture [54]. In vivo studies based on realistic patient’s data have confirmed these findings [52,53,55]. Abnormal WSS in geometrically susceptible coronary segments is thought to promote the development of atherosclerosis. Early plaque progression within a low WSS segment could result in gradual luminal narrowing that eventually leads to an increase in WSS into the physiologic range. As a plaque continues to encroach into the lumen, WSS is thought to increase into the high range [56]. The resultant high WSS has been demonstrated to induce apoptosis of smooth-muscle cells, increase matrix metalloproteinases which induce thinning of the plaque fibrous cap making it more vulnerable to fissuring or rupture [41,57–59]. Clinical studies support the
evidence that high WSS is associated with plaque vulnerability, plaque rupture and expansive vascular remodelling [60,61]. Vascular remodelling was first described in human autopsy studies as a compensatory change in vessel size as a result of increasing plaque burden and changes in local WSS in order to preserve the cross-sectional luminal area. It was reported that the development of luminal stenosis was delayed until the coronary plaque began to occupy 40% of the internal elastic lamina area [62]. This was confirmed by an in vivo study comparing intravascular ultrasound with clinical presentations with results showing that positive remodelling and larger plaque areas were associated with an unstable clinical presentation, while negative remodelling with a stable presentation [63]. Ohayon et al. studied the role of remodelling index on plaque stresses by considering positive and negative vascular remodelling as a parameter to affect plaque stability [64]. The authors investigated the biomechanical interaction between vessel and plaque geometry and the risk of plaque rupture. Their results showed that in the early stages of positive remodelling, lesions were more prone to rupture. This is supported by a recent study by Cilla et al. who examined the effect of positive and negative modelling on plaque stability by taking into account the arterial plaque morphologies [65]. Different stages and variations in atherosclerotic plaque remodelling were simulated with comparison of important determining parameters on plaque stresses including the fibrous cap thickness, the stenosis ratio, and lipid core dimensions and the atheroma plaque distribution. Their results support the findings by Ohayon et al. that coronary arteries with positive remodelling are more prone to rupture than those with negative remodelling for comparable geometrical conditions (Fig. 3).
4. CFD applications in coronary artery disease 4.1. Parameters used in CFD simulation For generation of coronary artery model, both fluid and structural domains are meshed with hexahedral cells so as to minimise numerical diffusion and lower the number of elements. Reliable blood flow data in the normal coronary arteries are required for conducting a fluid dynamics study, and this can be obtained with various imaging techniques, such as Doppler ultrasound or phasecontrast magnetic resonance imaging (MRI) [60,66]. The coronary artery flow pattern is significantly different from that in the aorta or peripheral arterial circulation, since blood supply to the heart occurs during diastole (Fig. 4). Flow in the coronary artery is unsteady as blood flow varies with the cardiac cycle. Most of the studies apply the cardiac cycle at the main aorta to provide both the inflow and outflow conditions to observe haemodynamic changes in the arterial system including coronary artery [67–72].
Please cite this article in press as: Sun Z, Xu L. Computational fluid dynamics in coronary artery disease. Comput Med Imaging Graph (2014), http://dx.doi.org/10.1016/j.compmedimag.2014.09.002
G Model CMIG-1305; No. of Pages 13 4
ARTICLE IN PRESS Z. Sun, L. Xu / Computerized Medical Imaging and Graphics xxx (2014) xxx–xxx
Fig. 4. Pulsatile blood flow waves used in the computational fluid dynamics simulation in the left coronary artery. T stands for one cardiac cycle; LCA, left coronary artery.
The inlet and outlet boundary conditions are based on a physiological pulsatile flow rate and pressure at the aorta [73]. As coronary artery disease most commonly occurs in the left coronary artery due to formation of bifurcation angle by two main coronary branches, left anterior descending (LAD) and left circumflex (LCx), a periodical pulsatile volume flow profile is applied as a velocity inlet boundary condition in the left coronary artery. Based on the average physiological human coronary flow data that are widely available in the literature [74], the average blood flow into the left coronary artery is reported to be 57 ml/min reaching a maximum value of 105 ml/min during the diastolic phases [75]. The blood flow distribution in the left bifurcation is calculated as 71% directed through the LAD and 29% through LCx according to previous reports [75,76]. Pulsatile aortic pressure is applied as an inlet boundary condition at the entrance of left main coronary artery, while pulsatile velocity conditions are imposed on both the LAD and LCx outlet boundaries (Fig. 4). A flow simulation is conducted over a time-span of several cardiac cycles, which are represented by time steps. The time steps can be divided into a number of coupling iterations, with each time step converged to a residual target of less than 1 × 10−4 by approximately 100 iterations [13,17,77]. A total of 600 time steps are required to achieve satisfactory convergence for fluid simulation when all velocity component changes from iteration to iteration are less than 10−6 [77,78]. The fluid–structure interaction (FSI) is an approach that can simultaneously model blood flow and arterial wall deformations, and it has received increasing interest in recent years due to its potential impact in the medical field [18,79–82]. Studies have shown the effect of vessel compliance on flow patterns and differences in WSS produced by FSI and rigid coronary models [83,84]. Dong et al. in their recent study based on idealised and realistic coronary models have demonstrated significant differences between FSI and rigid models, which are represented by noticeable qualitative discrepancies in WSS distributions over the bifurcation apex and the moderate narrow lumen site downstream to the LAD branch, and apparent quantitative differences in WSS profiles at the bifurcation apex over the diastole phase, respectively [76]. Furthermore, specific geometric features of coronary artery tree such as tortuosity or kinking would lead to flow alteration, thus, affecting the distribution of WSS [85–87]. These results further confirm the anatomical regions in the coronary artery which are prone to develop atherosclerotic lesions. 4.2. Clinical applications of CFD in coronary artery disease Mechanical forces and intravascular haemodynamics can affect and regulate blood vessels structure which induces a chronic
inflammatory response in the arterial walls resulting in atherosclerosis [6,88]. Applying CFD to study the blood flow in coronary artery disease is challenging as the coronary arteries are highly curved, movable and deformable during the heart beating. However, in recent years considerable insight has been gained on CFD studies in coronary artery disease, which is represented in two areas: development of CFD techniques and tremendous progress in applying CFD method to elucidate the role of haemodynamics in coronary artery disease development and progression [89]. Early CFD-based haemodynamic studies were conducted to represent in vitro conditions within restrictive assumptions [90–93]. Later reports demonstrated that CFD methods have been coupled with medical imaging techniques to provide detailed haemodynamic information which cannot be obtained from cardiac imaging alone. These studies have shown that CFD simulations have the potential to enhance the data obtained from in vivo methods (computed tomography [CT] or MRI) by providing a complete characterisation of haemodynamic conditions under precisely controlled conditions [94–98]. Knight et al. performed an analysis of the haemodynamic parameters including AWSSG, WSS and OSI obtained through a CFD study on the right coronary arteries of 30 patients undergoing coronary CT angiography [95]. These parameters were correlated to each patient’s specific plaque profile with the aim of predicting the particular plaque location. Their results showed a statistically significant difference between AWSSG and OSI in sensitivity and positive predictive value for the identification of atherosclerotic plaque sites in the right coronary artery. These findings further strengthen the theory that low WSS is a contributor to the initiation of atherosclerosis. In contrast to the right coronary artery, there is considerable variability in left main/LAD and LCx geometries among patients. A larger value of the angle between LAD and LCx would increase the disparity in WSS between medial and lateral walls. In other words, a larger bifurcation angle, with no change in the amount of branch flow, increases the likelihood of low and oscillating WSS at the lateral walls [96]. This has been confirmed by Chaichana and colleagues who showed a direct relationship between angulations of the left coronary artery and corresponding haemodynamic changes, based on simulated and realistic coronary models [13]. Low WSS and WSSG was found in the larger bifurcation angles ranging from 75◦ to 120◦ compared to the smaller bifurcation models ranging from 15◦ to 60◦ (Fig. 5). In addition to the CFD analysis of main coronary arteries, impact of side branches on local wall shear stress should not be neglected. Wellnhofer et al. studied the impact of side branches on WSS calculation in 17 patients and they concluded that side branches showed significant impact on coronary flow and WSS profile in the right coronary artery (Fig. 6) [96]. Chaichana et al. investigated the influence of realistic coronary plaques on the left coronary side branches, based on coronary CT angiography images in a patient with coronary artery stenosis at the left coronary bifurcation. Coronary plaques were found to be closely related to the subsequent WSS and wall pressure stress gradient changes in the coronary side branches [97]. These research findings improve the understanding of the development of atherosclerosis by exploring the haemodynamic effect of coronary plaques using CFD technique, although further studies based on a large cohort are required to verify these results. In summary, promising results have been achieved with use of CFD in patient-specific models for diagnosis of coronary artery disease. With the rapid developments of CFD methods and image processing techniques, the role of haemodynamic forces in the development of coronary artery disease will be elucidated with patient-specific CFD applications becoming more popular
Please cite this article in press as: Sun Z, Xu L. Computational fluid dynamics in coronary artery disease. Comput Med Imaging Graph (2014), http://dx.doi.org/10.1016/j.compmedimag.2014.09.002
G Model CMIG-1305; No. of Pages 13
ARTICLE IN PRESS Z. Sun, L. Xu / Computerized Medical Imaging and Graphics xxx (2014) xxx–xxx
5
4.3. Haemodynamic analysis of coronary plaques-biomechanics of atherosclerotic plaques
Fig. 5. Wall shear stress gradient observed with different angles of the realistic left coronary artery models generated from coronary computed tomography angiography at peak systolic phase of 0.4 s. The arrows indicate the wall shear stress gradient distributions, with a large region of the low magnitude present at a 120◦ model (A) and a small region at a 73◦ model (B). Reprint with permission from Ref. [13].
[94–102]. In the following sections, a review of the CFD applications in coronary plaques is provided with regard to the plaque biomechanics, plaque development, progression and rupture. A critical appraisal is given to the clinical value of a recently developed CFD-computed fractional flow reserve in the diagnosis of ischaemia-causing lesions.
It is generally accepted that in the presence of systematic risk factors, WSS plays a key role in the initiation of early atherosclerosis [103]. Low WSS is a well-recognised mechanical stimulus that promotes the inflammatory process by inducing an oxidative response in endothelial vascular cells and leads to endothelial dysfunction inducing a pro-atherogenic environment [37]. Thus, shear stress is regarded as an important factor leading to the focal distribution of atherosclerotic plaques, with extensive studies being conducted to investigate the relationship between shear stress and early phase of atherosclerosis [16,35,104]. Early studies were performed to understand associations between these elements with vessel pathology using CFD by modelling the formation and progression of atherosclerotic plaque [105,106]. In particular, the analysis of blood velocity and WSS, associated with artery geometry, has been recognised to provide early biomarkers of the atherosclerotic plaque formation and growth [107–109]. Later reports confirmed these findings regarding the role of wall shear stress in atherosclerosis development. Wahle et al. studied the correlation between WSS and plaque distribution in a set of 48 in vivo vessel segments based on a 3D reconstruction of intravascular ultrasound images [109]. The inverse relationship between local WSS and plaque thickness was significantly more pronounced in vessel cross sections demonstrating compensatory enlargement (positive remodelling) without lumen narrowing than when the full spectrum of disease severity was considered. Their findings confirmed that relatively lower WSS is associated with early plaque development. Similar studies have been performed using other imaging modalities to further verify the correlation between WSS and plaque development, such as extraction of vessel centreline from coronary CT angiography [110], investigation of relationship between WSS, plaque thickness and plaque burden by means of MRI images [111], reconstruction of side branch with coronary CT angiography data for investigation of 3D shear stress distribution in coronary bifurcations [112], and generation of 4D coronary arterial tree reconstructed from angiographic
Fig. 6. Three representative cases of transient simulations: oscillatory shear index (OSI) and wall shear stress gradient (WSSG) distributions. Reprint with permission from Ref. [96].
Please cite this article in press as: Sun Z, Xu L. Computational fluid dynamics in coronary artery disease. Comput Med Imaging Graph (2014), http://dx.doi.org/10.1016/j.compmedimag.2014.09.002
G Model CMIG-1305; No. of Pages 13 6
ARTICLE IN PRESS Z. Sun, L. Xu / Computerized Medical Imaging and Graphics xxx (2014) xxx–xxx
images [113]. These techniques may require further optimisation in order to be incorporated into an automatic processing system. In summary, these studies confirmed the role of biomechanics, in particular, WSS in the initiation of atherosclerotic plaques. Some early atherosclerotic plaques may progress either to non-stenotic high-risk rupture–prone plaques, which develop in areas characterised by low shear stress with excessive expansive remodelling, or to stenotic high-risk plaques. Thus, it is of paramount importance to explore the haemodynamics of coronary plaques. 4.4. Haemodynamic analysis of coronary plaques-prediction of plaque progression and vulnerability One of the main purposes of simulation studies is to identify the mechanical factors responsible for the plaque progression and/or rupture. Stone et al. reported data on plaque progression in six human coronary arteries over a 6-month period using wall thickness data obtained from intravascular ultrasound [114]. They found plaque progression and outward remodelling in the low shear stress patches, and in these patches shear stress increased after 6 months. The most recent study on this topic conducted by Samady et al. extended the framework proposed by other researchers to introduce plaque type analysis [60]. Authors studied the relation between WSS and the plaque progression in terms of type and amount of plaque in 20 patients who were monitored both on baseline and follow-up exams by intravascular ultrasound (virtual histology-intravascular ultrasound, VH-IVUS) and CFD modelling for WSS analysis. Over a period of 6-month follow-up, coronary segments with low WSS showed an increase in plaque area (0.12 ± 0.78 mm2 ) compared with those segments with intermediate WSS (−0.09 ± 1.16 mm2 ; p = 0.027) and high WSS (−0.16 ± 1.17 mm2 ; p = 0.002), both of which demonstrated plaque area regression. VH-IVUS-derived plaque composition indicated that coronary segments with low and high WSS were associated with significant increase in necrotic core area compared with segments with intermediate WSS (0.15 ± 0.26 and 0.09 ± 0.51 versus −0.03 ± 0.44 mm2 ; p < 0.001). VH-IVUS-derived plaque composition showed significant regression of fibrous area (−0.21 ± 0.83 versus 0.14 ± 0.56 mm2 ; p < 0.001) and fibro-fatty area (−0.14 ± 0.44 versus 0.01 ± 0.09 mm2 ; p < 0.001) and progression of dense calcium area (0.08 ± 0.34 versus 0.03 ± 0.09 mm2 ; p = 0.03) in the coronary segments with high WSS compared to those with low WSS. Furthermore, coronary segments with low WSS were found more likely to be associated with constrictive remodelling compared with those with intermediate (73% versus 42%; p = 0.15) and high WSS (73% versus 30%; p = 0.06), whereas high-WSS segments more likely underwent excessive expansive remodelling when compared with low-WSS segments (42% versus 15%; p = 0.16). These findings observed in low- and high-WSS segments suggest transformation of coronary plaque to a more vulnerable type. By far the most extensive follow-up study on shear stress in human coronary arteries was published by Stone et al. the Prediction of Progression of Coronary Artery Disease and Clinical Outcomes Using Vascular Profiling of Endothelial Shear Stress and Arterial Plaque Characteristics (PREDICTION) study [115]. The PREDICTION study is the largest and most comprehensive serial anatomic natural history study of coronary atherosclerosis with use of innovative methodologies to investigate potential pathophysiological mechanisms responsible for coronary artery disease progression. This study focused on the change in plaque thickness over a period of 6–10 months in 329 patients with high-risk coronary artery disease consisting of 824 coronary arteries. They observed that low shear stress was associated with a decrease in lumen area, while high shear stress segments showed an increase in lumen area. The positive predictive value of using baseline
vascular characteristics to identify a lesion that progresses clinically at follow-up (such as symptoms worsening, development of acute coronary syndrome due to lumen narrowing) increased from 22% to 41%, if only large plaque burden was presented when compared to the presence of both large plaque burden and low wall shear stress. The negative predictive value to identify a high-risk lesion was 92% if both large plaque burden and low shear stress were absent. This implies that plaque burden, defined as the ratio of plaque area and vessel area, and low local shear stress provide independent and additive prediction to identify plaques that develop progressive enlargement and lumen narrowing. Despite the PREDICTION study having both intravascular ultrasound and WSS data, no plaque phenotype data was included in their analysis. This limitation has been addressed by Corban et al. in their recent study who explored the incremental value of pathological intimal thickening, large plaque burden and high WSS for prediction of plaque progression and vulnerability [116]. Authors studied 20 patients with coronary artery disease who underwent baseline and 6-month follow-up for coronary virtual histologyintravascular ultrasound and CFD modelling for calculation of WSS. Among the three plaque phenotypes, coronary segments with pathological intimal thickening was found to lead to the greatest increase in necrotic core area (+0.31 ± 0.45 mm2 , p = 0.06) and greatest decrease in fibro-fatty area (−0.37 ± 0.48 mm2 , p < 0.0001). Their results showed that coronary segments with a combination of pathological intimal thickening, large plaque burden (>60%) and high WSS at baseline is associated with a higher rate of transformation to vulnerable plaque at follow-up when compared to the coronary segments with one or two of these variables. In summary, these studies emphasise the incremental value of regional WSS in combination with anatomic and composition characteristics of coronary plaques in predicting plaque progression and vulnerability. These findings warrant further investigation in large prospective trials, in particular, analysis of symptomatic clinical events between baseline vascular characteristics and symptomatic clinical outcome events as current studies lack such statistical inferences [60,115,116]. 4.5. Haemodynamic analysis of coronary plaques-correlation with plaque composition and location The distribution of ruptured or prone-to-rupture plaques is known to be non-uniform throughout the coronary tree [117,118]. According to several studies based on angiographic or intravascular ultrasound analyses, plaque rupture rarely occurs in the left main stem and the distal part of the coronary arteries, whereas it is far more common in the proximal part of the coronary arteries, especially in the left anterior descending artery [119–122]. Why vulnerable or ruptured plaques tend to spare the left main stem and distal segments of the left coronary arteries remain poorly understood. Plaque composition, favouring propensity to vulnerability, might also be non-uniformally distributed along the coronary arteries [123,124]. Nakazawa et al. studied histological data from 26 human coronary bifurcations and they observed the largest necrotic core in the proximal part of the plaque at the hips of coronary bifurcations, while the size of the necrotic core decreased gradually when going from proximal to distal [125]. Maehara et al. analysed the location of 300 plaque ruptures in culprit and non-culprit coronary arteries and they reported that the rupture location was upstream of the minimum lumen area [126]. Similarly, Fujii et al. confirmed these results in their study of plaque ruptures in 74 patients by showing that 80% of the ruptures were found upstream of the minimum lumen area in patients suffering from acute myocardial infarction [127]. The study by Chaichana et al. based on idealised and realistic coronary artery models showed a direct correlation between the
Please cite this article in press as: Sun Z, Xu L. Computational fluid dynamics in coronary artery disease. Comput Med Imaging Graph (2014), http://dx.doi.org/10.1016/j.compmedimag.2014.09.002
G Model CMIG-1305; No. of Pages 13
ARTICLE IN PRESS Z. Sun, L. Xu / Computerized Medical Imaging and Graphics xxx (2014) xxx–xxx
7
Fig. 7. The pressure gradient patterns inside left bifurcation at effective plaque locations with eight types of bifurcation plaques and normal condition during the diastolic phase (0.7 s). As shown in figure, the largest distribution of high pressure gradient is the type D which involves multiple coronary branches with significant lumen stenosis. Reprint with permission from Ref. [128].
low shear stress and left coronary bifurcation regions [13]. These studies confirm that at bifurcation level lower values of shear stress are present in regions opposite to flow divider, and these regions correspond to the sites mainly predisposed to the development of atherosclerosis [78]. Coronary plaque generally originates in the bifurcation region due to the angulations. The angulations cause a region of low wall shear stress, as confirmed by previous reports [13,99–101,121,122]. CFD provides an opportunity to predict the haemodynamic behaviour. Thus, the characterisation of haemodynamic variations due to the various types of bifurcation plaque in the configurations can be further explored with flow visualisations; this exceeds the traditional anatomical analysis of coronary stenosis or occlusion. Chaichana et al. studied various types of plaques in different anatomical positions of the left coronary artery to represent realistic distribution of coronary plaques [128]. The WSS, velocity and wall pressure gradient were computed and compared using the CFD method. These findings indicate that extra plaques located in the left coronary artery may increase the risk of plaque rupture (Fig. 7), although further studies are needed to analyse the realistic plaque at the coronary artery based on different configurations (concentric versus eccentric plaques) and compositions (calcified versus non-calcified or mixed plaques).
In summary, clinical and biomechanical studies have identified plaque composition, location and morphology as key predictors of plaque formation, vulnerability and likelihood of rupture [129,130]. However, identifying lesions vulnerable to rupture and characterising them remains a major issue for the prevention of acute cardiac events. A multimodality approach comprising coronary image postprocessing, computer modelling and simulation, and clinical indices of patient outcomes will improve risk assessment of coronary plaques, in particular, accurate identification of vulnerable plaques. 5. CFD-computed fractional flow reserve The fractional flow reserve (FFR) technique uses a pressure sensitive catheter to assess the rate of maximal myocardial blood flow through a stenotic artery relative to the flow through a normal portion of the aorta in a hyperemic state [131–133]. The FFR is a lesion-specific technique and is widely considered the gold standard physiologic test for assessment of the functional significance of a coronary stenosis. The FFR is considered diagnostic of ischaemia at values
