Annals of Biomedical Engineering, Vol. 43, No. 1, January 2015 ( 2014) pp. 94–106 DOI: 10.1007/s10439-014-1155-9

Focal Association Between Wall Shear Stress and Clinical Coronary Artery Disease Progression LUCAS H. TIMMINS,1,2 DAVID S. MOLONY,1 PARHAM ESHTEHARDI,3 MICHAEL C. MCDANIEL,2 JOHN N. OSHINSKI,1,4 HABIB SAMADY,2 and DON P. GIDDENS1 1 Wallace H. Coulter Department of Biomedical Engineering, Georgia Institute of Technology and Emory University, 313 Ferst Drive, Room 2127, Atlanta, GA 30332, USA; 2Division of Cardiology, Department of Medicine, Emory University School of Medicine, Atlanta, GA 30322, USA; 3Department of Medicine, Jacobi Medical Center, Albert Einstein College of Medicine, Bronx, NY 10461, USA; and 4Department of Radiology and Imaging Sciences, Emory University School of Medicine, Atlanta, GA 30322, USA

(Received 1 May 2014; accepted 7 October 2014; published online 15 October 2014) Associate Editor Diego Gallo oversaw the review of this article.

INTRODUCTION

Abstract—Wall shear stress (WSS) has been investigated as a potential prospective marker to identify rapidly progressing coronary artery disease (CAD) and potential for lesions to acquire vulnerable characteristics. Previous investigations, however, are limited by a lack of understanding of the focal association between WSS and CAD progression (i.e., data are notably spatially averaged). Thus, the aim of this investigation was to examine the focal association between WSS and coronary atherosclerosis progression, and compare these results to those determined by spatial averaging. Five patients with CAD underwent baseline and 6-month follow-up angiographic and virtual histology-intravascular ultrasound imaging to quantify CAD progression. Patient-specific computational fluid dynamics models were constructed to compute baseline WSS values, which were either averaged around the entire artery circumference or examined in focal regions (sectors). Analysis of data within each sector (n = 3871) indicated that circumferentially averaged and sector WSS values were statistically different (p < 0.05) and exhibited poor agreement (concordance correlation coefficient = 0.69). Furthermore, differences were observed between the analysis techniques when examining the association of WSS and CAD progression. This investigation highlights the importance of examining spatially heterogeneous variables at a focal level to reduce the affect of data reduction and warrants implementation in a larger clinical study to determine the predictive power in prospectively identifying rapidly progressing and/or vulnerable coronary plaques.

Coronary artery disease (CAD) is the leading cause of death in the United States, accounting for nearly 400,000 deaths annually.9 While cardiovascular disease risk factors are systemic, coronary plaque development is a focal, primarily eccentric, phenomenon with a predilection towards areas of high curvature and branching vessels.6,12,31 Once formed, however, each coronary plaque independently evolves (e.g., progresses, ruptures, stabilizes),33 which implies that local determinants are responsible for plaque evolution. Of significant clinical interests and importance are the atherosclerotic plaques that rapidly progress and/or become unstable by gaining vulnerable features (i.e., thin cap fibroatheroma). If these high-risk or so-called vulnerable plaques rupture, the resulting thrombotic event, which presents clinically as an acute coronary syndrome (e.g., acute myocardial infarction, unstable angina), can be fatal. Therefore, early detection of vulnerable plaques is of critical importance and may lead to the development of preemptive treatment strategies in reducing major adverse cardiac events. Near-wall blood flow patterns that induce low and oscillatory wall shear stress (WSS) have been positively correlated with the spatial distribution of CAD,7,11 and, most recently, have been proposed as a prognostic marker in identifying vulnerable plaques.35 The development of techniques to reconstruct the complex 3D geometry of the coronary arteries from clinical imaging data now allow for patient-specific computational fluid dynamics (CFD) models to quantify coronary WSS distributions.14,17,34 In combination with

Keywords—Hemodynamics, Atherosclerosis, Computational fluid dynamics, Biomechanics, Intravascular ultrasound, Virtual histology.

Address correspondence to Lucas H. Timmins, Wallace H. Coulter Department of Biomedical Engineering, Georgia Institute of Technology and Emory University, 313 Ferst Drive, Room 2127, Atlanta, GA 30332, USA. Electronic mail: lucas.timmins@bme. gatech.edu

94 0090-6964/15/0100-0094/0

 2014 Biomedical Engineering Society

Focal Examination of WSS and CAD Progression

intravascular imaging modalities (e.g., intravascular ultrasound (IVUS), optical coherence tomography (OCT)), which allow for evaluation of plaque morphology and composition, longitudinal clinical investigations can be conducted to understand the association of coronary atherosclerosis evolution and WSS. Previous clinical studies have shown that regions of low WSS are associated with plaque progression and expansive or constrictive remodeling at 6–8 months follow-up.27,28 Furthermore, the PREDICTION Study, a multicenter prospective that examined 506 patients with follow-up ranging from 6–10 months, concluded that low WSS provided independent predictive value in identifying CAD lesions that enlarge and cause lumen narrowing requiring percutaneous coronary intervention.30 With the aid of virtual histology-IVUS (VH-IVUS), an imaging technique that allows for the identification of plaque constituents (e.g., fibrous tissue, fibro-fatty tissue, necrotic core, and dense calcium),22 we have shown that at 6-months follow-up regions of low and high WSS exhibit significant total plaque area progression and regression, respectively.26 We also showed that regions of high WSS exhibit an increase in necrotic core area, implying that a phenotypic transformation characteristic of vulnerable plaque occurs in these coronary regions. Moreover, the combination of WSS, plaque burden, and plaque phenotype exhibited incremental value for prediction of plaque progression and vulnerability.4 Thus, there is considerable clinical evidence that, similar to CAD development, local hemodynamics modulate the evolution of CAD. Although previous studies have provided insight on the role of WSS in clinical CAD progression, they are limited by a lack of understanding of the focal association between these highly spatially heterogeneous variables. Specifically, WSS values are averaged over relatively large arterial segments27,28,30 or circumferentially averaged over cross-sections26 and correlated with plaque progression values that are an aggregate over these regions. It has been suggested that the effects of spatial averaging, or data reduction, can considerably limit the range of hemodynamic data or disease stages and lead to conclusions that contradict accepted principles.23 To investigate the effects of spatial averaging on the association between WSS and clinical CAD progression, the aim of this study was to compare the associations when data are either examined around the complete arterial circumference or in focal regions, herein circumferential or sector analysis, respectively. Furthermore, comparisons in computed hemodynamic values, plaque morphology, and plaque progression are examined. The results from this investigation will provide understanding on whether the refinement of analysis techniques offers greater

95

insight on the effects of WSS on the natural history of CAD.

MATERIALS AND METHODS Patient Data Collection Patients (n = 5) were randomly selected from a previously investigated larger cohort.26 These patients presented to the Cardiac Catheterization Laboratory at Emory University Hospital with stable angina or an abnormal non-invasive stress test and found to have a non-obstructive lesion requiring invasive physiologic evaluation were enrolled. Clinical data acquisition, physiologic examination, and enrollment criteria are described in considerable detail elsewhere.26 At baseline, all patients underwent biplane coronary angiography (Philips Medical Systems, Andover, MA) and EKG-gated (R-wave peak) VH-IVUS image acquisition of the proximal left anterior descending (LAD) and left main (LM) coronary arteries (20 MHz Eagle Eye Gold Catheter, Volcano Corp., Rancho Cordova, CA). Briefly, following intracoronary administration of 200 lg nitroglycerin, the IVUS catheter was guided approximately 60 mm down the LAD. An automated motorized pullback (0.5 mm/s) was performed, and VH-IVUS images were acquired up to the guide catheter in the aorta. Due to slight changes in a patient’s heart rate during the IVUS pullback, VHIVUS images do not have a perfectly uniform distribution along the coronary tree; however, these minimal variations have no affect on the resulting accuracy of the geometry reconstruction. In addition, Doppler derived velocity and pressure data were acquired in the LM and distal LAD coronary arteries with a 0.014¢¢ monitoring guidewire (ComboWire, Volcano Corp.). Following 6-months, during which time optimal medical therapy was prescribed (e.g., cardiovascular risk factor modification an atorvastatin), patients returned for repeat catheterization and VH-IVUS image acquisition of the same coronary segment. Offline volumetric reconstruction and analysis of the VHIVUS data, which included semi-automatic segmentation of the internal and external elastic laminae, were performed by a single experienced VH-IVUS reader who was blinded to the patients’ clinical data (echoPlaque 4.0, INDEC Medical Systems, Santa Clara, CA).4,5,26 Boundary segmentation was performed according to criteria of the American College of Cardiology Clinical Consensus documents on IVUS and good intraobserver reproducibility has been previously reported.21,26 Eligible patients provided written informed consent. The Emory University Institutional Review Board approved the study.

96

TIMMINS et al.

Coronary Geometry Reconstruction The end-diastolic three-dimensional (3D) geometry of the coronary segment of interest was constructed by fusion of angiographic and VH-IVUS imaging data (i.e., the ANGUS technique) as previously described and validated.17,34 Briefly, biplane angiographic images of the IVUS catheter were acquired in its most distal location prior to pullback (Figs. 1a–1d). The 3D spatial location of catheter was determined via backprojection imaging analysis techniques (IC-PRO, Paieon, Inc. Rosh Ha’ayin, Israel), and served as a backbone for vessel construction. The specific location of each segmented VH-IVUS image (Fig. 1e) on the catheter path was determined from its time-stamp and automatic catheter pullback speed (0.5 mm/s). At this location, images were stacked perpendicular to the catheter. As a result of catheter twisting during pullback, the relative rotation between IVUS images (i.e., the rotation of the catheter tip between consecutive images) was quantified by the sequential triangulation algorithm, which is adapted from the Frenet– Serret formulas from differential geometry.34 Collectively, this reconstruction technique resulted in an anatomically correct 3D reconstruction of the coronary segment interrogated by the IVUS catheter (Fig. 1f).

(a)

(b)

(c)

(d)

(e)

Branches

Computational Modeling Following main vessel reconstruction, coronary branches were extended perpendicular to the catheter path. Branch locations and diameters were determined from the VH-IVUS and angiographic imaging data, respectively. To ensure an accurate representation of the modeled system, all major branches identifiable on angiography and IVUS were included, such as the left circumflex, diagonal, and septal coronary arteries. Incorporation of branch data into the geometry reconstruction allowed for flow division at branches, as occurs in vivo, thus preventing incorrectly high flow rates, and WSS values, in distal vessels if branches were excluded. The resulting 3D lumen geometry point cloud was wrapped with a surface, and the surface was defined by non-uniform rational B-splines (Geomagic Studio 11, Geomagic, Inc., Research Triangle Park, NC). Prior to geometry discretization, flow extensions were added to ensure a smooth transition into the computational domain (1 diameter) and fully developed flow at all outlets (7 diameter; Fig. 2a; ICEM CFD, ANSYS 14, Ansys, Inc., Canonsburg, PA). The volume was discretized with an unstructured tetrahedral mesh (ICEM CFD), which included an 8 volume thick boundary layer (initial height = 0.01 mm, height ratio = 1.5, total height = 0.22 mm) to capture

(f)

Lumen

IVUS Catheter IVUS Images FIGURE 1. Example of an ANGUS derived coronary geometry reconstruction. Biplane angiographic images of left coronary tree following contrast injection (a, b) and segmented IVUS catheter prior to pullback (c, d). (e) Representative VHIVUS images acquired in the left anterior descending coronary artery. Note the identifiable coronary branches. (f) Complete reconstructed coronary geometry. For illustrative purposes, only every 10th VH-IVUS image is shown.

near-wall flow patterns, and imported into the commercial solver Fluent (ANSYS 14). Patient-specific pulsatile inlet velocity values were digitized from Doppler data (Fig. 2b) and temporally interpolated at

Focal Examination of WSS and CAD Progression

(a)

Computational Mesh

Main Body

(b)

97 Aortic Pressure Coronary Pressure Velocity

Flo Exte w nsio

n

(c)

(d)

TAWSS 2 (dynes/cm ) 70

Velocity (m/s)

60 0.6

50

0.5

40 30

0.4

20 0.3

10 0

0.2 0

0.18

0.35

0.53

0.71

0.88

Time (s) FIGURE 2. Computational modeling techniques. (a) Addition of flow extensions to reconstructed coronary segment. (b) In vivo Doppler derived velocity (cyan) and pressure (yellow) data in the left main coronary artery. (c) Digitized velocity data over 1 cardiac cycle applied at the inlet of the computational grid. (d) Resulting TAWSS distribution across computational model.

300 equally spaced time-points in one cardiac cycle (Fig. 2c). These velocity values were applied over 300 increments in the simulation as uniform velocity profiles (i.e., plug profile). All outlets were assumed pressurefree. A no-slip boundary condition (m = 0) was applied at the rigid wall. The fluid (blood) was assumed to be an incompressible Newtonian fluid, which is valid under the pulsatile, moderate Reynolds number flow conditions in the coronary arteries, with a density of 1.06 g/ cm3 and dynamic viscosity of 3.5 cP.15 The pressure– velocity equations were coupled with the SIMPLE scheme and second-order spatial discretization was used. Convergence criteria were set to residual errors