Physiological Flow Simulation in Residual Human Stenoses After Coronary Angioplasty

Rupak K. Banerjee Staff Scientist, Bioengineering and Physical Science Program, Bldg. 13, Rm. 3N17, National Institute of Health (NIH), Bethesda, MD 20892 Lloyd H. Back Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109 Fellow ASME Martin R. Back Assistant Professor of Surgery, Division of Vascular Surgery, University of South Florida, College of Medicine, Tampa, FL 33606 Young I. Cho Professor, Mechanical Engineering and Mechanics Department, Drexel University, Philadelphia, PA 19104 Physiological Flow Simulation in Residual Human Stenoses After Coronary Angioplasty To evaluate the local hemodynamic implications of coronary artery balloon angioplasty, computational fluid dynamics (CFD) was applied in a group of patients previously reported by [Wilson et al. (1988), 77, pp. 873 885] with representative stenosis geometry post-angioplasty and with measured values of coronary flow reserve returning to a normal range (3.6 0.3). During undisturbed flow in the absence of diagnostic catheter sensors within the lesions, the computed mean pressure drop p was only about 1 mmhg at basal flow, and increased moderately to about 8 mmhg for hyperemic flow. Corresponding elevated levels of mean wall shear stress in the midthroat region of the residual stenoses, which are common after angioplasty procedures, increased from about 60 to 290 dynes/cm 2 during hyperemia. The computations R e e 100 400; e 2.25 indicated that the pulsatile flow field was principally quasi-steady during the cardiac cycle, but there was phase lag in the pressure drop mean velocity p ū relation. Timeaveraged pressure drop values, p, were about 20 percent higher than calculated pressure drop values, p s, for steady flow, similar to previous in vitro measurements by Cho et al. (1983). In the throat region, viscous effects were confined to the near-wall region, and entrance effects were evident during the cardiac cycle. Proximal to the lesion, velocity profiles deviated from parabolic shape at lower velocities during the cardiac cycle. The flow field was very complex in the oscillatory separated flow reattachment region in the distal vessel where pressure recovery occurred. These results may also serve as a useful reference against catheter-measured pressure drops and velocity ratios (hemodynamic endpoints) and arteriographic (anatomic) endpoints post-angioplasty. Some comparisons to previous studies of flow through stenoses models are also shown for perspective purposes. S0148-0731 00 00304-6 Introduction Successful luminal recanalization of coronary or peripheral arterial occlusive lesions during endovascular procedures is commonly defined in terms of anatomic endpoints. Contrast angiography and newer alternative imaging techniques e.g., intravascular ultrasonography, angioscopy are routinely used to assess the adequacy of remodeling of stenoses by various angioplasty or other obliterative techniques. The hemodynamic implications of stenosis remodeling, are, however, more difficult to measure. Since the function of the perfused distal organ or tissue depends on restoration of near-normal flow conditions across stenotic lesions, hemodynamic endpoints are of interest and may serve predictive roles in maintaining short and long term luminal patency following angioplasty. Accurate quantitation of volumetric blood flow rates in stenotic vessels is difficult to determine clinically. While in principle the pulsed Doppler ultrasound tip catheter can be used to assess coronary artery velocity 1, absolute measurement of mean flow rate is complicated by uncertainties in velocity distributions across the lumen. Uncertainties in velocity may be caused due to lumen curvature or nearby branches, transducer positioning and sample volume, blockage and disturbance effects of the catheter, determination of vessel cross-sectional area, and in vivo device calibration. Pulsed Doppler catheters have been used in patients to measure changes in coronary artery velocity in response to infusion of vasodilator agents such as papaverine or adenosine and estimate coronary flow reserve ratio of maximal mean Doppler shift frequency between hyperemic and basal conditions before and after Contributed by the Bioengineering Division for publication in the JOURNAL OF BIOMECHANICAL ENGINEERING. Manuscript received by the Bioengineering Division March 23, 1999; revised manuscript received March 20, 2000. Associate Technical Editor: S. E. Rittgers. angioplasty of stenotic lesions e.g., 2. Small Doppler guide wire systems (d 0.46 mm are also susceptible to inaccuracies e.g., 3 and have been used distal as well as proximal to coronary lesions e.g., 4. Measurements of the mean pressure gradient across a stenosis have been used during angioplasty procedures as a hemodynamic endpoint in gaging the severity of the lesion and the effectiveness of the intervention. Pressure gradients often referred to in the medical literature are pressure drops, i.e., not divided by an axial length. Whereas clinical investigators have acknowledged the limitations of translesional pressure gradient measurements because of flow obstruction by catheters, limited information is available on what degrees of flow blockage exist with currently used catheters e.g., 2,5. Wilson and Laxson 6 note that the specific effects of changes in hemodynamic conditions on pressure gradient measurements are not well described in the human coronary circulation. Small diameter guide-wire optic sensors (d 0.46 mm; 7 and fluid-filled guide wires (d 0.38 mm; 8 hold promise for reducing flow blockage effects in measuring mean trans-stenotic pressure gradients. Earlier in vitro pressure measurements by Young and Tsai 9, Mates et al. 10, and Cho et al. 11, in particular, have contributed to an improved understanding of stenotic flows, as also have the earlier computational studies by Deshpande et al. 12 and later by Siegel et al. 13 for steady laminar flows through various stenoses models, by Back et al. 14 for pulsatile flow in a diffuse plaque region model obtained from a coronary artery casting, and later by Cho and Kensey 15 for blood flow with shear-ratedependent viscosity ( ( )) through a casting of a diseased coronary artery. Spatial resolution of numerical flow computations have improved significantly over the years, but at the expense of relatively long computational times for pulsatile blood flow through vessels. 310 Õ Vol. 122, AUGUST 2000 Copyright 2000 by ASME Transactions of the ASME

Fig. 1 Stenoses configuration with dimensions and finite element mesh This investigation utilizes computational fluid dynamics CFD to evaluate hemodynamics in human coronary stenoses remodeled by balloon angioplasty. The simulation is for physiological conditions without catheters present to disturb/alter the flow in the residual lesions. Banerjee et al. 16 calculated the pressure, in conjunction with the pressure measurements by Wilson et al. 2, using the angioplasty catheter, which significantly elevated the pressure drop and reduced blood flow in the tighter artifactual stenoses during the measurements. The in vivo data set of Wilson et al. 2 was used with proximal lesion measurements of coronary flow reserve in the normal range after angioplasty. The flow regime was basically laminar mean flow Reynolds numbers R e e 100 400) and the frequency parameter e 2.25. The results of the simulations at basal and hyperemic flow conditions give numerical data on pulsatile velocity profiles proximally, within the lesion, and distally; phasic and mean wall shear stress; pressure drop distributions and mean pressure drops; and phase relations. The results provide a wealth of numerical data on hemodynamic variables that cannot be measured in human coronary lesions, and are descriptive of the physiologic conditions that may have existed on average in the patient group of Wilson et al. 2 after the balloon procedures. The results also may be useful for comparison to measurements from endovascular diagnostic sensors utilized to gage hemodynamic results of angioplasty. Simpler and much faster steady flow calculations at the same mean flow rates were also made for comparison purposes, and to determine the necessity of slower and more difficult pulsatile flow simulations. In the discussion section some of the numerical data for the angiographic model Fig. 1 are shown in nondimensional form over a range of Re e from 100 to 400 and compared to some previous numerical and experimental flow studies with stenoses models that have somewhat similar characteristic length scales and relatively mild to moderate lumen narrowing, d m /d e, by using mean flow Reynolds number similarity. Methods The in vivo data set of Wilson et al. 2 in a 32 patient group undergoing percutaneous transluminal balloon coronary angioplasty PTCA was used. The patients had single-vessel, singlelesion coronary artery disease. Dimensions and shape of the coronary stenosis after angioplasty were obtained from biplanar angiograms Fig. 1. After PTCA the average minimal diameter increased to d m 1.8 mm (A m 2.5 0.1 mm 2 ) within treated lesions; initially d m 0.95 mm (A m 0.7 0.1 mm 2 ). Average proximal diameter d e 3 mm was relatively unchanged in the procedure, producing a residual 40 percent mean diameter stenosis after PTCA with distal diameter d r d e. Additional dimensional data on the lesion shape were used from a similar stenosis described by Back and Denton 17. Balloon angioplasty both lengthened and widened the narrowest region of the stenosis with resulting l m 3 mm. Despite axial redistribution of plaque away from the narrowest region, the constriction length l e 6 mm and divergence length l r 1.5 mm were roughly unchanged by balloon dilation. The constriction segment was approximately conical and the divergence segment relatively abrupt. Separate measurements of coronary flow reserve by Wilson et al. 2 with a 3F pulsed Doppler ultrasound catheter (d 1.0 mm with tip positioned proximal to the coronary lesions with minimal blockage improved to a value of 3.6 0.3 in the patent group after angioplasty, compared to 2.3 0.1 before PTCA. The coronary flow waveform used in the flow simulations Fig. 2 was obtained in our laboratory from in vitro calibration 11 and smoothing the fluctuating Doppler signal. The spatial average velocity (ū(t)) across the flow needed for flow simulations is similar to that from Doppler catheter measurements in the left anterior descending LAD and circumflex LCX of patients undergoing PTCA 18. The ratio of peak systolic to peak diastolic velocity was 0.4 with ratios from the right coronary artery RCA often larger than for the LAD and LCX. The pulsatility index, PI (ū p t ū m t )/u, for the waveform Fig. 2 was 1.86 since the minimum value ū m t was set to zero. In Fig. 2, the peak velocity ū p t corresponds to a normalized velocity of 1.0, so that the ratio of mean to peak velocity u /ū p t is 0.537. For hyperemic response, values of PI often decrease as flow rate increases. Numerical Method. The flow simulations were carried out by solving the mass and momentum equations for pulsatile blood Fig. 2 Coronary flow waveform ūõū pàt versus t Journal of Biomechanical Engineering AUGUST 2000, Vol. 122 Õ 311

flow using a Galerkin finite element method FEM 19. The Carreau model was used for shear-rate-dependent non-newtonian viscosity of blood with local shear rate calculated from the velocity gradient through the second invariant of the rate of strain tensor 15. The conservation equations for mass and momentum for axisymmetric flow, relations for local shear rate and blood viscosity ( ), and details of the finite element method have been previously described 16. In the simulation, the composite lesion was assumed to have a smooth, rigid plaque wall, and round concentric shape with mean diameter d 0. Plaque geometry was presumed to remain the same for hyperemic conditions, consistent with studies documenting the failure of flow-dependent dilation mechanisms in atherosclerotic coronary arteries secondary to endothelial dysfunction 20,21. Arterial wall motion associated with pressure pulsation is believed to be relatively small in plaque regions because of the decreased wall elasticity. Hence, the arterial wall was considered to be essentially rigid. Estimated wall shear stress and translesional pressure drops are expected to be the lower bound since plaque irregularities including wall roughness, lumen dissections, and intimal flaps that may occur after angioplasty procedures were not accounted for in the flow simulation, nor were unknown lesion curvatures and vessel bending due to heart motion. The no-slip boundary condition u i 0 was specified on the remodeled plaque wall assumed to be rigid, and the symmetry condition was used along the axis of the lesion. However, the proximal and distal bounding conditions are inherently difficult to control. In the proximal vessel, the spatial velocity profile was initially taken to be the Poiseuille flow relation for the axial velocity u u ū e 2 1 r 2 r e (1) where ū e (t) is the spatial average velocity across the flow. The calculations were initiated a distance proximal to the lesion 2 cm to allow the pulsatile non-newtonian blood velocity profile to develop before the inlet of the lesion not given simply by Eq. 1 consistent with the mean flow waveform shape, i.e., ū e (t) Fig. 2. This numerical procedure, an extension of the earlier simpler mathematical treatment by Womersley 22, produces stable numerical calculations in the lesion. Since proximal flow development distances were much shorter than 2 cm, upstream side branches may not have significantly altered the inlet profile. Branch geometries proximal and distal to lesions are not given in most clinical reports including Wilson et al. 2. Wilson and Laxson 6 stress the importance of making blood velocity measurements immediately proximal to a stenosis, and distal to any major proximal side branch. In the distal vessel, oscillatory spatial flow reattachment processes occur during the cardiac cycle, and fluctuations in the flow may arise depending on the magnitude of the Reynolds number e.g., 23 and expansion diameter ratio d m /d r. Mean flow Reynolds numbers R e e ranged moderately from 100 300 over the mean flow rate range Q 50 150 ml/min used in the pulsatile calculations. Fluctuations in the shear layer between the central jet and the reverse flow region along the wall in the distal vessel, however, are not believed to have occurred based on the dye filament streamline observations of Azuma and Fukushima 24 in a stenosis model with the same values of d m /d r 0.6, and d e d r as in the present case. They observed that for proximal Re e to 400 (Re m 670), the flow was laminar. At Re e 530, the distal streamline began to oscillate indicating disturbed flow, and at a higher Re e 830, the oscillation was amplified to a more chaotic state typical of turbulence. Yongchareon and Young 25 give the critical Reynolds number Re e * 2384 d m /d e 2 for the initiation of distal turbulence in models of arterial stenoses (d e d r ), which is consistent with the data of Azuma and Fukushima 24 and correlated the dog aorta measurements by Sacks et al. 26, with implanted orifice plates by detection of murmurs with a catheter microphone. Based on these experimental observations, we believe the laminar flow code is adequate to describe the pulsatile blood flow field along the residual lesion and in the distal vessel after angioplasty during hyperemic flow conditions. In principal, the outflow boundary condition in the distal vessel need not be specified in detail, since its values are effectively determined by extrapolation similar to finite difference schemes. The main concern is where the calculations should be truncated, bearing in mind the usual presence of distal vessel side branches. As a guide, simpler steady flow calculations were carried out at the same mean flow rates for a distal distance 2.2 cm. The peak pressure recovery location, which occurs downstream of the flow reattachment location Fig. 4, was found to be 0.8 and 1.65 cm distal to the lesion as the flow rate increased from 50 to 100 ml/min, respectively. At the higher mean flow rates, pressure recovery was incomplete in the calculation domain, but the major component, p r was indicated Fig. 6. Conversely, if any major distal side branch was nearer the lesion, one could roughly estimate the smaller pressure recovery that would effectively increase the lesion pressure drop. However, there may be a further pressure rise, p d in the distal vessel beyond any major side branch because of a loss of momentum therein owing to flow through the branch. Earlier measurements in our laboratory in simple branch models without proximal lesions 27, indicated that the magnitude of p d scaled on the dynamic pressure (1/2) u r 2 in the main lumen distal herein and increased primarily with the ratio of branch to main lumen flow rate, with weaker dependence on Reynolds number and branch angle. Thus, in principal, a fraction of the dynamic pressure in the lesion throat region was expected to be recovered distal to the lesion. In an earlier flow study by Mates et al. 10 in isolated stenosis models, some pressure recovery was also observed in the axisymmetric models, though not in the highly asymmetric models. The pulsatile flow calculations were truncated at a distal distance 2.2 cm as a compromise. Reported pressure drops p(t) p r p e are instantaneous pressure differences between the lesion inlet and distal region, thereby including pressure recovery. The calculations were done for two consecutive pulse cycles in order to compare them and to obtain accurate results with numerical data being reported for the second cycle where convergence was obtained. Heart rate was 75 beats/minute period of heartbeat T 0.8 s and the density of blood, 1.05 g/cm 3. Depending on the velocity pulse shape, the calculation time steps are varied between 1 10 4 to 1 10 5 s. The finite element computer code 28 was used to formulate and solve the matrix equations. The mixed formulation FEM was used so that both velocity and pressure in the conservation equations are calculated simultaneously. The mesh plot for the stenosis is shown in Fig. 1. For this simulation a Sun Ultrasparc 2 computer, having 200 MHz, 256 MB RAM, 4 GB disk was used with results downloaded to an IBM-PC computer for plotting. The CPU time for each time step was approximately 2.7 s. Calculation time steps decreased below 1 10 5 s and hence, the number of time steps increased significantly with increasing mean flow rate, which precluded obtaining converged solutions at a higher value of 200 ml/min. The appendix in Banerjee et al. 16 contains previous investigations that contribute to the validation of the methodology. Results Axial velocity components of the calculated flow are reported as a function of radius r along the axial z direction Fig. 3, and wall shear stress Figs. 4 and 5 and pressure distributions Figs. 6 and 7 are shown along the entire lesion and distal vessel. The origin of the axial coordinate (z 0) is just proximal to the lesion inlet a distance of 0.2 cm Fig. 1 ; values of z 0.5 and 0.95 cm correspond to the midpoints of the constriction and throat segments, respectively, and z 2.2 cm is distal to the lesion. The pulsatile data depict increasing times during the second cycle cor- 312 Õ Vol. 122, AUGUST 2000 Transactions of the ASME

Fig. 3 Velocity profiles at some locations along the stenosis at various times during the cardiac cycle for Q Ä50, 100, and 150 mlõmin responding to near peak systolic (t 0.89 s, and diastolic (t 1.20 s and times t 1.41 and 1.53 s during the deceleration phase in diastole, as denoted by points I, II, III, and IV in the waveform Fig. 2 for which values of ū/ū p t are respectively 0.35, 0.99, 0.70, and 0.17. Velocity Profiles. The variation of axial velocity profiles during the cardiac cycle at progressive locations along the lesion are shown vertically upward in Figs. 3 A 3 D, 3 E 3 H, and 3 I 3 L at mean flow rates Q of 50, 100, and 150 ml/min, respectively. Proximal to the lesion where Doppler catheters are used, the profiles if normalized by the centerline velocity u cl ) are close to parabolic in shape at peak diastolic point II and at point III, but deviate considerably near peak systolic point I and the latter portion of the deceleration phase point IV where values of ū(t) are lower. This trend is evident over the range of mean flow rates, Q 50 150 ml/min. Along the constriction region where spatial flow acceleration occurs, the velocity profiles at the midpoint are steeper in the wall region and somewhat more uniform in the higher velocity core flow, although the shapes of the pro- Fig. 4 Arterial wall shear stress w distributions along the stenosis for the reference steady flow calculations at QÄ50, 100, 150, and 200 mlõmin Journal of Biomechanical Engineering AUGUST 2000, Vol. 122 Õ 313

Fig. 5 Arterial wall shear stress w distributions along the stenosis at various times during the cardiac cycle for Q Ä50, 100, and 150 mlõmin Fig. 7 Axial pressure drop pàp e along the stenosis at various times during the cardiac cycle for Q Ä50, 100, and 150 mlõmin Fig. 6 Axial pressure drop pàp e along the stenosis for the reference steady flow calculations at QÄ50, 100, 150, and 200 mlõmin files vary during the cardiac cycle, and with mean flow rate. In the throat region where mean flow velocities are highest along the lesion, the velocity profiles at the midpoint indicate that viscous effects are more confined to the near wall region with a more uniform core flow like boundary layer type flow during the cardiac cycle. The viscous wall shear layer thickness decreases at the higher mean flow rates. Another way to view these profiles, which depict the spatial transition of pulsatile viscous blood flow proximal to the lesion to a viscous wall shear layer type flow in the throat region, is to follow the instantaneous velocity profiles shown in Figs. 3 A 3 D, 3 E 3 H, 3 I 3 L, at the particular times I, II, III, IV during the cardiac cycle, thus providing a snapshot picture of the flow simulation. The flow separated from the lesion wall in the divergent region and reattached to the distal vessel wall. The uppermost velocity profiles in Figs. 3 D, 3 H, 3 L at z 2.2 cm about 1 cm distal to the lesion were primarily S-shaped. The reattachment location oscillated spatially during the cardiac cycle, and moved downstream with increasing mean flow rate. The particular profiles shown in Fig. 3 D are downstream of reattachment in the spatial flow redevelopment region, while at the higher mean flow rates, 314 Õ Vol. 122, AUGUST 2000 Transactions of the ASME

Table 1 Steady flow pressure recovery Figs. 3 H, 3 L, the profiles are within the reattachment region as indicated by the reverse flow velocities in the wall vicinity during phases of the cardiac cycle. The pulsatile flow field was very complex in the spatial flow deceleration region in the divergent segment of the lesion and in the distal vessel where pressure recovery occurred. Wall Shear Stress Distributions. The variation of shear stress w along the arterial wall is shown in Fig. 4 as a reference datum for steady flow rates Q of 50, 100, 150, and 200 ml/min. In the constriction region, w increases as flow velocities increase; a local spatial peak value occurs at the end of the constriction region. In the throat region, w decreases as the wall shear layer grows in thickness along the throat, and core flow acceleration occurs. Values of w decrease sharply in the divergent region where spatial flow deceleration occurs, and the flow separated from the wall ( w 0). The flow reattached along the wall of the distal vessel ( w 0) and w increased again in the flow redevelopment region. In the flow separation region, values of w 0, due to the reverse flow velocities in the wall region of the recirculation zone. The wall shear stress levels increased nonlinearly with flow rate Fig. 4. For example, at the midpoint of the throat region, values of wm increased from 62 to 315 dynes/cm 2, or by a factor of 5.1 for the fourfold increase in Q. In the proximal vessel (z 0) the corresponding increase of w from 12.6 to 45.1 dynes/cm 2 was by a factor of 3.6. In the reverse flow region, minimum values of w 0 also increased with flow rate, and the reattachment location moved downstream nearly linearly from the throat exit with increasing flow rate. The pulsate flow calculations of wall shear stress distributions during the cardiac cycle are shown in Figs. 5 A, 5 B, and 5 C at corresponding mean flow rates Q of 50, 100, and 150 ml/min. By using the time-integrated waveform factor u /ū p t Q /Q p t 0.537 shown in Fig. 2, the corresponding peak diastolic flow rates Q p t 93, 186, and 279 ml/min. During the cardiac cycle there was wide variation in the magnitude of oscillatory values of w at time points I, II, III, and IV. The relative level of these oscillations increased with mean flow rate. For example, at the midpoint of the throat region, values of wm at near-peak diastolic flow point II increased from 126 to 476 dynes/cm 2 factor of 3.8 for the threefold increase in mean flow rate in the numerical calculations. Elevation in phasic peak values wp are higher at the inlet of the throat region (z 0.8 cm as seen in Fig. 5 C where values reached about 700 dynes/cm 2 at an elevated flow rate Q 150 ml/min. The flow separation location in the divergent region varied little during the cardiac cycle, nor with mean flow rate. The flow reattachment location in the distal vessel moved downstream with increasing instantaneous flow rates during the cardiac cycle, and the largest distance near peak diastolic flow point II increased with mean flow rate. For these results, reattachment locations occurred within the calculation domain. In the oscillating recirculation zones, the level of values of w 0 increased with instantaneous flow rate, and with mean flow rate. On an overall basis the pulsatile distributions of w shown in Fig. 5 are similar in general shape as the reference values for steady flow Fig. 4. Although there are some differences in detail, the pulsatile flow field appears to be principally quasi-steady. Phasic and mean flow values of wm in the throat region are discussed subsequently. Pressure Distributions. The variation of the pressure distribution p p e along the arterial wall is shown in Fig. 6 as a reference datum for steady flow rates Q of 50, 100, 150, and 200 ml/min. Here, pressure differences are referenced to p e at the lesion inlet, z 0.2 cm. The general shape of the pressure variation consists of increasing pressure drops along the constriction region, a nearly linear decrease in the throat region, and pressure recovery in the divergent region and distal vessel. The predominant pressure drop contribution occurs across the constriction region due to wall friction shear stress acting on the arterial wall and changes in momentum as the flow undergoes spatial acceleration. In the throat region, the pressure drop contribution is due to wall friction, and the additional momentum change as the core flow accelerated due to wall shear layer growth. In the divergent region, the initial rise in pressure is associated with flow deceleration, and the plateau region with flow separation. The major component of pressure recovery occurs in the distal vessel due to viscous processes in the separated flow region, and downstream of flow reattachment in the flow redevelopment region. Since there is little information on pressure recovery coefficients c pr for blood flow in the literature, these numerical calculations were used to evaluate a factor k in the following relation: c pr p r 1/2 ū kc 2 pr (2) m Here, p r p r p m2 where p r is distal peak pressure, and p m2 the pressure at the throat exit z 1.1 cm, 2 ū 1 2 m is the dynamic pressure in the throat region, and c pr the often cited high Reynolds number limit across a sudden expansion e.g., 29. c pr 2 A m A r 1 A m A r (3) Values of k are given in Table 1, exceed unity, and decrease with throat Reynolds number Re m d e d m Re e Overall pressure drops, p s p e p r, increased appreciably with flow rate, as evident in Fig. 6. Values of p s, shown in Table 2, increased by a factor of 8 for the fourfold increase in Q. The pulsatile flow pressure distributions during the cardiac cycle are shown in Figs. 7 A, 7 B, 7 C at corresponding mean flow rates Q of 50, 100, and 150 ml/min. As with the wall shear stress distributions, there was wide variation in magnitude of these oscillating distributions at time points I, II, III, IV. The relative level of these oscillations increased with mean flow rate. At the elevated mean flow rate of 150 ml/min, pressure was not fully recovered in the calculation domain at the higher instantaneous flow rates during the cardiac cycle. The shapes of the pressure distributions retain the features of the reference values for steady flow Fig. 6 in the various regions of the lesion and distal vessel. However, as will be seen, moderate increases in p above the steady flow values p s are found. Phasic and Mean Flow Relations. The instantaneous overall pressure drops p p r p e across the lesion including pressure recovery is shown in Fig. 8 during the cardiac cycle at mean flow rates of 50, 100, and 150 ml/min. Although the shape of p with time is similar to the waveform velocity ū shown in Fig. 2, there is some phase lag in the p(t) ū(t) relation since the pressure force must overcome the wall shear force to provide the net force Journal of Biomechanical Engineering AUGUST 2000, Vol. 122 Õ 315

Table 2 Comparison of calculated magnitudes of mean pressure drops with Poiseuille flow theory Fig. 8 Overall pressure drop p across the stenosis during the cardiac cycle for Q Ä50, 100, and 150 mlõmin Fig. 10 Time mean pressure drop-flow rate relation, p ÀQ, for physiologic flow, other comparisons, and with the catheter present after angioplasty Fig. 9 Phasic variation of the wall shear stress wm at the midpoint of the throat region for Q Ä50, 100, and 150 mlõmin for convective and unsteady flow accelerations and decelerations. The phase angle p between the local peak p and ū values were calculated in the usual way. That is, if a peak p value occurs at t 1, and a peak ū value occurs at t 2 t 1 p, then since the circular frequency, 2 f 2 /T, the phase angle p 360(t 2 t 1 )/T deg. At the beginning of the systolic phase where ū o, the phase angle p 14 deg and the phase shifted p is positive rise in pressure. The value of p at the local systolic peak is about 20 deg, but at peak diastolic flow there is virtually no phase lag in the p ū relation. Thus, the phase angle p varied during the cardiac cycle, exhibiting hysteresis effects in the p ū relation. The phasic variation of the wall shear stress wm at the midpoint in the throat region is shown in Fig. 9 over the mean flow rate range. The shape of wm (t) is like ū(t), and the corresponding phase angle in the wm ū relation is only 4 deg at ū 0 and at peak systole. At the diastolic peak, the values were in phase. Integration of p(t) and wm (t) over the cardiac cycle gives the time average values shown in Figs. 8 and 9 by the dotted lines over the range of mean flow rates. The mean wall shear stress levels of wm of 64, 143, and 234 dynes/cm 2 correspond to Q of 50, 100, and 150 ml/min are within 6 percent of the steady flow values wms. The mean pressure drop p flow rate Q relation is shown in Fig. 10 as curve A, and values are given in Table 2. In the usual way, the absolute magnitudes of pressure drops or losses, p e p r, are obviously p in our nomenclature. The value of p at Q 150 ml/min was adjusted to 5.5 mmhg because pressure recovery was not complete at the higher instantaneous flow rates, and a curve fit by plotting the pressure drop coefficients c pm p / 1 2 u 2 m versus R e m on a log log plot was used for extrapolation to Q 200 ml/min to give a value of 8.8 mmhg. Discussion The present calculations for steady flow allow appraisal of simpler mean flow analyses to estimate the pressure drop due to wall friction and spatial changes of flow momentum. In the approximate method used by Back et al. 30 with and without angioplasty catheters present, local mean Poiseuille flow was assumed in the constriction and throat regions, whereas in the divergence and distal regions, pressure recovery was specified in terms of a pressure recovery coefficient. At relatively low Reynolds numbers 316 Õ Vol. 122, AUGUST 2000 Transactions of the ASME

due to flow blockage, there was good correlation between the approximate method with an angioplasty catheter present in the flow and detailed solutions of the mass and momentum equations for pulsatile flow 16. In the presence of a catheter, flow was viscous dominated and quasi-steady with a negligible phase lag in the p ū relation. For physiological flow in the absence of a catheter where flow rates and Reynolds numbers were higher and momentum changes significant, the calculated wall shear stress was higher along the constriction and throat regions Fig. 4 than inferred from the Poiseuille relation, w (32/ )( /d 0 3 )Q,and thus frictional pressure drops were larger, and there was an additional momentum change in the throat region due to entrance effects. These are the main reasons why those estimated mean pressure drops shown in Fig. 10 as curve C were about 30 percent less than the values of p s obtained from the more accurate numerical calculations for steady flow Fig. 10, curve B, which also more precisely determine net momentum changes and pressure recovery for blood flow. The Poiseuille relation is often used in codes to calculate the wall frictional component of mean pressure losses across stenoses, e.g., Brown et al. 31. Most codes emanate from the earlier correlation recommended by Young and Tsai 9, and are used clinically to compute hemodynamic parameters from quantitative coronary angiography. Direct measurements of coronary wall shear stress are not possible in man, and are very difficult to make in stenosis models. Consequently, there is interest in methods that more accurately predict levels of wall shear stress in coronary lesions, such as described herein. For pulsatile flow, inertial effects were also important, and along with spatial boundary layer like phenomena during the cardiac cycle, lead to calculated increases in p of about 20 percent above the steady flow values p s at the same mean flow rate or Reynolds number Fig. 10, curves A and B. This trend is similar to the pressure drop measurements by Cho et al. 11 in an atherosclerotic human coronary artery casting with a moderate stenosis, for which values of p were about 30 percent larger than p s at the same mean flow rate or Reynolds number in the physiological flow range. The moderate increases in p / p s calculated after angioplasty, correspond to values of the frequency parameter e 2.25, proximal to the lesion, and a smaller value of m 1.35 in the throat region. In the physiologic flow regime there was a phase lag in the p ū relation. The important result of the physiological flow simulation is that with the enlarged luminal cross-sectional areas after angioplasty in the patient group of Wilson et al. 2, the calculated p was only about 1 mmhg at a mean resting flow rate Q 50 ml/min, and increased moderately to a value of about 8 mmhg for hyperemic flow. Thus, for the measured value of coronary flow reserve of 3.6 0.3 in the patient group after PTCA, the estimated increase in p was about a factor of 7. In a simple power law relation, p Q n, the exponent n was about 3/2. The corresponding calculated wall shear stress levels wm at the midpoint of the throat region increased from about 60 to 290 dynes/cm 2,orby about a factor of 4.5. To aid interpretation of data derived from clinical studies using diagnostic catheter sensors, the particular results of our flow simulations are shown in Fig. 10 by the mean p Q relation for two situations. The lower curve A is for physiological flow without a catheter present within the residual lesions to disturb/change the flow as described in this investigation. Insertion of a catheter across a lesion to measure pressure drops, increases the flow resistance because of the tighter artifactual stenosis, thereby significantly shifting the p Q relation upward and to the left as indicated by the upper most curves in Fig. 10 from the flow simulation by Banerjee et al. 16 in conjunction with the pressure measurements by Wilson et al. 2 with the angioplasty catheter (d 1.4 mm. The proximal measurements of coronary flow reserve by Wilson et al. 2 with minimal blockage corresponds to the lower curve A in Fig. 10. The hemodynamic consequences of catheter presence or absence are evident and necessitate development of smaller diameter catheter sensors for measurement of transluminal pressure drops to improve the accuracy of predicted hemodynamics across lesions after angioplasty. Stenosis Geometries. From quantitative angiography, the important geometric parameters are the degree to which atheromas narrow vessel lumens, smallest widths, and lesion lengths. Even biplane angiography resolves an average vessel width, with crosssectional area usually calculated from the equation for an ellipse, as done by Wilson et al. 2, which we converted to mean diameters. Interpretation of angiograms may also include measuring on average the constriction angle and the exit angle of the divergent segment. For the angiographic model Fig. 1 with rounded segment junctions, radial and axial lengths relative to proximal vessel widths (d e 3 mm and half-angles are (d m /d e ) 0.6, c (l c /d e ) 2, m (l m /d e ) 1, tan c 0.5 ((d e d m )/l c ) ( c 5.7 deg ; and in the divergent segment, r (l r /d e ) 0.5, tan r 0.5 ((d r d m )/l r )( r 21.8 deg since d r d e. Much work has been carried out for stenosis models with cosine-shaped walls e.g., 9, which are symmetric about the throat plane, and are characterized by the geometric ratios, 1 (r m /r e ) and l 0 /r e, where l 0 is the model half-length. These models have circular arc throat segments of relative radius of curvature (r c /r e ) (2 2 )/( 2 ), a mean taper angle tan 1 /, and a maximum slope of the contour at l 0 /2, where m tan 1 ( )/(2 ). In general, these models for mild to moderate lumen narrowing d m /d e have contours with relatively gradual throat regions, and relatively small taper. One way to view the angiographic model Fig. 1 is to pass a plane through the midpoint of the throat region, which gives a relative upstream length of (l u /d e ) c ( m /2) 2.5. The contour is obviously not symmetric about this plane, with a much smaller downstream portion of relative length (l d /d e ) ( m /2) r 1.0. The constriction angle c of the nearly conical portion is relatively small, but the divergence angle r is much larger, being about 4 c. Elevated Wall Shear Stress. Persistent elevations in mean wall shear stress and pulsatile variations in residual stenotic segments after coronary angioplasty may contribute to early thrombotic complications and later restenosis phenomena e.g., 17 although the responsible flow-related mechanisms remain to be elucidated. Here, we focus on the reference steady flow calculations Fig. 4 and show, in Fig. 11, the peak wall shear stress wp in the near vicinity of the throat inlet normalized by the proximal value we as a function of proximal Reynolds number Re e from 100 to 400, on a log log basis. The ratio wp / we increases by a Fig. 11 Normalized peak wall shear stress wp Õ we variation with proximal Reynolds number Re e for steady flow Journal of Biomechanical Engineering AUGUST 2000, Vol. 122 Õ 317

factor of 6.6 to 10.2 over the range of Re e. A kinematic viscosity 0.035 cm 2 /s and viscosity 0.0368 Poise were used in our reported values of Re e. In the blood flow simulations, the shear stress with ( ). In the Carreau model with asymptote 0.0345 Poise as, the value 0.0368 Poise corresponds to a shear rate of 1400 s 1 and 52 dynes/cm 2, while for 300 s 1, 0.041 Poise and 12.2 dynes/cm 2. Thus the variation of we Fig. 4 with Re e corresponds to a variation in we of about 10 percent. Pritchard et al. 32 reported numerical calculations of the Navier Stokes equations for steady flow of a Newtonian fluid at a particular Re e 200 Poiseuille flow upstream for a larger model very similar in shape to our angiographic model up to the end of the throat region ((d m /d e ) 0.632, c 2, m 1, c 5.4 deg but with a sudden downstream expansion, r 0, r 90 deg, and d r d e. The shape of the normalized wall shear stress w / we distribution along the conical constriction segment and uniform diameter throat segment was very similar to our model, and their peak value wp / we very near the throat inlet of 8.6 was near our value of 8.3 at Re e 200. Some differences are expected because of shear dependent blood viscosity mentioned above and the differences in degrees of stenosis. Estimates of the magnitude of peak values wp / we are shown by the dotted curve in Fig. 11 from the laminar boundary layer similarity analysis by Back 33 and Back and Crawford 34 for conical constrictions. These values are in fair agreement with our peak values, being about 15 percent low at Re e 100, and 10 percent high at Re e 400. In their analysis for steady flow of a Newtonian fluid, wp was normalized by the proximal Poiseuille value, we, so that the ratio wp / we is proportional to (Re e ) 1/2, and wp Q 3/2, typical of laminar boundary layer theory. Other applications of the similarity method were given by Back and Crawford 34 for coronary artery atherosclerotic lesion models, and compared to numerical solutions of the Navier Stokes equations in constriction segments including pulsatile flow 14. Siegel et al. 13 solved the Navier Stokes equations for steady flow of a Newtonian fluid through cosine-shaped stenoses models ((d m /d e ) 0.707, 0.500, and 0.316; (l 0 /d e ) 1.5 and 3 over a Re e range from 100 to 400, assuming proximal Poiseuille flow. Because they used a proximal wall shear rate of 4ū e /d e rather than the Poiseuille value of 8ū e /d e ) to normalize the calculated wall shear rates, the normalized values of ( w / we ) are half the normalized wall shear rates they show. In general, the normalized wall shear rates increased to a peak value upstream of the throat plane, decreased in the throat region, and the flow separated in the symmetric divergent portion at a Re es dependent upon the model shape (d m /d e and l 0 /d e ). These trends are also observable in our wall shear stress distributions Fig. 4 if viewed relative to the throat midplane. Using their correlation equation, values of peak ( wp / we 0.5(a Re 0.5 e b) are shown in Fig. 11 for two models ((d m /d e ) 0.707 and 0.500 with relative half-length ((l 0 /d e ) 3) near our value of ((l u /d e ) 2.5). These models have relatively large values of (r c /r e ) 25 and 14.6 very gradually curved throat regions and relatively small taper. Their results shown in Fig. 11 span our values of ( wp / we ) for the 40 percent diameter stenosis as expected. The Re 1/2 e dependence in their correlation is consistent with laminar boundary layer theory. The relative length dependence in the coefficient a roughly scales on 1/(l 0 /d e ) 0.5, which is similar to the similarity solutions of Back and Crawford 34 where ( wp / we ) scales on the relative length of the constriction as 1/ 0.5 c for constrictions where (sin c ) 0.5 (tan c ) 0.5 less than 3 percent error for c to 20 deg. Moreover, (Re e / c ) (Re e d e /l c ) (4Q/ l c ) is a characteristic length Reynolds number. The coefficient b depended primarily on degree of stenosis. Pressure Drop Coefficients. The calculated pressure drops p normalized by the dynamic pressure in the proximal vessel Fig. 12 Pressure drop coefficient c pe variation with proximal Reynolds number Re e for pulsatile and reference steady flow 1 2 u 2 e (c pe ) are shown in Fig. 12 as a function of the proximal Reynolds number R e e from 100 to 400 on a log log basis. In the c pe (R e e ) representation, values of c pe decrease with increasing R e e since the variation of p with Q is less than Q 2, the asymptotic, very high Re e limit. The reference steady flow calculations c pse, which lie below the pulsatile values, are also shown. The other reference curve is for Poiseuille flow through a tube without a stenosis of diameter d e 3 mm and l t /d e 10.8, where l t includes the stenosis length 1.05 cm plus the distal length 2.2 cm, the effective region of influence of the stenosis. The steady flow values of c p lie above the Poiseuille values by a factor of 2.5 and 5.0 at Re e 100 and 400, respectively. In our steady flow calculations, flow separation occurred in the divergent segment over the Re e range Fig. 4 in the adverse pressure gradient region Fig. 6. The other curve shown in Fig. 12 is from the correlation by Young and Tsai 9 of their steady flow pressure drop measurements for the cosine shaped wall model M-1 ((d m /d e ) 0.667, (l 0 /d e ) 2 with (r c /r e ) 9.7 and 4.8 deg in a relatively large tube using distilled water. Flow separation occurred in the symmetric divergent region for Re e 200. Initial flow instabilities, measured with a hot-film probe located at l/d e 2 distally, were observed at Re e 300, but not at a more distal location l/d e 8 until Re e 800. The pressure taps were located at l/d e 8 relative to the throat plane, or l t /d e 16, so that some pressure drop proximal to the model was included. Conversion of their values of p/ ū 2 e by multiplying by 2, result in values of c pe Fig. 12 for the mildly curved 33 percent diameter stenosis model that lie below our steady flow values for the 40 percent diameter stenosis. This trend is expected because of the strong dependence of p on the degree to which atheromas narrow vessel lumens d m /d e. The pressure drop calculations by Deshpande et al. 12 for the M-1 model with l t /d e 16 are in good agreement with the Young and Tasi 9 data for Re e 200 to 400 considering the accuracy of p measurements at relatively low flow rates through the relatively large models, and distal flow fluctuations that were damped further downstream. Much work has been reported on the initiation, damping, and amplification of distal flow oscillations e.g., 35, which were discussed in connection with our angiographic model in the numerical method section. Mates et al. 10 also made pressure measurements in their model with the same cosine shape as the M-1 model of Young and Tsai 9, that was also placed in a relatively large tube, using a more viscous fluid. Their reported pressure drop coefficients c pe relative to the upstream tap at l/d e 6 are higher than those shown in Fig. 12 from the Young and Tsai 9 measurements, yet they show the Young and Tsai 9 measurements of c pe in rough 318 Õ Vol. 122, AUGUST 2000 Transactions of the ASME

agreement with their values for Re e from about 200 to 500. Some of these discrepancies are associated with the location of the upstream pressure tap in the tube and the pressure drop to the model inlet which are included in their values of c pe. In our flow simulations the pressure drop is referenced to the stenosis inlet value Figs. 8 and 10. In both of these experimental investigations, values of c pe referenced to the stenosis inlets would lie below their reported values for the M-1 stenosis models. Our in vitro flow studies e.g., 11 used human coronary artery castings of diseased segments of actual size, and therefore, pressure drops were much larger and more accurately measured than in larger model studies with relatively mild degrees of stenosis. The preceding discussion illustrates the particular importance of tap location in interpreting pressure drop measurements. Endovascular diagnostic catheter sensors used to measure lesion pressure drops also depend on measurement locations primarily distal, and complex hemodynamic flow-sensor interactions that elevate the pressure drop and reduce blood flow during measurements. Nomenclature A flow cross-sectional area c pr pressure recovery coefficient c pe pressure drop coefficient based on proximal dynamic pressure c pm pressure drop coefficient based on throat dynamic pressure d e proximal vessel diameter d m minimal lesion throat diameter d 0 mean vessel diameter d r distal vessel diameter l c length of constriction region l m length of narrowest throat region l 0 half-length of symmetric cosine wall models l r length of divergence region p pressure p overall pressure drop Q blood flow rate r radial distance r c throat radius of curvature for symmetric cosine wall models r 0 mean vessel radius R e e proximal mean Reynolds number (4/ )(Q / d e ) R e m throat mean Reynolds number (4/ )(Q / d m ) t time u i velocity vector u axial velocity z axial distance frequency parameter (d 0 /2)( / ) 0.5 c, r constriction and divergent segment half-angles shear rate degree of stenosis 1 (d m /d e ) blood viscosity phase angle relative half-length l 0 /r e c, m, r relative lengths l c /d e, l m /d e, l r /d e kinematic viscosity / blood density shear stress period of cardiac cycle circular frequency, 2 /T Subscripts e proximal vessel m narrowest throat condition o vessel p t peak temporal value r distal vessel w wall condition Superscripts time average mean over cardiac cycle spatial average across flow References 1 Cole, J. S., and Hartley, C. J., 1977, The Pulsed Doppler Coronary Artery Catheter: Preliminary Report of a Technique for Measuring Rapid Changes in Coronary Flow Velocity in Man, Circulation, 56, pp. 18 25. 2 Wilson, R. F., Johnson, M. R., Marcus, M. L., Alyward, P. E. G., Skorton, D. J., Collins, S., and White, C. W., 1988, The Effect of Coronary Angioplasty on Coronary Flow Reserve, Circulation, 77, pp. 873 885. 3 Doucette, J. W., Corl, P. D., Payne, H. M., Flynn, A. E., Goto, M. N., Nassi, M., and Segal, J., 1992, Validation of a Doppler Guide Wire for Intravascular Measurement of Coronary Artery Flow Velocity, Circulation, 85, pp. 1899 1911. 4 Segal, J., Kern, M. J., Scott, N. A., King, S. B., III, Doucette, J. W., Heuser, R. R., Ofili, E., and Siegel, R., 1992, Alterations of Phasic Coronary Artery Flow Velocity in Humans During Percutaneous Coronary Angioplasty, J. Am. Coll. Cardiol., 20, pp. 276 286. 5 Anderson, H. V., Roubin, G. S., Leimgruber, P. P., Cox, W. R., Douglas, J. S., Jr., King, S. B., and Gruentzig, A. R., 1986, Measurement of Transstenotic Pressure Gradient During Percutaneous Transluminal Coronary Angioplasty, Circulation, 73, pp. 1223 1230. 6 Wilson, R. F., and Laxson, D. D., 1993, Caveat Emptor, A Clinician s Guide to Assessing the Physiologic Significance of Arterial Stenoses, Cathet. Cardiovasc. Diagn., 29, pp. 93 98. 7 Emanuelsson, H., Lamm, C., Dohnal, M., and Serruys, P. W., 1993, High Fidelity Translesional Pressure Gradients During PTCA Correlation With Quantitative Coronary Angiography, J. Am. Coll. Cardiol., 21, p. 340A. 8 DeBruyne, B., Pijls, N. H. J., Paulus, W. J., Vantrimpont, P. J., Sys, S. U., and Heyndrickx, G. R., 1993, Transstenotic Coronary Pressure Gradient Measurement in Humans: In Vitro and In Vivo Evaluation of a New Pressure Monitoring Angioplasty Guide Wire, J. Am. Coll. Cardiol., 22, pp. 119 126. 9 Young, D. F., and Tsai, F. Y., 1973, Flow Characteristics in Models of Arterial Stenosis 1 Steady Flow, J. Biomech., 6, pp. 395 410. 10 Mates, R. E., Gupta, R. L., Bell, A. C., and Klocke, F. J., 1978, Fluid Dynamics of Coronary Artery Stenosis, Circ. Res., 42, pp. 152 162. 11 Cho, Y. I., Back, L. H., Crawford, D. W., and Cuffel, R. F., 1983, Experimental Study of Pulsatile and Steady Flow Through a Smooth Tube and an Atherosclerotic Coronary Artery Casting of Man, J. Biomech., 16, pp. 933 946. 12 Deshpande, M. D., Giddens, D. P., and Mabon, R. F., 1976, Steady Laminar Flow Through Modelled Vascular Stenoses, J. Biomech., 9, pp. 165 174. 13 Siegel, J. M., Markou, C. P., Ku, D. N., and Hanson, S. R., 1994, A Scaling Law for Wall Shear Rate Through an Arterial Stenosis, ASME J. Biomech. Eng., 116, pp. 446 451. 14 Back, L. H., Radbill, J. R., and Crawford, D. W., 1977, Analysis of Pulsatile Viscous Blood Flow Through Diseased Coronary Arteries of Man, J. Biomech., 10, pp. 339 353. 15 Cho, Y. I., and Kensey, K. R., 1991, Effects of the Non-Newtonian Viscosity of Blood on Flows in a Diseased Arterial Vessel: Part 1, Steady Flows, Biorheology, 28, pp. 241 262. 16 Banerjee, R. K., Back, L. H., Back, M. R., and Cho, Y. I., 1999, Catheter Obstruction Effect on Pulsatile Flow Rate Pressure Drop During Coronary Angioplasty, ASME J. Biomech. Eng., 121, pp. 281 289. 17 Back, L. H., and Denton, T. A., 1992, Some Arterial Wall Shear Stress Estimates in Coronary Angioplasty, Advances in Bioengineering, ASME BED-Vol. 22, pp. 337 340. 18 Sibley, D. H., Millar, H. D., Hartley, C. J., and Whitlow, P. L., 1986, Subselective Measurement of Coronary Blood Flow Velocity Using a Steerable Doppler Catheter, J. Am. Coll. Cardiol., 8, pp. 1332 1340. 19 Baker, A. J., 1983, Finite Element Computational Fluid Mechanics, Hemisphere, New York, Chap. 4, pp. 153 230. 20 Drexler, H., Zeiher, A. M., Wollschlager, H., Meinertz, T., Just, H., and Bonzel, T., 1989, Flow Dependent Coronary Artery Dilation in Humans, Circulation, 80, pp. 466 474. 21 Vita, J. A., Treasure, C. B., Ganz, P., Cox, D. A., Fish, R. D., and Selwyn, A. P., 1989, Control of Shear Stress in the Epicardial Coronary Arteries of Humans: Impairment by Atherosclerosis, J. Am. Coll. Cardiol., 14, pp. 1193 1199. 22 Womersly, J. R., 1955, Method for the Calculation of Velocity, Rate of Flow and Viscous Drag in Arteries When the Pressure Gradient is Known, J. Physiol. London, 127, pp. 553 563. 23 Back, L. H., and Roschke, E. J., 1972, Shear Layer Flow Regimes and Wave Instabilities and Reattachment Lengths Downstream of an Abrupt Circular Channel Expansion, ASME J. Appl. Mech., 39, pp. 677 681. 24 Azuma, T., and Fukushima, T., 1976, Flow Patterns in Stenotic Blood Vessel Models, Biorheology, 13, pp. 337 355. 25 Yongchareon, W., and Young, D. F., 1979, Initiation of Turbulence in Models of Arterial Stenoses, J. Biomech., 12, pp. 185 196. Journal of Biomechanical Engineering AUGUST 2000, Vol. 122 Õ 319