J Artif Organs (2006) 9:165–178 © The Japanese Society for Artificial Organs 2006DOI 10.1007/s10047-006-0340-5

ORIGINAL ARTICLE

Received: January 18, 2006 / Accepted: May 30, 2006

T. Akutsu (*) · J. SaitoDepartment of Mechanical Engineering, Kanto Gakuin University,1-50-1 Mutsuurahigashi, Kanazawa-ku, Yokohama 236-8501, JapanTel. +81-45-786-7114; Fax +81-45-786-7098e-mail: akutsu@kanto-gakuin.ac.jp

Toshinosuke Akutsu, PhD · Jun Saito, BE

Dynamic particle image velocimetry flow analysis of the flow fieldimmediately downstream of bileaflet mechanical mitral prostheses

Abstract New dynamic particle image velocimetry (PIV)technology was applied to the study of the flow field associ-ated with prosthetic heart valves. Four bileaflet prostheses,the St. Jude Medical (SJM) valve, the On-X valve withstraight leaflets, the Jyros (JR) valve, and the EdwardsMIRA (MIRA) valve with curved leaflets, were tested inthe mitral position under pulsatile flow conditions to findthe effect of the leaflet shape and overall valve design on theflow field, particularly in terms of the turbulent stress distri-bution, which may influence hemolysis, platelet activation,and thrombus formation. Comparison of the time-resolvedflow fields associated with the opening, accelerating, peak,and closing phases of the diastolic flow revealed the effectsof the leaflet shape and overall valve design on the flowfield. Anatomically and antianatomically oriented bileafletvalves were also compared in the mitral position to studythe effects of the orientation on the downstream flow field.The experimental program used a dynamic PIV system uti-lizing a high-speed, high-resolution video camera to mapthe true time-resolved velocity field inside the simulatedventricle. Based on the experimental data, the followinggeneral conclusions can be made. High-resolution dynamicPIV can capture true chronological changes in the velocityand turbulence fields. In the vertical measuring plane thatpasses the centers of both the aortic and mitral valves (A-Asection), bileaflet valves show clear and simple circula-tory flow patterns when the valve is installed in theantianatomical orientation. The SJM, the On-X, and theMIRA valves maintain a relatively high velocity throughthe central orifice. The curved leaflets of the JR valve gen-erate higher velocities with a divergent flow during the ac-celerating and peak flow phases when the valve is installedin the anatomical orientation. In the velocity field directlybelow the mitral valve and normal to the previous measur-

ing plane (B-B section), where characteristic differences invalve design on the three-dimensional flow should be vis-ible, the symmetrical divergent nature of the flow generatedby the two inclined half-disks installed in the antianatomicalorientation was evident. The SJM valve, with a centraldownward flow near the valve, is contrasted with the JRvalve, which has a peripherally strong downward circulationwith higher turbulent stresses. The On-X valve has a strongcentral downward flow attributable to its large openingangle and flared inlet shape. The MIRA valve also has arelatively strong downward central flow. The MIRA valve,however, diverts the flow three-dimensionally due to itsperipherally curved leaflets.

Key words Prosthetic heart valve · Dynamic PIV · Flowvisualization · Biofluid mechanics · Medical equipment

Introduction

It has been more than 50 years since the first introduction ofthe prosthetic heart valve by Hufnagel1 in 1952 (and later byStarr2) to correct aortic valve insufficiency. Since then, vari-ous types of prosthetic heart valves have been introduced inthe market place. Major effort has been directed towarddeveloping a hemodynamically better and structurally supe-rior prosthetic heart valve. Obviously, the fluid dynamics ofthe valve plays an important role in developing moresuccessful prosthetic heart valves. A wide variety of flowanalyzing techniques such as, laser Doppler anemometry(LDA), particle tracking velocimetry (PTV), and particleimage velocimetry (PIV) has been used in analyzing theflow fields associated with prosthetic heart valves. Es-timation of the cause of hemodynamically related complica-tions such as hemolysis, platelet activation, and thrombusformation is only possible by knowing the precise velocityfield and stress distribution obtained by using thesetechniques.

It is worth pointing out the correlation between themedical experience and the in vitro experimental results.

166

A number of clinical studies3–5 have suggested that anti-anatomical orientation is preferable for mitral replacementvalves of bileaflet design. In our previous in vitro experi-mental study,6–9 it was shown that the simple circulatoryflow offered by the antianatomical orientation of the St.Jude Medical (SJM) valve seems to provide a smoother flowfield and generate less turbulence during the cardiac cycle; itis thus easy to speculate that the lower turbulent stress isless stressful to the blood components, which seems in goodagreement with the clinical experience.3,4

Flow field analysis, particularly related to the stress level,is becoming essential to understand the causes of hemolysis,platelet activation, thrombus formation, and other clinicalproblems. Quantitative results obtained by using LDA andPIV are ideally suited to elucidating such a flow field fromwhich one can estimate the effect on blood components. Forthe platelet activation study, one has to know not only thestress level but also the duration of the stress. LDA cancapture very precise velocity values and temporal resolutionis excellent. On the other hand, complete analysis of therelevant flow area using an LDA system requires severaltraverses. Thus, this approach is not suitable for a time-resolved analysis of the entire flow field. Normal PIV is avery powerful experimental tool that can capture the entireflow field of interest in a very short time and provides usefulinformation such as the velocity vector, viscous shear stress,and other useful fluid dynamical information. One couldrecord several cycles worth of information and later recon-struct sequential information by rearranging timed informa-tion using recorded cycle information, but this is not truesequential information. The introduction of the high-speeddigital camera and, particularly, more powerful continuouslaser systems, which only recently became available, hasmade it possible to continuously capture sequential imagesat high speed, allowing a true time-resolved analysis of theturbulent flow. Use of a very high resolution (1024 × 1024pixels or more) and high-speed (1000Hz or more) videocamera enables very precise analysis of the true turbulentflow associated with prosthetic heart valves, which was previ-ously only possible with the LDA system until the introduc-tion of this dynamic PIV system. We have demonstrated thetime-resolved capability of this new dynamic PIV system10

and the results were comparable to the LDA system.10–12

In this in vitro investigation, therefore, the effect of valveorientation and the leaflet design of the four bileafletmechanical valves on the three-dimensional flow field inthe downstream area was studied in more detail using adynamic PIV system utilizing a high-speed, high-resolutionvideo system to map the velocity and turbulence field.

Experimental models and test facilities

The four valves studied were 27- and 29-mm mitral valves asdescribed below (Fig. 1):

1. St. Jude Medical valve (SJM, 29mm, opening angle =85°, St. Jude Medical, St. Paul, MN, USA)

2. Jyros valve (JR, 27mm, opening angle = 80°, Manley-Western, Chichester, UK)

3. On-X valve (On-X, 27mm, opening angle = 90°, MedicalCarbon Research Institute, Austin, TX, USA)

4. Edwards MIRA valve (MIRA, 27mm, opening angle =80°, Edwards Lifescience, Irvine, CA, USA)

The SJM valve, selected as the reference valve for thisstudy, is a typical bileaflet mechanical prosthetic heart valvethat has been in successful clinical use for over 20 years. TheSJM valve has two straight half-disk leaflets with a relativelyhigh hinge location. The JR valve has a unique design withcurved leaflets and a leaflet support ring that allows theleaflets to self-rotate. This arrangement is aimed at reduc-ing the valve orientation problem. The manufacturer alsoclaims quieter operation because of the design. Althoughthe valve is not now widely used clinically, design featuressuch as the curved leaflets provide interesting flow phenom-ena, thus warranting its inclusion in this study. The pivotlocation is lower than for the SJM valve. Because of thisdifference in hinge location, the tips of the leaflets of the JRvalve protrude into the ventricle when the valve is fullyopen. The On-X valve has two straight half-disk leafletswith a relatively long support ring. This valve opens to 90°and the inlet edge forms a flare shape, both ensuring smoothflow through the valve. The MIRA valve has unique periph-erally curved leaflets that maximize the central orifice areawhen fully open.

The pulsatile flow facility and associated instrumentation(Fig. 2) are basically the same as that used in previousstudies. A computer-controlled homemade stepping motordrive system was used for this experiment. Please refer toprevious reports10 for a detailed description. The pulsatileflow facility used in this study is essentially a cardiac pulseduplicator designed to permit hydrodynamic performanceassessment of both mitral and aortic heart valves. The hous-ing (Fig. 3) containing the mitral and aortic valves is madefrom solid Plexiglas with a thin conical transparent polyure-thane film ventricle simulating the shape and the flexibilitycharacteristics of the natural ventricle (Du Plessis andMarchandl3).

The circulation system is an adaptation of the lumpedcircuit proposed by Westerhof et al.14 with an adjustableWindkessel. An adjustable constant head tank maintainsthe necessary backpressure. Primarily, the system consistsof a model left ventricle made from polyurethane and acomputer-controlled (Toshiba Dynabook486EZ, Tokyo,Japan) stepping motor.10 The test arrangement with thehigh-speed video system used during the experiment isshown in Fig. 2.

Capture of the flow field was carried out using a high-speed (normally 1000 frames/s), high-resolution (1024 ×1024) video system (Phantom V5.0; Vision Research,Stuart, FL, USA). The video capture rate used in this ex-perimental study was nominally set at 1000 frames/s for aduration of about 4s (4GB memory), and captured datawere analyzed frame by frame using the Insight V3.5 pro-gram (TSI, St. Paul, MN, USA). Calculation of mean veloci-ties for a certain time frame was accomplished by applying

167

Fig. 1. The test models used for this study. SJM, St. Jude Medical; JR, Jyros; MIRA, Edwards MIRA

Fig. 2. The pulsatile facility, light sheet arrangement, and operatingconditions. The stroke volume is 64ml with a sinusoidal waveform. Asaline solution was used and the pressure was set at 120/70mmHg at

72bpm. The flow rate was 4 l/min and the tracer was 50-µm polyvinylchloride spheres

168

the moving-average method utilizing the neighboring 20frames. Calculation of the instantaneous turbulent velocitywas then done by finding the difference between the localvelocity and the calculated mean velocity. These calcula-tions were done using the Tecplot (Amtec Engineering,Bellevue, WA, USA) program. Continuous laser light wasprovided using a DMI0-527Nd:YFL laser (Photonics,Bohemia, NY, USA).

Dynamic PIV utilizes continuous tracer images to obtaincontinuous velocity vectors in the flow field. The ability tocapture the entire flow field velocity information in a veryshort time is unique to this method. PIV calculates correla-tions of successive images, which allows the use of a widevariety of tracer materials.

Pressure monitoring was conducted using physiologicalpressure transducers (Statham P-23 series; Gould Instru-ment, Oxnard, CA, USA) and the flow rate was monitored

using a pair of magnetic flow transducers (In Vivo MetricSystems Model K; Healdsburg, CA, USA with a GouldStatham blood flow meter model SP2202).

Because the Reynolds number effect (effect of changesin viscosity) on the velocity profile for the Reynolds numberrange experienced during this study is considered to be verysmall, saline solution was selected as the working fluid.Systolic and diastolic pressures were adjusted to 130 and70mmHg, respectively, with a pump frequency of 72bpmand a stroke volume of 64ml. The average flow rate wasabout 4 l/min at 72bpm. The waveform used was sinusoidal.

Results and discussion

Flow field downstream of the SJM, the JR, the On-X, andthe MIRA valves in the A-A section

The SJM valve installed in the anatomical orientation

Under pulsatile flow conditions, three distinct flow phasesmay be identified, as in previous cases:11 the acceleratingflow phase, the peak flow phase, and the decelerating phase.These phases affect the downstream velocity profile and thepresent comprehensive test program aims at studying thecomplete time history of these velocity profile changes.

Although some of the sequential flow events inside asimulated ventricle using the SJM and the JR valves havealready been reported in previous articles,10 results fromthis new experiment using the SJM and the JR valves aregiven to compare the characteristic nature of the eventsobtained using four bileaflet valves.

Figure 4 shows sequential changes in velocity vector andturbulent stress. At the onset of valve opening, flow throughthe mitral valve was fairly disorganized because of the

Fig. 3. Schematic showing the cross section of the simulated hearthousing and the transparent polyurethane ventricle

Fig. 4. Sequential presentationof the velocity profiles andturbulent stress distributionsobtained using dynamic particleimage velocimetry (PIV) for theSJM valve installed in theanatomical orientation (A-Asection)

169

movement of the leaflets. Flow was directed to the sides,one part toward the aortic valve and the other part towardthe ventricle wall, to form diverse flow patterns because thebileaflet valve opens from the outside first (Fig. 4a, earlyacceleration phase). The flow collided with the ventriclewall and showed high levels of instability with high levels ofturbulent stress, indicated by shading, immediately down-stream of the tested mitral valve. Note that the Reynoldsstress (turbulent stress) shown in this paper is u¢v¢ ratherthan the usual −ρu¢v¢, resulting in rather small values for thestress since the density term is missing. It is also worthpointing out that the turbulent stress level depends on thecalculated average value, because the turbulent velocitycomponent is calculated by subtracting the instantaneousvelocity from the average velocity. Consequently, the tur-bulent velocity component is affected by the way the aver-age velocity is determined. The process adopted in thispaper (the moving average of 20 neighboring frames) tendsto estimate a lower turbulent velocity component than themore commonly adopted window method. It is also worthpointing out that the turbulent stress scale used in thefigures is limited to 0.001 and −0.001. The actual maximumturbulent stress level exceeds this several fold. The inten-tion is to show the area of high stress more vividly ratherthan the value itself, which will be discussed in a futurearticle. Although the maximum stress value is not shown inthis paper, the sample calculated value seemed to be of thesame order of magnitude as the published data and ourprevious experimental data using an LDA system.10 Theobserved velocity vectors in general were short, indicatinga low velocity. Subsequently, as the flow accelerated, thevalve opened fully but the flow at the measured location stillseemed to be strongly diverted. Generally, flow was moreorganized, forming a clear velocity pattern with increasedstability (Fig. 4b, late acceleration phase). The increase in

velocity was also associated with an increase in turbulence.Note that the velocity immediately downstream was stillperipheral in nature. As the central velocity increased, thefigure begins to show the effectiveness of the central orificeof the SJM valve. As a result of the direction of the initialflow generated by the two half-disks, the flow tended toform two circulatory patterns, one located close to the aor-tic valve and other located close the center of the ventricle.This trend gradually weakened but continued until peakflow was reached.

Once the flow reached its peak value (Fig. 4c, peak flowphase), the flow from the central orifice was greater still andthe velocity became more stable with two circulatory pat-terns as described above. The decelerating flow phase wascategorized by diminishing velocity vectors but with a stablecirculatory flow pattern, indicated by the shorter arrows(Fig. 4d, deceleration flow phase).

The JR valve installed in the anatomical orientation

Self-rotation in the JR valve is the most interesting andunique feature of this valve. Indeed, it self-rotates duringnormal operation. The effect of self-rotation on the flowfield can be very complex and further research should beconducted on this subject. In this study, however, the effectof valve orientation is the factor we are focusing on, so a finewire was applied at the pivot location to restrict valve rota-tion. The wire was attached to the upstream side of thevalve with one end anchored to the O-ring that replaces thesuture portion of the valve and other end placed at thepivot. The wire was fine enough to allow smooth valveopening and closing.

Four representative flow phases are also presented forthe JR valve installed in the anatomical orientation in Fig. 5.At the onset of valve opening, the flow through the curved

Fig. 5. Sequential presentationof the velocity profiles andturbulent stress distributionsobtained using dynamic PIV forthe JR valve installed in theanatomical orientation (A-Asection)

170

mitral valve leaflets of the JR valve was mainly downwardbut diverted to the side, as in the SJM case. The dominantflow direction was toward the side of the ventricle wall in astrongly divergent flow (Fig. 5a, early acceleration phase).The velocity vectors were slightly longer at these dominantregions, indicating higher velocities. The degree of instabil-ity was also slightly higher than for the SJM valve. Subse-quently, as the flow accelerated, the flow was observed tobecome strongly divergent, forming two circulatory regionswithin the ventricle (Fig. 5b, late acceleration phase). Theselocations were also correlated with high instability. Notethat the center velocity immediately downstream of thevalve location remained weak, indicating inefficient usageof the central orifice of the JR valve. Because of the initialdirectional flow, flow tended to form two clear circulatoryflows, one located close to the aortic valve and other locatedslightly lower. This trend gradually weakened, but contin-ued until peak flow was reached.

When the flow reached its peak value (Fig. 5c, peak flowphase), the velocity pattern reinforced this twin-circulatoryflow with strong instability indicated by the wider morethicker shaded area. Note that the flow corresponding to thecentral orifice remained very weak at this flow phase. Thedecelerating flow phase was categorized by diminishingvelocity vectors and an unorganized flow pattern, indicatedby the shorter arrows and unclear flow pattern (Fig. 5d,deceleration phase).

The On-X valve installed in the anatomical orientation

Four representative flow phases for the On-X valve areshown in Fig. 6. At the onset of valve opening, the flowthrough the straight mitral valve leaflets of the On-X valvewas mainly downward with a slightly diverted nature,smaller than those in the SJM and the JR valves. The di-

verted flow was limited to the region immediately down-stream of the valve. The dominant flow direction was to-ward the center of the ventricle (Fig. 6a, early accelerationphase). The velocity vectors in general were slightly longerat these dominant regions than for the SJM and the JRvalves, indicating higher velocities. The degree of instabilitywas slightly smaller than for the SJM and the JR valves.Subsequently, as the flow accelerated, the flow was slightlydivergent close to the valve location, forming two weakcirculatory regions within the ventricle (Fig. 6b, late accel-eration phase). This time, the mitral side of the circulationwas weaker than the other circulation close to the aorticvalve. These locations were also correlated with high insta-bility, but at a lower level than that seen in the other twovalves. Note that the velocity immediately downstream ofthe valve location was weakly diverted but with a very highvelocity, indicating efficient usage of the central orifice ofthe On-X valve. This generated two clear circulatory flowsas a result, one located close to the aortic valve and theother located close to the apex region. This trend graduallyweakened, but continued until peak flow was reached.

When the flow reached its peak value (Fig. 6c, peak flowphase), the velocity pattern reinforced this twin-circulatoryflow with thicker shade indicating strong instability. Thedecelerating flow phase was categorized by diminishingvelocity vectors and an unorganized flow pattern (Fig. 6d,deceleration phase).

The MIRA valve installed in the anatomical orientation

Four representative flow phases are also presented for theMIRA valve in Fig. 7. At the onset of valve opening, theflow through the peripherally curved mitral valve leafletsof the MIRA valve was mainly downward with a slightlydiverted nature similar to that of the On-X valve. The

Fig. 6. Sequential presenta-tion of the velocity profilesand turbulent stressdistributions obtained usingdynamic PIV for the On-Xvalve installed in theanatomical orientation (A-Asection)

171

diverted flow was limited to the region immediately down-stream of the valve but was slightly stronger than that seenwith the On-X valve. The dominant flow direction was to-ward the center of the ventricle (Fig. 7a, early accelerationphase). The velocities in general were slightly higher atthese dominant regions than for the SJM and JR valves. Thedegree of instability was similar to that of the On-X valvebut slightly smaller than for the SJM and the JR valves.Subsequently, as the flow accelerated, the flow was slightlydivergent close to the valve location (Fig. 7b, late accelera-tion phase). This time, the two circulatory flow patternswere absent. Note that the strong velocity area was limitedto the region immediately downstream of the valve loca-tion. This trend continued until peak flow was reached.

When the flow reached its peak value (Fig. 7c, peak flowphase), the weakly diverted flow was almost absent, the flowforming strong unified downward velocity vectors. A circu-latory flow pattern close to the aortic valve was generated asa result. Strong instability was limited to these high-velocityareas. The decelerating flow phase was categorized by di-minishing velocity vectors and a very organized flow pattern(Fig. 7d, deceleration phase).

The SJM valve installed in the antianatomical orientation

The valve is rotated 90° to achieve the antianatomical orien-tation. Because the measuring plane is located immediatelydownstream of the central orifice of the three orifices of abileaflet valve when the valve is fully open, we are nowmeasuring flow through the central orifice. In other words,we are examining the effectiveness of the central orifice asa flow passage.

At the onset of valve opening, flow through the mitralvalve was fairly strong; however, the directional flow to theside observed in the anatomical orientation was absent be-

cause the affected area was away from the observed plane(Fig. 8a, early acceleration phase). It is also worth notingthat flow through the central orifice had a relatively largevelocity. Higher jet-like velocities into the relatively slowmoving flow region during this phase resulted in relativelyhigh turbulence (Fig. 8a, early acceleration phase). Subse-quently, as the flow accelerated, it became more organizedand formed a clearer velocity pattern (Fig. 8b, late accelera-tion phase). Here, a single circulatory flow was observedrather than the two circulatory flows of anatomical orien-tation case (Fig. 4b). Again, high-velocity flow into therelatively stagnant region during this phase resulted inrelatively high turbulence (Fig. 8b, late acceleration phase).During the peak flow phase, the velocity pattern becamemore stable with a large circulatory flow pattern (Fig. 8c,peak flow phase). The circulatory flow pattern was morestable than for the anatomical orientation. This observationindicates that the utilization of the central orifice of the SJMvalve is rather good.

It is worth noting that the SJM valve in both orientationsproduced relatively simple circulatory flow patterns com-pared with the JR valve. The straight disks of the SJM valvedid not deflect the flow too much and the flows through thethree orifices seemed to merge to form a single downstreamvelocity profile. This probably indicates that the utilizationof the central orifice of the SJM valve is quite good and thedeflected nature of the flow caused by the inclined half-disks appears to be minimal when the valve is fully open.

The JR valve installed in the antianatomical orientation

When the valve is rotated 90° into the antianatomical orien-tation, a similar general trend is seen to that of the SJMvalve in the antianatomical position, in that it tends to forma single circulatory flow at the end of the acceleration phase

Fig. 7. Sequential presentationof the velocity profiles andturbulent stress distributionsobtained using dynamic PIV forthe MIRA valve installed in theanatomical orientation (A-Asection)

172

(Fig. 9c). At the onset of valve opening, flow through themitral valve was predominantly downward. Velocity vec-tors were similar in direction but shorter than for the SJMvalve. Flow to the side was weak (Fig. 9a, early accelerationphase). Similarly to the SJM valve, the higher velocity flowinto the relatively slow flow region during this phase re-sulted in relatively high turbulence (Fig. 9a, early accelera-tion phase). Subsequently, as the flow accelerated, it formeda circulatory velocity pattern similar to but weaker than thatfor the SJM valve (Fig. 9b, late acceleration phase). At thepeak flow phase, the velocity pattern was a clear, large,circulatory flow pattern (Fig. 9c, peak flow phase).

The On-X valve installed in the antianatomicalorientation

At the onset of valve opening, flow through the mitral valvewas predominantly downward. Velocity vectors were simi-lar to the SJM valve but where wider than, indicating thatutilization of the central orifice is rather good. Flow to theside was absent, partially because the affected area wasaway from the observed plane (Fig. 10a, early accelerationphase). Higher velocity flow into the relatively slow flowregion during this phase resulted in relatively large areas ofturbulence, but at a low level because the jet-like nature of

Fig. 8. Sequential presentationof the velocity profiles andturbulent stress distributionsobtained using dynamic PIV forthe SJM valve installed in theantianatomical orientation (A-Asection)

Fig. 9. Sequential presentationof the velocity profiles andturbulent stress distributionsobtained using dynamic PIV forthe JR valve installed in theantianatomical orientation (A-Asection)

173

Fig. 10. Sequential presentationof the velocity profiles andturbulent stress distributionsobtained using dynamic PIV forthe On-X valve installed in theantianatomical orientation (A-Asection)

the flow was relatively weak for this valve (Fig. 10a, earlyacceleration phase). Subsequently, as the flow accelerated,it became very organized and quickly formed a circulatoryvelocity pattern similar to the other valves, but more accen-tuated than that for the SJM and the JR valves (Fig. 10b,late acceleration phase). The flared inlet shape and the 90°opening angle of the leaflets may have had favorable effectson the flow. Again, large areas of turbulence were observedin this region. At the peak flow phase, the velocity patternwas a very clear, large, single circulatory flow pattern(Fig. 10c, peak flow phase). The decelerating flow phase iscategorized by maintained velocity strength and a veryorganized flow pattern (Fig. 10d, deceleration phase).

The MIRA valve installed in the antianatomicalorientation

It is interesting to see similar results for the MIRA valve likethe SJM valve installed in the antianatomical orientation,even though the leaflet configuration is curved rather thanstraight. At the onset of valve opening, flow through themitral valve was predominantly downward. Velocity vectorswere similar to, yet narrower than, the On-X valve. Again,utilization of the central orifice is rather good. Flow to theside was absent (Fig. 11a, early acceleration phase). Theperipherally curved leaflets of the MIRA valve produced alarge opening area at the central orifice when fully opened.

Fig. 11. Sequential presentationof the velocity profiles andturbulent stress distributionsobtained using dynamic PIV forthe MIRA valve installed in theantianatomical orientation (A-A section)

174

Better utilization of the central orifice of the MIRA valveresults from this design feature (Fig. 11a, early accelerationphase). Subsequently, as the flow accelerated, it was veryorganized and quickly formed a circulatory velocity patternsimilar to, but slightly weaker than, the On-X valve. It was,however, stronger than that for the SJM and the JR valves(Fig. 11b, late acceleration phase). The peripherally curvedleaflets of the MIRA valve also acted as a three-dimensionaldiffuser for the flow through the valve. This feature tendedto make the velocity profile uniform but with a lower veloc-ity. This may be the reason why the observed velocity for theMIRA valve was lower than that for the On-X valve. Again,higher turbulence was observed in the high velocity region.At the peak flow phase, the velocity pattern was a very clear,large, circulatory flow pattern similar to the On-X valve(Fig. 11c, peak flow phase). The decelerating flow phase wascategorized by maintained velocity strength and a very orga-nized flow pattern (Fig. 11d, deceleration phase).

Observation of the flow patterns for the JR valve in bothorientations revealed that utilization of the central orifice ofthe JR valve is rather weak. Because of the curved leafletsof the JR valve, flow tended to be deflected more than wasthe case for the SJM valve, resulting in a more complex anddisturbed flow pattern. The two circulatory flow patternsobserved for the anatomical orientation must result in somecolliding motion of the flow, and the unclear flow patternfor the antianatomical orientation was the result of verydisturbed flow. These factors should have a negative effecton the turbulent stresses, which often result from collidingtype flow. Indeed the turbulent stress during the late accel-eration to peak flow phase showed higher values than forother flow phases.

Observation of the flow patterns for the On-X valve andthe MIRA valve in both orientations revealed that utiliza-tion of the central orifice of both valves is rather good.

Because of the flared inlet shape and 90° opening angle ofthe leaflets of the On-X valve and the peripherally curvedleaflets of the MIRA valve, flow improved, particularlythrough the central orifice, resulting in a higher velocity anda very organized flow pattern. These factors should have apositive effect on the turbulent stresses, which often resultfrom colliding type flow.

Flow field downstream of the SJM, the JR, the On-X, andthe MIRA valves in the B-B section

As shown in the above results, antianatomically installedvalves have a simpler flow field with less turbulent stress inthe vertical measuring plane that passes the centers of boththe aortic and the mitral valves (A-A section). This seems tobe the prime reason for the lower hemolysis found inantianatomically oriented valves.

To draw conclusions from observations made in only onemeasuring plane is misleading because the flow field insidethe ventricle is three dimensional. To capture the character-istic nature of the three-dimensional flow field correctly, weneed a special measuring device such as a three-dimensionalPIV system, the so-called stereo PIV system. This, however,is not a simple matter, because the cost involved in usingsuch a device is high and technical difficulties still existwhen analyzing three-dimensional flow. One can capturethe three-dimensional nature of a flow field by prudentlyselecting the measuring locations. For this purpose, an addi-tional measuring plane (B-B section) was selected.

The SJM, the JR, the On-X, and the MIRA valves installedin the antianatomical orientation

Typical velocity and turbulent stress changes for the fourvalves are shown in Figs. 12–15. At the onset of valve open-

Fig. 12. Sequential presentationof the velocity profiles andturbulent stress distributionsobtained using dynamic PIV forthe SJM valve installed in theantianatomical orientation (B-Bsection)

175

ing, flow through the mitral valve was fairly disorganizedbecause of the opening movement of the leaflets. This disor-ganized flow, however, did not initially affect most of theventricle and was limited to locations very close to thevalve. All valve leaflets open from the outside first andmove downward. Because of this, outward-directed flowswere shown at the top of the figures for all valves. Flowsthrough the SJM and the JR valves (Fig. 12a, 13a, earlyacceleration phase) showed strong peripheral flow. Flowsthrough the On-X and the MIRA valves (Fig. 14a, 15a, early

acceleration phase) were diverted but were relativelyweaker than the previous two valves. The observed velocityvectors in the general area were relatively short, indicatinglower velocities throughout most of the ventricle, except forthe jet-like flow areas. Subsequently, as the flow acceleratedand the valve angle changed, the flow was directed more tothe side for the SJM valve (Fig. 12b, late acceleration phase)and the flow through the JR valve was directed even moreto the side and reached further downstream than that forthe SJM valve (Fig. 13b, late acceleration phase). Note that

Fig. 13. Sequential presentationof the velocity profiles andturbulent stress distributionsobtained using dynamic PIV forthe JR valve installed in theantianatomical orientation (B-Bsection)

Fig. 14. Sequential presentationof the velocity profiles andturbulent stress distributionsobtained using dynamic PIV forthe On-X valve installed in theantianatomical orientation (B-Bsection)

176

the strong peripheral flow through the JR valve generatedflow along the ventricle wall and an upward flow at thecenter of the ventricle. The flow through the On-X and theMIRA valves was directed to the side but to a lesser degree.It seemed to merge gradually (Fig. 14b, 15b, late accelera-tion phase). The highly directed jet-like flows of the SJMand the JR valves were associated with higher turbulencestresses (Fig. 12b, 13b, late acceleration phase). At this flowphase, the flow though the central orifice of the SJM valvebegan to dominate the downward flow close to the valve butwas still peripheral in nature (Fig. 12a, late accelerationphase), whereas the JR valve continued to demonstrate aflow with a significant peripheral nature (Fig. 13b, lateacceleration phase) resulting in an upward central flowclose to the valve. This is opposite to the SJM case. Exceptfor the JR valve, the other three valves began to generatetwo circulatory flows, one at the very top end and one at thecorners of the simulated ventricle.

Once the flow reached its peak value (Fig. 12c–15c, peakflow phase), characteristic differences in the flow field wereapparent. The velocity pattern for the SJM valve becamemore stable with a central downward velocity close to thevalve, because the valve was fully open at this time (Fig. 12c,peak flow phase). The velocity pattern for the JR valve (Fig.13c, peak flow phase) also became more stable with a twincirculatory flow pattern with peripheral downward velocityand weak central flow because the flow was directed to theside by the curvature of the leaflets. The merging nature ofthe On-X and the MIRA valve was apparent. Becauseof the flared inlet shape and 90° opening angle of the leafletsof the On-X valve and of the peripherally curved leaflets ofthe MIRA valve, flow tended to be improved, particularlythrough the central orifice. The strong downward flow ofthe On-X valve underwent no divergence because the flared

Fig. 15. Sequential presentationof the velocity profiles andturbulent stress distributionsobtained using dynamic PIV forthe MIRA valve installed in theantianatomical orientation (B-Bsection)

shape of the inlet and the opening angle of the leaflets (90°)helped to achieve this type of flow (Fig. 14c, peak flowphase). The MIRA valve also produced relatively strongdownward flow. Because the peripherally curved leaflets ofthe MIRA valve offered a large central orifice area whenfully open, this feature helped to produce relatively strongdownward flow immediately downstream of the mitralvalve. This curved leaflet, however, also helped divert theflow through peripheral orifices, showing diverted flowclose to the center of the ventricle (Fig. 15c, peak flowphase). The decelerating flow phase showed continuation ofthe peak flow phase characteristics with diminishing veloc-ity vectors (Fig. 12–15d, deceleration phase). Again, thehighly jet-1ike flows observed during the acceleration phasewere associated with high turbulent stresses.

To summarize, the JR valve showed symmetrical twincirculations caused by the divergent nature of the flow gen-erated by the two curved half-disks; the JR valve with pe-ripheral downward circulation contrasts with the SJM valvewith central downward circulation close to the valve. Thetwo newer valves, the On-X and the MIRA valves, showedweakly diverted flow initially, but strong central flow at thepeak flow phase.

Discussion

Based on the experimental results obtained for bileafletvalves using the dynamic PIV system, general trends in theflow pattern have become clear. Because of the basic de-sign, all bileaflet valves opened from the outside first. Thisfeature generates diverted flow at the initial stage. The de-gree of the influence of this feature on the valve perfor-mance was rather surprising. The straight-leaflet SJM valve

177

and the curved-leaflet JR valve diverted the flow more thanthe newer-generation valves (Figs. 4–7). Of the two oldervalves, the curved-leaflet JR valve diverted flow more thanthe SJM valve because of its curved leaflet, and usage of thecentral orifice was better for the SJM valve than for the JRvalve. Generally, both valves behaved in a similar fashionup to the peak flow phase when the valve was installedanatomically (Fig. 4b,c and Fig. 5b,c). The JR valve gener-ated less-clear flow patterns at the deceleration flow phase(Fig. 4d and Fig. 5d). The newer straight-leaflet On-X valveand the peripherally curved-leaflet MIRA valve did notdivert the flow as much as the SJM and the JR valves didand generated stronger and more unified flows (Figs. 6, 7).As a result, both newer valves generated simpler flow pat-terns than older the valves within the ventricle when thevalve was installed anatomically.

All valves showed clearer and simpler circulatory flowpatterns when installed in the antianatomical orientation(Figs. 8–11). Differences can be found in the degree ofutilization of the central orifice. The SJM valve showedclearer and stronger flow patterns than the JR valve (Fig.8b,c and Fig. 9b,c). The newer two valves, the On-X and theMIRA valves, showed even clearer and stronger circulatorypatterns in the same orientation (Fig. 10b,c and Fig. 11b,c).It was surprising to find that small differences in hingelocation, leaflet configuration, and overall design betweenthese valves could ultimately generate such noticeable dif-ferences. However, it is difficult to rank the valves fromthese findings because the study is qualitative in nature andthe observed velocity field was confined to the vertical planethat passed the center of the aortic and the mitral valves(A-A section).

For flow in the B-B section, the directed and highly jet-like flows observed during the acceleration phase are asso-ciated with high turbulent stresses (Fig. 12a,b–15a,b). Thehighly deflected flow of the JR valve, in particular, main-tained high turbulent stress levels until the end of the peakflow phase (Fig. 13c).

The highly complicated and highly turbulent flow pat-terns experienced during the early accelerating phase withthe valve in the anatomical orientation seem to correlatewell with the clinical experiences reported by Aoyagi atKurume Medical School that valve orientation seems to bea factor that affects the level of hemolysis. Other results4,5

also suggest that antianatomical orientation is the optimumorientation for mitral valve replacements. If higher turbu-lence is the cause of hemolysis, the observations in this invitro study using dynamic PIV show that the antianatomicalorientation of the SJM valve produced less turbulence andis consistent with the clinical findings. In this context, thenewer two valves, the On-X and the MIRA, showed morecentral flows, which is expected to be better clinically be-cause of its improved and simpler flow patterns. One caneasily speculate that this kind of flow with lower turbulentstresses will result in less hemolysis and fewer emboliccomplications.

It is worth pointing out that the results of the On-X andthe MIRA valves seem to show some discrepancy betweenthe results for the A-A section and the results for the B-B

section: the magnitude of the velocity was lower for the B-B section than for the A-A section. From the experimentalresults, both valves showed excellent central orifice flowutilization. Therefore, one might expect a much higher ve-locity for the B-B section. Remember that the flow insidethe ventricle is three dimensional. Flow from the mitralvalve is not always downward but can be inclined due to thecirculatory nature of the main flow. Flow in the B-B sectionwas only presenting one section of this constantly changingthree-dimensional flow. In fact, the B-B section was 10°offset from the mitral valve center axis. We were looking atthe velocity component along the B-B section. Neverthe-less, both valves showed excellent central flow properties.

Conclusion

Based on the experimental results obtained for bileafletvalves, the general flow trends have become clear. The dy-namic PIV system has been successfully applied to examinethe flow fields of prosthetic heart valves. Antianatomicallyinstalled bileaflet valves generate a simpler circulatory flowpattern with lower stresses in the A-A section, whereasthese valves generate, characteristically, a very differentflow field immediately downstream in respect of the mitralvalve in the B-B section. The newer On-X and MIRAvalves show a cleaner and stronger central flow than theolder designs. In both orientations, the JR valve generatedlarge areas of high turbulent stress.

Acknowledgments This work was partly supported by a Grant-in Aidfor Scientific Research from the Japan Society for the Promotion ofScience (Grant C-17591491).

References

1. Hufnagel CA, Harvey WP, Rabil PJ, McDermott TF. Surgicalcorrection of aortic insufficiency. Surgery 1954;35:673–683

2. Starr A, Edwards ML, McCord CW, Griswold HE. Aortic replace-ment clinical experience with a semirigid ball-valve prosthesis.Circulation 1963;27:779–783

3. Aoyagi N, Tanaka I, Nishi Y, Yamashita M, Oryouji A, Hara T,Kosuga K, Ooishi K. Long-term results of MRV using theSJM valve. J Jpn Thorac Cardiovasc Surg 1991;39:1126–1130 (inJapanese)

4. Baudet EM, Oca CC, Roques XF, Laborde MN, Hafez AS, CollotMA, Ghidoni IM. A 5.5 year experience with the St. Jude Medicalcardiac valve prosthesis. Early and late results of 737 valve replace-ments in 671 patients. J Thorac Cardiovasc Surg 1985;90:137–144

5. Duveau D, Michaud JL, Despins P, Patra P, Train M, Dupon H,Rozo L, Carlier R. Mitral valve replacement with the St. JudeMedical prosthesis: 242 cases with clinical results and an evaluationof prosthesis positioning. In: DeBakey ME (ed) Advances incardiac valves, clinical perspectives (Proceedings of the Third In-ternational Symposium on the St. Jude Valve, November 1982,Scottsdale, Arizona), New York: Yorke Medical Books, 1983;183–190

6. Akutsu T, Higuchi D. Effect of mechanical prosthetic heart valveorientation on the flow field inside a simulated ventricle: compari-son between St. Jude Medical valve and Medtronic-Hall valve.J Artif Organs 1999;2:39–45

178

7. Akutsu T, Higuchi D. Effect of the mechanical prosthetic mono-and bi-leaflet heart valve orientation on the flow field inside asimulated ventricle. J Artif Organs 2000;3:126–135

8. Akutsu T, Higuchi D. Flow analysis of bi-leaflet mechanical pros-thetic heart valves using laser Doppler anemometry: effect of thevalve design and installed orientation on the flow inside a simu-lated left ventricle. J Artif Organs 2001;4:113–125

9. Akutsu T, Masuda T. Three-dimensional flow analysis of amechanical bi-leaflet mitral prostheses. J Artif Organs 2003;6:112–123

10. Akutsu T, Fukuda T. Time-resolved particle image velocimetryand laser Doppler anemometry study of the turbulent flow field of

bileaflet mechanical mitral prostheses. J Artif Organs 2005; 8:171–183

11. Modi VJ, Bishop WF, Akutsu T. Unsteady fluid dynamics of threecontemporary heart valves using a two-component LDA system.Artif Organs 1991;14:103–107

12. Akutsu T, Modi VJ. Unsteady fluid dynamics of several mechani-cal prosthetic heart valves using a two-component laser Doppleranemometer system. Artif Organs 1997;21:1110–1120

13. Du Plessis LA, Marchand P. The anatomy of the mitral valve andits associated structures. Thorax 1964;19:221–227

14. Westerhof N, Elzinga G, Sipkema P. An artificial arterial systemfor the pumping heart. J Appl Physiol 1971;31:776–781