Retina Today - Fluidics Behavior During Vitrectomy (November/December 2010)

[摘要] Advancements in vitreoretinal instrumentationhave expanded the selection of surgicalparameters during vitrectomy procedures.With numerous combinations of system settings,the task of optimizing aspiration flow rates whilemaintaining precise tissue cutting can be challenging. Tosafely utilize new-generation instruments, surgeonsshould understand fluidic behavior during vitrectomy.Poiseuille’s law can be used to describe flow rate througha vitrectomy probe (see Equation 1).1 The variable R is theinner lumen radius, P is the pressure drop across the cutter,μ is the dynamic viscosity of aspirated fluid, and l is thelength of cylindrical tubing through which flow is measured.Poiseuille’s equation is valid for laminar flow of an incompressiblefluid such as balanced salt solution through theprobe when the cutter is disabled. Under realistic surgicalconditions, the heterogeneous composition of vitreouscomplicates fluid dynamics. Although the nature of vitreousbehavior during aspiration is complex, the relationshipbetween flow rate, probe geometry, applied vacuum pressure,and viscosity of aspirated fluid still provide insight intofluid dynamics during vitrectomy.The evolution of probe design continues to expandthe variety of surgical parameters governing flow rate.Vitreous surgery was originally performed using20-gauge pneumatic, spring-return cutters.2 Designimprovements, such as reduced outer probe diameter,led to 23- and 25-gauge transconjunctival sutureless vitrectomyinstruments that reduced sclerotomy inflammation.3,4 Although smaller-diameter instrumentsimproved postoperative patient comfort, surgeonsexperienced reduced flow rates during vitrectomy.4,5Spring-return mechanisms were engineered to achievemaximum cut speeds of 2500 cpm and demonstratedincreased vitreous flow rates at higher cut rates.6,7 Now,new-generation cutters operate using a dual pneumaticdrive rather than spring-return mechanism which allowsthe cutter to achieve even higher cut rates of 5000 cpmand to modulate duty cycle or percent of port opentime. The dual pneumatic probe technology has introducedhigher cut rate capabilities and a variety of dutycycle settings that may affect surgical technique andflow behavior during surgery.As previous studies evaluated flow performance ofspring-return cutters at the maximum cut rate of 2500 cpm,it is important to evaluate the performance of new generationcutters at 5000 cpm and at various duty cycle modes.In this study, two laboratory bench tests were performed todetermine the effects of high speed cutting on flow ratesduring vitrectomy. The purpose of the first test was tomeasure flow rates in pure porcine vitreous at various cutrates. The purpose of the second test was to measure thepeak velocity and peak acceleration of clear fluid flow intothe probe port.EXPERIMENT 1: PORCINE VITREOUS FLOWPurposeThe purpose of the first experiment was to investigateporcine vitreous flow performance of 23-gaugeUltraVit probes (Alcon Laboratories, Inc., Fort Worth, TX) at various duty cycle modes, cut speeds, and vacuumpressures.MethodsA Constellation Vision System (Alcon Laboratories,Inc.) was used to test six dual-pneumatic UltraVitprobes. Three duty cycle modes were evaluated withthe vitrectomy system: core (maximum port opentime), 50% port open, and shave (minimum port opentime). Each probe was tested at vacuum settings of 250mm Hg, 450 mm Hg, and 650 mm Hg and at cut ratesof 500 cpm to 5000 cpm.Porcine eyes from Sierra for Medical Science (Whittier,CA) were tested within 48 hours of slaughter. Each eyewas kept refrigerated until immediately before testing.Prior to flow rate testing, vitreous density was measuredby weighing the mass of 3 mL to 4 mL of extracted vitreousin a previously weighed graduated cylinder. Densitymeasurements were repeated for 10 eyes and averaged.A Styrofoam cube with 3-inch edges was used as amount for the porcine eyes. A spherical recess 1 inch indiameter was created on the top of the cube to securethe eye. Multiple T-pins were used to fasten one eye inthe spherical recess with the cornea facing upward.Using a scalpel, an initial incision was made through thepars plana, 3 mm from the limbus. The initial incisionwas extended in an annulus around the cornea and theanterior chamber of the eye was removed. Any vitreousattachments to the posterior capsule were carefullydetached using the scalpel. The Styrofoam block withthe porcine eye secured into it was placed on an electronicbalance (model EK-600i, A&D Engineering, Inc.,San Jose, CA). The vitrectomy probe to be tested wassecured vertically over the balance using a clamp andlab stand. The vitrectomy probe port was inserted intothe center of the vitreous bolus. One porcine eye wasused per test.The electronic balance was connected directly to apersonal computer using a RS232 cable (Figure 1). ALabVIEW program (National Instruments, Austin, TX)was written to measure mass with time. Volume flowrates were calculated using previously obtained vitreousdensity measurements. The LabVIEW program dividedthe mass flow by the average porcine vitreous density.Volume flow data was exported from the program as aMicrosoft Excel file. Prior to each test, the balance waszeroed and the appropriate Constellation Vision Systemsettings were chosen. LabVIEW data capture was activatedafter the vitreous cutter was engaged and vitreousflow stabilized (after ~5 seconds).Vitreous flow rates were compared with previouslypresented flow rates of balanced salt solution (BSS irrigatingsolution, Alcon Laboratories, Inc.).8 For BSS flowtests, key differences in experimental set up included aclosed-globe, rigid model eye; applied infusion pressureof 30 mm Hg; and a flow meter instead of an electricbalance.For each test, the vitreous flow rate was calculated asthe average of data points after the initial time to stabilizeand before all vitreous was removed from the eye.All data are presented as mean ± standard deviationunless otherwise noted. Flow rates were comparedusing Student’s t-test with statistical significance ofP<.05. Trend lines were examined by linear regression and correlation coefficients were calculated usingMicrosoft Excel.ResultsFigures 2 to 4 depict vitreous flow rates through 23-gauge UltraVit probes at the 3 duty cycle modes of theConstellation Vision System. Each data point on thegraph illustrates the average flow rate of six probes anderror bars represent 95% confidence intervals.Core Duty Cycle. The core duty cycle mode generatedrelatively stable vitreous flow rates across all cut ratesand vacuums (R2≤.2). For all vacuum pressures, therewas no significant difference between flow at the minimumand maximum tested cut speeds of 500 cpm and5000 cpm. In the core mode, flow at low cut rates weresignificantly higher than flow at low cut rates in of the50% and shave modes at equivalent settings (P<.05). Amaximum flow rate of 3.69 ≤ 0.92 cc/min was achievedat 650 mm Hg vacuum and 5000 cpm.50% Duty Cycle. In the 50% duty cycle mode, flowincreased with increasing cut rate. For all vacuum pressures,flow at 5000 cpm was significantly higher thanflow rates at 500 cpm (P<.05). At 50% duty cycle,450 and 650 vacuum, low cut rate (500 cpm), flow wassignificantly higher than shave (P<.05). Minimum vacuumand cut rate settings (250 mm Hg vacuum and500 cpm) generated flow rates of 0.20 ± 0.19 cc/min.Maximum flow of 4.12 ± 0.81 cc/min was achieved atmaximum vacuum and cut rate settings (650 mm Hgand 5000 cpm). Maximum flow was statistically similarto flow in the core and shave duty cycle at equivalentsettings.Shave Duty Cycle. Flow in the shave mode increasedwith increasing cut rate. For all vacuum settings, flowrate at 5000 cpm was significantly higher than 500 cpm(P<.05). Flow in the shave mode increased more rapidlythan flow of 50% duty cycle. For 50% duty cycle, theslope of trend lines for 250 mm Hg, 450 mm Hg, and650 mm Hg vacuum were 0.1µL/cut, 0.2 µL/cut, and0.5 µL/cut, respectively (all R2≥.7). For shave duty cycle,slope of trend lines for 250 mm Hg, 450 mm Hg, and650 mm Hg vacuum were 0.2 µL/cut, 0.5 µL/cut, and0.9 µL/cut, respectively (all R2≥.9). At maximum vacuumand cut rate (650 mm Hg and 5000 cpm), flow wasstatistically similar to flow in the core mode at equivalentsettings. For the shave mode, flow rates rangedfrom 0.05 ± 0.12 cc/min (at 250 mm Hg vacuum and500 cpm) to a maximum flow rate of 4.82 ± 0.77 cc/min(at 650 mm Hg vacuum and 5000 cpm).Vitreous flow of 23-gauge UltraVit probes in the coremode was not strongly dependent on cut rate; flowremained relatively constant for all cut rates. Vitreousflow in the 50% and shave duty cycle modes was influencedby cut rate; flow increased with increasing cutrate. The rate of increase for the shave mode was higherthan that of the 50% duty cycle. At the maximum vacuumand cut-rate settings, flow rates were statisticallysimilar for all duty cycle modes.DISCUSSIONThis study demonstrated the effects of cut rate,duty cycle, and vacuum on vitreous flow rates. Figure5 illustrates the average duty cycle of six 23-gaugeUltraVit cutters.9 Figures 6 to 8 compare the findingsof this study to the average clear fluid flow rates of six23-gauge UltraVit® cutters operating at 650 mm Hgvacuum and 500-5000 cpm.8 Error bars illustrate a confidence interval of 95%.Core Duty Cycle. At 500 cpm, the core duty cycle ofthe Constellation Vision System generated a maximumof 87.34 ± 0.86% duty cycle that decreased to 53.63± 1.82% at 5000 cpm.9 It was reported that clear fluidflow generated a maximum flow rate of 20.8 ± 0.9cc/min at low cut rates (500 cpm) that decreased to14.45 ± 1.32 cc/min at higher cut rates (5000 cpm).8Both clear fluid flow rates and duty cycle decreasedwith increasing cut rate in the core mode. Since aqueousflow and core duty cycle are dependent on cut rate,we might expect to see a decrease in vitreous flow ascut rate increases; however, vitreous exhibited constantflow at all cut rates (Figure 7). High cut rates fragmentvitreous into smaller segments resulting in reduced flowobstructions.10 Conversely, low cut rates generate lessvitreous fragmentation and more flow obstruction dueto the reduced cut frequency. At low cut rates, theeffects of high duty cycle may have offset reduced vitreousflow (from decreased fragmentation) resulting inconstant flow rates for all cut speeds. In other words,constant vitreous flow in the core duty cycle mode suggeststhat there is an increased resistance to flow at lowcut rates.50% Duty Cycle. In the 50/50 mode, clear fluid flowrates reflect the constant duty cycle and maintain relativelyconstant flow ranging from 13.58 ± 0.37 to 14.3 ±0.83 cc/min at 500 cpm to 5000 cpm, 650 mm Hg vacuum(Figure 9).8,9 During 50% duty cycle, the port opentime proportionally decreases with increasing cut ratein order to maintain a constant 50% duty cycle for allcut rates. The duration of port open time is longer atlower cut rates than higher cut rates, but the ratio ofport open time to cut cycle time remains 50% for all cutrates. We might expect vitreous flow to maintain constantflow similar to clear fluid behavior, but vitreousflow rates increased from 1.83 ± 0.68 (500 cpm) to4.12 ± 0.81 cc/min (5000 cpm), suggesting that thedecreased port open time at high cut rates is associatedwith reduced aspirated volume, increased vitreous fragmentation,and reduced flow resistance.Shave Duty Cycle. The shave duty cycle maintained aconstant port open time for all cut rates. Recall thatduty cycle was defined as the percent ratio of portopen time to cut cycle time. With increasing cut rate,the port open time was held constant while cut cycletime decreased resulting in a duty cycle that climbedfrom 12.80 ± 1.05% to 52.15 ± 1.06%.9 Aspiration duringone cut cycle can be described as a cut “bite” wherevacuum pressure draws vitreous into the open port andthe guillotine cutter fragments and aspirates a discretevolume of vitreous. Because the port open time isequivalent for all cut rates, the shave duty cycle behaviorsuggests the discrete volume of vitreous aspiratedper cut bite is also equivalent for all cut rates. If the frequency of these cut bites increase at high cut rates,then the amount of aspirated vitreous fragments alsoincreases with high cut rates. Thus, increased vitreousflow may be associated with high cut rates in the shaveduty cycle as the probe engages vitreous more frequentlythan low cut rates. This study demonstrated anincrease in vitreous flow at 650 mm Hg vacuum from0.59 ± 0.76 cc/min at 500 cpm to 4.82 ± 0.77 cc/min at5000 cpm (Figure 8). Although the shave duty cyclemaintained an equivalent cut bite size for all cut rates,increased vitreous flow and increased intake of cut bites(associated with high cut rates) suggests that vitreousaspiration is more effective at high cut rates.It is important to address the difference in test methodsbetween vitreous and clear fluid flow data. Vitreousflow rates were captured using an electronic balance tomeasure open-sky vitrectomy rates while aqueous flowdata was collected using a BSS infusion flow meter in arigid, closed-system eye model. A closed system can bepressurized—an open system cannot. Although thepressure response differs between the two systems,both experimental methods generate measurementsthat help characterize flow rates during vitrectomy.Another important difference to note is the compositionof aspirated materials. The heterogenous propertiesof vitreous drastically differ from BSS irrigating solution.Vitreous is a semisolid mixture of water, collagenfibers, and hyaluronic acid. The complex nature of vitreousis unpredictable during vitrectomy and can sometimesgenerate a high standard deviation due to vitreousflow obstructions. Clear fluid flow tests wererepeatable, predictable, and more accurately reflectPoiseuille’s law of flow. Despite differences in the compositionof aspirated fluids, a comparison between clearfluid and vitreous helps characterize vitrectomy flowbehavior.CONCLUSIONIn summary, the 50% duty cycle mode generated anincrease in vitreous flow at high cut rates suggestingthat high cut rates generate less flow resistance.Constant vitreous flow in the core duty cycle modeacross all cut rates implies that the high flow effects ofthe core duty cycle are offset by the increased resistanceto vitreous flow associated with larger bite sizes.In the shave mode, vitreous flow increased at high cutrates as the frequency of constant-volume bitesincreased with high cut rates. Overall, the operation of23-gauge UltraVit probes at high cut rates yieldedreduced resistance to flow at high cut rates and moreeffective aspiration of the vitreous body than lowercut rates.EXPERIMENT 2: MEASUREMENT OF PEAKFLOW VELOCITY AND PEAK ACCELERATIONPurposeThe purpose of the second experiment was to determinethe effects of cut rate on peak flow velocity andpeak acceleration of clear fluid and microbeads through25+-gauge dual-pneumatic cutters.MethodsA Constellation Vision System was used to test anUltraVit probe in a closed-system, test chamber. Flowwas evaluated at 5000 cpm, 650 mm Hg vacuum and at2500 cpm, 575 mm Hg vacuum to achieve equivalentflow rates of approximately 7 cc/min. All testing wasperformed under 30 mm Hg infusion pressure and 50%duty cycle.The test chamber was designed to simulate theclosed system of a compliant human eye while allowingfor visualization of fluid entering the probe port. Thetest chamber was a 1x1x2-inch open box made of clearacrylic. A square 1.5-inch silicone sheet of 0.005-inchthickness with a reading of 55A durometer was used tocover and seal the box (McMaster-Carr, Santa FeSprings, CA). The silicone sheet was secured over thetop of the acrylic test chamber using a metal clampingmechanism (Figure 12). The flexible properties of thesilicone cover simulated the dampening effect of compliantsclera. The test chamber was filled with aqueoussolution and 75 μm microbeads were injected to helpvisualize fluid motion during aspiration.To replicate fluid input and output during vitrectomy,the test chamber was modified to incorporate to twotrocar cannula entry ports—one for BSS fluid infusionand a second port for the vitrectomy probe. A 0.02-inchdiameter hole was drilled in the side wall of the testchamber. A 25+-gauge trocar cannula (AlconLaboratories Inc.) was secured in the drilled hole usingLoctite 4014 (Henkel, Berkeley, CA). A 25+ infusion cannula(Alcon Laboratories Inc.) was inserted into the25+-gauge trocar cannula and supplied BSS infusion tothe test chamber. A second port was created by insertinga 25+-gauge trocar cannula into the silicone coverusing the 25+-gauge trocar blade. The vitrectomy probeto be tested was secured vertically over the test chamberusing a three-screw adjustable ring mount (EdmundOptics, Barrington, NJ). The probe tip was inserted intothe test chamber through the second entry port of thesilicone cover and centered in the test chamber.Two different camera perspectives were required for3-D motion analysis. Two Phantom high-speed cameras(Vision Research, Wayne, NJ) were oriented orthogonallyaround the test chamber. Each camera was secured toan Edmund Optics X-Y-Z axis metric stage for precise position adjustment. A Computar 10X magnificationlens was fastened to each camera and focused on thevitrectomy probe port. Two fluorescent backlight panelsilluminated the test chamber.Both cameras were tethered to one another in a masterand slave configuration to synchronize recording.An external pickle switch was connected to both camerasin order to manually trigger video recording at thesame time. Both cameras were directly connected to apersonal computer with an Ethernet cable. High-speedvideo was recorded at 1000 frames per second. Thevideo resolution for each camera was 640x480 pixels.Brightness, contrast, and exposure time was adjusted toclearly define microbeads around the probe port.Prior to high-speed recording, the appropriateConstellation Vision System settings were chosen. Aninfusion pressure of 30 mm Hg was applied to the testchamber. Video capture was triggered with the pickleswitch after the cutter was engaged and fluid flowstabilized. Recording was stopped after approximately20 seconds of recording and each camera saved anaudio video interleave (AVI) file.Prior to video analysis, 3-D measurements of TEMAMotion Software (Photo-Sonics Inc, Burbank, CA) werecalibrated. Relative camera orientations were manuallyentered into the program. Six points on a calibrationtarget visible from both cameras were identified. Thesoftware translated 2-D measurements from each camerainto 3-D motion. During analysis, AVI video files ofeach camera perspective were imported into the TEMAsoftware program. Each file was synchronized andviewed simultaneously in the analysis window.Individual microbeads were manually selected fromeach perspective and the software tracked microbeadmotion into the vitrectomy probe port frame-by-frame.Velocity and acceleration data was exported from theprogram as a Microsoft Excel file. Peak velocity andpeak acceleration values of five tracked microbeadswere averaged for each cut rate. All data are presentedas mean ± standard deviation. Velocities and accelerationswere compared using a two-tailed, Student’s t-testwith statistical significance of P<.05.RESULTSFigure 10 and 11 depict peak velocity and peak accelerationgraphs of microbeads entering a 25+-gaugeUltraVit cutter at 2500 cpm and 5000 cpm, 50% dutycycle. Each data point on Figure 10 illustrates the averagepeak velocity of 5 microbeads during aspirationinto the vitrectomy probe port. Each data point inFigure 11 illustrates the average peak acceleration offive microbeads. The time axis of Figures 10 and 11 canrelate to microbead distance from the port and rangefrom 0.1 to 0.25 for clear illustration of velocity andacceleration profiles. Error bars represent 95% confidenceintervals.At 2500 cpm, the average peak velocity ofmicrobeads increased as the beads approached theport. Just immediately before entering the probeport, the average maximum peak velocity was 52.98± 11.89 mm/s. With 5000 cpm cut rate and vacuumpressure of 650 mm Hg, the average peak velocity alsoincreased with time. The average maximum peak velocitybefore aspiration into the probe port was 42.06± 10.51 mm/s. Statistical evidence suggests that peakvelocities during aspiration at 2500 cpm were significantlyhigher than 5000 cpm (P<.05). On average, peakvelocities occurring at 2500 cpm were approximately24.28% faster than 5000 cpm peak velocities.For 2500 and 5000 cpm cut rates, the average peakacceleration increased as the beads approached the port.Maximum peak acceleration of microbeads reached5442.46 ± 1479.84 mm/s2 prior to entering the probeport. At 5000 cpm, peak accelerations reached an average value of 2909.88 ± 538.82 mm/s2. Statistical evidence suggeststhat aspiration at 2500 cpm generated significantlyhigher average peak accelerations than beads aspirated at5000 cpm (P<.05). Average acceleration for 2500 cpm cutspeeds generated approximately 34.80% higher peak accelerationvalues than 5000 cpm cut speeds.Aspiration with the 25+-gauge UltraVit probe demonstratedan increase in velocity and accelerationas beads approached the port for both 2500 and 5000cpm. The 2500 cpm cut rate generated faster peak flowvelocities and higher peak accelerations than 5000 cpm.DISCUSSIONThis study demonstrated the effects of cut rateon peak velocity and peak acceleration of fluid in aclosed-system test chamber. We evaluated the motionof 5 microbeads into an UltraVit 25+-gauge probeat 2500 cpm, 575 mm Hg vacuum and 5000 cpm,650 mm Hg vacuum. All testing was performed under50% duty cycle and 30 mm Hg infusion.During a cut cycle, the velocity of fluid increases anddecreases according to port open and close times. Whenthe port is fully open, fluid reaches peak velocity. Thisvelocity can be directly related to flow using a basicprinciple of volume flow rate: Flow rate = orifice area xfluid velocity. Reduced peak velocity at high cut ratestranslates into reduced clear fluid flow at high cut rates.Under surgical conditions, reduced peak flow velocitiesand flow rates of BSS suggests that operation at 5000cpm may reduce pulsatile disturbance of tissue and fluidin the eye. The data shows fluid and microbeads aspiratedat 2500 cpm approached the probe port with 24.28%faster peak velocity than 5000 cpm.According to Newton’s second law relating accelerationand inertial mass to force (Force = mass x acceleration),lower acceleration leads to reduced forces on fluidand tissue around the port. As high cut rate yields loweraspiration forces at the probe port, surgeons can preciselyoperate closer to the surface of tissue and membranes,reduce vitreoretinal traction, and reduce the incidence ofiatrogenic retinal breaks. Data analysis of fluid accelerationdemonstrated that higher cut rates generated loweraccelerations and associated forces, as peak accelerationsat 5000 cpm were 34.80% lower than 2500 cpm.Limitations of this study included a test chamberdesign that did not perfectly replicate the conditions ina human eye. The rigid test chamber does not exhibitthe same flexibility and dampening as scleral tissue.Although reported measurements using the rigid modelmay not exactly represent testing in an eye, a siliconewall of the test chamber was used to replicate somedegree of tissue compliance.CONCLUSIONSThe first test measured flow rates of porcine vitreousat various cut rates and duty cycle settings. Results indicatedhigher vitreous flow at high cut rates suggestingreduced flow resistance and more efficient vitreousaspiration. The second test measured peak flow velocitiesand peak accelerations of microbeads and BSS fluidentering the vitrectomy probe port. It was determinedthat 2500 cpm generated faster peak velocities andhigher peak accelerations than 5000 cpm. This suggeststhat high cut rate reduces BSS fluid flow and hence,reduces aspiration forces, pulsatile vitreoretinal traction,and iatrogenic tears while allowing for more precisecontrol of fluid surrounding the port.Overall, high cut rates of dual pneumatic probes generatedclinically effective vitreous flow and reducedaspiration forces. By investigating the effects of highcutting rates on vitreous and clear fluid, this study characterizesflow behavior of advanced probe technologyand helps surgeons optimize system settings to improvethe efficiency and precision of surgical technique.Dina Abulon, MS; and David Buboltz, BS, MBA, are withAlcon Laboratories in Fort Worth, TX.Steve Charles, MD, is Founder of the CharlesRetina Institute in Memphis, TN, and is aClinical Professor in the Department ofOphthalmology at the University of TennesseeCollege of Medicine. He is a Retina TodayEditorial Board member and states that he is a consultantfor Alcon Laboratories, Inc. Dr. Charles can be reachedvia e-mail at [e-mail] BR. Fundamentals of Fluid Mechanics. New York: John Wiley & Sons, Inc. 2002. O’Malley C, Heintz R. Vitrectomy with an alternative instrument system. Ann Ophthalmol.1975;585-588, 591-594. Eckardt, C. Transconjunctival sutureless 23-gauge vitrectomy. Retina. 2005;25:208-211. Didi FI, Moisseiev J. Sutureless vitrectomy: evolution and current practices. Br JOphthalmol. 2010 Aug 23. [Epub ahead of print) Hubschman JP, Gupta A, Bourla D. Culjat M, Schwartz SD. 20-, 23-, and 25-gauge vitreouscutters: performance and characteristics evaluation. Retina. 2008;249-257. Magalhaes O, Chong L, DeBoer C, et al. Vitreous flow analysis in 20-, 23-, and 25-gaugecutters. Retina. 2008;236-241. Ho, A. Next generation vitrectomy technology for microincision vitrectomy surgery.Presented at the Retina Congress; September 30-October 4, 2009; New York, NY. Novack, R. Relationship of duty cycle versus cut rate for two commercially available systems.Presented at: the American Society of Retina Specialists’ 28th Annual Meeting; August26-September 1, 2010; Vancouver, Canada. Teixeira A, Chong L, Matuoka N, et al. An experimental protocol of the model to quantifytraction applied to the retina by vitreous cutters. Invest Ophthalmol Vis Sci. 2010;4181-4186.

[学科分类] 眼科学

[关键词] retinatoday;bryn mawr communications