Share this article on:

Fiber Bundle Design for an Integrated Wearable Artificial Lung

Madhani, Shalv P.*†; Frankowski, Brian J.*†; Federspiel, William J.*†‡§

doi: 10.1097/MAT.0000000000000542
Pulmonary

Mechanical ventilation (MV) and extracorporeal membrane oxygenation (ECMO) are the only viable treatment options for lung failure patients at the end-stage, including acute respiratory distress syndrome (ARDS) and chronic obstructive pulmonary disease (COPD). These treatments, however, are associated with high morbidity and mortality because of long wait times for lung transplant. Contemporary clinical literature has shown ambulation improves post-transplant outcomes in lung failure patients. Given this, we are developing the Pittsburgh Ambulatory Assist Lung (PAAL), a truly wearable artificial lung that allows for ambulation. In this study, we targeted 180 ml/min oxygenation and determined the form factor for a hollow fiber membrane (HFM) bundle for the PAAL. Based on a previously published mass transfer correlation, we modeled oxygenation efficiency as a function of fiber bundle diameter. Three benchmark fiber bundles were fabricated to validate the model through in vitro blood gas exchange at blood flow rates from 1 to 4 L/min according to ASTM standards. We used the model to determine a final design, which was characterized in vitro through a gas exchange as well as a hemolysis study at 3.5 L/min. The percent difference between model predictions and experiment for the benchmark bundles ranged from 3% to 17.5% at the flow rates tested. Using the model, we predicted a 1.75 in diameter bundle with 0.65 m2 surface area would produce 180 ml/min at 3.5 L/min blood flow rate. The oxygenation efficiency was 278 ml/min/m2 and the Normalized Index of Hemolysis (NIH) was less than 0.05 g/100 L. Future work involves integrating this bundle into the PAAL for which an experimental prototype is under development in our laboratory.

From the *McGowan Institute for Regenerative Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania; Department of Bioengineering, University of Pittsburgh, Pittsburgh, Pennsylvania; Department of Chemical and Petroleum Engineering, University of Pittsburgh, Pittsburgh, Pennsylvania; and §Department of Critical Care Medicine, University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania.

Submitted for consideration August 2016; accepted for publication in revised form January 2017.

Disclosure: William J. Federspiel chairs the scientific advisory board and is the co-founder of ALung Technologies. Other authors do not have any financial disclosures related to the work presented in this article.

This study was supported by NIH Grant Number RO1 HL117637, the Commonwealth of PA and the McGowan Institute for Regenerative Medicine.

Correspondence: William J. Federspiel, McGowan Institute for Regenerative Medicine, Departments of Bioengineering, Chemical and Petroleum Engineering and Critical Care Medicine University of Pittsburgh, 3025 East Carson Street, Pittsburgh, PA 15203. Email: federspielwj@upmc.edu.

Acute and chronic lung diseases are major healthcare problems. The Centers for Disease Control and Prevention (CDC) report lung disease is the third leading cause of death in the United States.1 Acute respiratory distress syndrome (ARDS) affects 190,000 patients annually and is associated with a mortality of 64%.2 Chronic lung disease affects 12.7 million annually and is associated with a mortality of 135,000.3,4 As lung disease becomes end-stage, lung transplant is the only viable treatment.5 The number of lung transplants have been growing at the rate of 179 transplants annually (3,519 in 2010),6 but the supply of organs is not sufficient to meet the need for lung transplants. The Organ Procurement and Transplant Network (OPTN) database states that approximately 2,500 patients are added to the wait list annually. The average time of the wait list in 2013 was 4 months varying from 2.6 to 9.7 months depending on the level of sickness of a patient, and the wait list mortality is 10–15 deaths per 100 patient-years of waiting.6

Current intervention techniques include mechanical ventilation (MV) and extracorporeal membrane oxygenation (ECMO). Prolonged MV injures the patient in the form of barotrauma and volutrauma to the lung, and results in poor post-transplant outcomes.7,8 Conventional ECMO can be used as a bridge to transplant but is cumbersome and expensive in addition to being associated with high morbidity and mortality.9–11 This morbidity and mortality is only exacerbated through progressive deconditioning as patients are confined in MV and ECMO.9,10 Recently clinical implementation of the Maquet Cardiohelp, or Quadrox along with centrifugal,11,12 and novel cannula such as the Avalon Elite (Maquet Cardiovascular LLC, Wayne, NJ) dual-lumen cannula have simplified ambulation on ECMO.13 Ambulation using such systems improves patient outcomes as this allows patients to walk, eat, and exercise during therapy—reducing muscle deconditioning.14–17 Yet, the newer generation of ECMO systems remain bulky and cumbersome.

We are developing a highly integrated blood pump and oxygenator as a wearable artificial lung or respiratory assist device. By integrating a hollow fiber membrane (HFM) bundle for gas exchange directly with an efficient centrifugal pump, the Pittsburgh Ambulatory Assist Lung (PAAL) is a truly wearable device that allows for patient ambulation. The PAAL requirements based on other device used clinically and under research development includes18,19 a small form factor, long-term (1–3 month) durability, 180 ml/min oxygenation at 3.5 L/min for providing partial to complete lung support, compatibility with the Avalon Elite Dual Lumen Cannula (DLC) and a Normalized Index of Hemolysis (NIH) under 0.05 g/100 L.20 The small form factor can be achieved by minimizing the size of the HFM bundle, which typically represents the largest component of the pump oxygenator system. In turn, a smaller HFM bundle requires a design with increased gas exchange efficiency. Diffusional boundary layers dictate the gas exchange in HFM bundles, with thicknesses that scale approximately as the square root of fluid velocity past the fiber surfaces.21 Increasing velocity of fluid flow through the fiber bundle thus increases gas exchange efficiency. In other respiratory assist applications, “active mixing” has been used as a means to increase the fluid velocity past fiber surfaces by using ancillary components like rotating impellers adjacent to the fiber bundle or rotating the fiber bundle itself within a stationary housing.23–26 In this study, we investigated a simpler, passive means to improve mass transfer efficiency in hollow fiber bundles by manipulating their form factor to increase the fluid velocity past fiber surfaces. We used a simple 1D model of blood flow and gas exchange in hollow fiber bundles based on a previously published mass transfer correlation22 to characterize oxygenation efficiency as a function of fiber bundle diameter. Various fiber bundles were fabricated to validate the model and to help determine a fiber bundle form factor (diameter–gas exchange surface area) that would oxygenate blood 180 ml/min at 3.5 L/min. The final design was characterized in vitro through a gas exchange and hemolysis study.

Back to Top | Article Outline

Methods

Oxygen Transfer Model

The PAAL-specific HFM geometry was modeled using a previously published23 mass transfer correlation. The mass balance on O2 in the fiber bundle is as follows:

where Q is the flow rate through the bundle, R is the bundle radius, av is the surface area to volume ratio, z is the axial coordinate,

is the oxygen partial pressure difference between the fluid and gas sides, k is the mass transfer coefficient.

The oxygen concentration

blood accounting for the dissolved and bound concentration is given by as follows:

where

is the partial pressure of oxygen, CT is the hemoglobin binding capacity and

is oxygen saturation. The mass transfer coefficient k for oxygen transfer from the inside of the fiber to the blood is as follows:

where Sh is the Sherwood number defined as

, where

and

are the solubility and diffusivity of oxygen in blood, respectively, and dh the hydraulic diameter, which is equivalent to the fiber diameter. Re is the Reynold’s number and is defined as

, where

and

are fluid density and viscosity, respectively, V is the superficial blood flow velocity, Sc is Schmidt number which is

, where

is the kinematic viscosity of blood.

The differential Equation in 1 was solved using Equations 2 and 3 in Matlab (MathWorks, Natick, MA) with the built-in ODE solver based on the Runge–Kutta method. Oxygenation efficiency (oxygenation normalized to surface area) was calculated for the three benchmark bundles described in Table 1 (FB-1 to FB-3). After validation, final bundle geometry (FB-F) was designed using the model.

Back to Top | Article Outline

Fiber Bundle Manufacturing

Commercially available Membrana PMP 90/200 type hollow fiber sheets (44 fibers/in) (Membrana GmbH, Wuppertal, Germany) were used for manufacturing the four fiber bundles as the described in Table 1. Bundles were designed to have blood flow over fibers, with gas flowing through the fiber lumen. Square sheets cut from a spool of fiber were die cut and sealed with an arbor press. These sheets then fit in a round custom potting fixture. The potting fixture contained a glue reservoir attached to a mold in which fibers are stacked alternatingly at a 14° crossing angle; fibers were oriented perpendicular to the principal direction of blood flow. The mold was spun on its axis at 1,400 rpm for 12 h until the polyurethane potting adhesive (Vertellus Performance Materials Inc., Greensboro, NC) cured. Void fraction of the bundles was 0.5. Pressure drop for fiber bundles was estimated using a modified Blake–Kozeny equation. Custom test fixtures are shown in Figure 1 housed bundles during experiments.

Back to Top | Article Outline

In Vitro Gas Exchange

Gas exchange testing followed ISO 7199 standards24 using 7 L of locally collected slaughterhouse blood. Bovine or porcine blood was used interchangeably as hill dissociation curves across experiments overlapped. Blood was passed through a 40 μm filter (Pall Biomedical, Inc., Fajardo, PR), and treated with heparin (10 IU/ml) and gentamycin (0.1 mg/ml). The experimental setup consisted of the single pass loop system is shown in Figure 2. The loop contained two custom manufactured compliant 6 L blood reservoir connected to a Biomedicus BP 80-X pump (Medtronic, Minneapolis, MN) and the test device. Oxygenated blood was deoxygenated with a Medtronic Affinity NT 2.5 m2 oxygenator (Medtronic, Minneapolis, MN) placed downstream of the test device. Blood temperature was maintained at 37°C with a PolyScience 210 heater (PolyScience Inc., Niles, IL) connected to the deoxygenator’s built-in heat exchanger. Standard R-3603 Tygon tubing (Cole-Parmer, Vernon Hills, IL) connected loop components.

Before collecting a data point, a blend of N2, CO2, O2 sweep gas were flowed through the deoxygenator, conditioning blood to have an oxygen saturation of 65% ± 5% and a

of 45 mm Hg ± 5 mm Hg. Once conditioned, blood passed from the inlet reservoir through the loop into the outlet reservoir such that the post device blood was separate from the conditioned blood at all times. Flow rates tested were 1, 2, 3, 3.5, and 4 L/min. An ultrasound flow probe (Transonic Systems Inc., Ithaca, NY) measured flow. Pure oxygen sweep gas flowed through the test device at 7.5 L/min, measured with a GR Series mass flow controller (Fathom Technologies, Georgetown, TX). Each point was repeated once. One sample was drawn from each of the sampling ports is shown in Figure 2. A Rapid Point 405 Blood Gas Analyzer with Co-oximetry (Siemens Healthcare Diagnostics Inc., Tarrytown, NY) measured blood gases and oxygen saturation.

Oxygen transfer rates were calculated as follows:

where

is the rate of oxygenation, Q is the blood flow rate,

is the oxygen solubility in blood

,

is the partial pressure difference across the device, CT is the binding capacity

, and

is the saturation difference across the device.

Back to Top | Article Outline

In Vitro Hemolysis

Hemolysis testing followed established standards.25,26 Testing comprised of two identical loops in which flow was driven using a Centrimag blood pump (Thoratec, Pleasanton, CA) and measured using an ultrasound flow probe. A 1,200 ml compliant blood reservoir (Medtronic, Minneapolis, MN) was used for each loop. Temperature was maintained at 37°C using a water bath and heat exchanger. The test loop comprised of a fiber bundle module and a 27 Fr. Avalon Elite DLC (Maquet Cardiovascular LLC, Wayne, NJ) in addition to the Centrimag pump. The control loop comprised of just the cannula and pump.

Both loops ran simultaneously at 3.5 L/min for a shortened duration of 3 h owing to the linearity in the trend between plasma-free hemoglobin (PfHb) and time (R2 > 0.9). Every half-hour, one 3 ml waste sample was pulled from each loop before drawing one 5 ml sample. Plasma-free hemoglobin was tracked using this 5 ml sample. The supernatant was taken from the sample after centrifuging at 800g for 15 min, and then spun at 7,200g for 10 min. The absorbance of the purified plasma was measured at 540 nm using a Genesys 10 UV-Vis spectrophotometer (Thermo Fisher Scientific, Waltham, MA). Plasma-free hemoglobin was calibrated to absorbance by generating linear standard curves (n = 3) before the experiment. These standard curves correlated PfHb to absorbance via a slope of 0.11 g/dl/A. Hematocrit was measured using a capillary tube in an IEC Mb microcentrifuge (International Equipment Co., Needham Hts, MA).

Blood damage was then expressed as a NIH, which normalizes the rate of change of PfHb

to loop volume (V), hematocrit (Hct), and blood flow rate (Q) using the following relationship26:

Back to Top | Article Outline

Results

Benchmark HFM bundle model calculations and experiment values are shown in Figure 3. The percent difference between the model calculations and experiment results ranged from 4.9% to 13.3% for FB-1 is shown in Figure 3A, 3–17.5% for FB-2 is shown in Figure 3B, and 10.4–14.6% for FB-3 is shown in Figure 3C. Oxygenation efficiency increased as fiber bundle diameter was reduced. At 3.5 L/min

(mm Hg) across fiber bundle increased from 48 to 68 for FB-1, 43–76 for FB-2, and 47–78 for FB-3.

The model was then used to predict the FB-F geometry that would achieve our target oxygenation performance of 180 ml/min at 3.5 L/min. Manufacturing constraints relating to the centrifugal potting of fiber bundles prevented further reduction of fiber bundle diameter below 1.75 in. Oxygenation performance of FB-F is shown in Figure 4. Oxygenation increased from 79 to 207 ml/min as flow rate was increased from 1 to 4 L/min. Oxygenation of 180.7 ml/min, oxygenation efficiency of 278 ml/min/m2 and a

(mm Hg) change from 40 to 79 was achieved at 3.5 L/min. Figure 5 shows the NIH of FB-F. The test condition had an NIH of 0.021 g/100 L, whereas the control had an NIH of 0.018 g/100 L.

Back to Top | Article Outline

Discussion

Mechanical ventilation and ECMO are the only viable treatment options for lung failure patients at the end-stage, including ARDS and COPD. These treatments, however, are associated with high morbidity and mortality because of long wait times for lung transplant.9,10,14,15,27 Contemporary clinical literature has shown ambulation improves outcomes in lung failure patients.14–17 Given this, we are developing the PAAL, a truly wearable, compact artificial lung that allows for patient ambulation during bridge to recovery or transplant. In this study, FB-F met our design target of 180 ml/min oxygenation despite a low (0.65 m2) surface area. The FB-F design has a relatively high efficiency of 278 ml/min/m2, almost two times higher than standard blood oxygenators used for ECMO today.28 Final bundle will be incorporated in the final PAAL device for future bench and animal testing.

Our study found that decreasing the fiber bundle diameter and increasing fiber bundle length increases oxygenation efficiency although still maintaining a low level of hemolysis (NIH 0.021 g/100 L). The overall contribution of the bundle to hemolysis is ~14% of the total measured hemolysis, given that the baseline level of hemolysis is NIH 0.018 g/100 L. Albeit low, the cannula and loop generate a larger part of the measured hemolysis. Further, these NIH values are within the acceptable limits (NIH < 0.05 g/100 L) of hemolysis for clinically approved oxygenators.20 As blood flows over HFMs, a fluid boundary layer forms at the surface of the fibers. The thickness of the boundary layer is related to flow velocity past fibers.21 In this study, the flow velocity is increased by maintaining constant blood flow but reducing frontal area of fiber bundles. Although path length is increased, the increased enhancement is caused by velocity increase as residence time is reduced with increasing path length (residence time is 2.1 s for FB-F, 2.8 s for FB-1). Devices in the past have used “active mixing” to reduce boundary layer thickness;22,29–31; however, in this study we achieved this through a simple geometric means thus ensuring low hemolysis.

There are other artificial lung devices under research development for treating patients with lung failure.11,18,19,28,32,33 The ambulatory pump lung (APL) device features a fully magnetically levitated centrifugal pump integrated into a 0.8 m2 surface area annular fiber bundle18 having 200 ml/min/m2 oxygenation efficiency. The compliant thoracic artificial lung (cTAL) features a pumpless device with a 2.4 m2 surface area bundle having ~95 ml/min/m2 oxygenation efficiency. The device is implanted in the patient’s thoracic cavity, relying upon the patient’s right ventricular function to pump blood through the device.19 FB-F designed in this study has 40% higher efficiency than the APL and 200% higher efficiency than the cTAL. Higher oxygenation efficiency ultimately translates into lower required fiber bundle surface area, which not only helps create a more compact artificial lung but also potentially reduces the adverse blood–material interactions associated with a larger blood contacting area.

Our design leads to a path length of 3.12 in across the fiber bundle. This is longer than the APL (0.85 in) and cTAL (1.49 in) devices. At each of their respective operating conditions residences times in these devices are: 1.97 s in the APL and 6–9 s in the.18,19 However, the residence time in FB-F is 2.1 s at the operating flow rate of 3.5 L/min. This time is within the range of the APL and cTAL devices, which have been tested in vivo up to 30 days with few biocompatibility issues. Further the surface area of FB-F is smallest compared with and other device which potentially mitigates biocompatibility issues. Additionally, some oxygenators used clinically such as the Sorin Inspire have long path lengths as well.34 As part of the PAAL project, we are also developing novel thromboresistant coatings35 for the PAAL fiber bundle. Overall, we do not expect that the long bundle length used in the PAAL will induce significant hemocompatibility issues. One final point is that the circular cross section of our fiber bundle minimizes dead flow zones which can occur in square cross section oxygenators such as the Quadrox, which is currently the oxygenator used in the Maquet Cardiohelp portable ECMO system.

In conclusion, studies based on three benchmark fiber bundles validated our gas exchange model and based on this model, we developed a 0.65 m2 fiber bundle that met our target of providing 180 ml/min of oxygenation at 3.5 L/min blood flow rate. The overall oxygenation efficiency was high (278 ml/min/m2), whereas blood hemolysis was low (NIH 0.021 g/100 L). Future work involves integrating this fiber bundle design with a centrifugal pump we are developing into a single housing to create a highly integrated, compact, and wearable artificial lung. The integration will be guided by computational fluid dynamics analysis to create a first-generation PAAL prototype for bench and animal validation studies.

Back to Top | Article Outline

References

1. US Department of Health and Human Services: Health, United States, 2015. 2016.DHHS Publication No. 2016–1232.
2. Figueira JF, Oliveros MO, López JL, Civantos BC, Fernández LF. Acute respiratory distress syndrome: Analysis of incidence and mortality in a university hospital critical care unit. Critl Care 2012.16: P396.
3. Mannino DM. COPD: Epidemiology, prevalence, morbidity and mortality, and disease heterogeneity. Chest 2002.121(5 suppl): 121S–126S.
4. Chronic Obstructive Pulmonary Disease (COPD) Fact Sheet. American Lung Association. Available at: http://www.lung.org/lung-disease/copd/resources/facts-figures/COPD-Fact-Sheet.html. Accessed August 9, 2015.
5. Kotloff RM, Thabut G. Lung transplantation. Am J Respir Crit Care Med 2011.184: 159–171.
6. Christie JD, Edwards LB, Kucheryavaya AY, et al. The Registry of the International Society for Heart and Lung Transplantation: 29th adult lung and heart-lung transplant report-2012. J Heart Lung Transplant 2012.31: 1073–1086.
7. Villar J, Blanco J, Añón JM, et al. The ALIEN study: Incidence and outcome of acute respiratory distress syndrome in the era of lung protective ventilation. Intensive Care Med 2011.37: 1932–1941.
8. Nieman GF, Gatto LA, Bates JH, Habashi NM. Mechanical ventilation as a therapeutic tool to reduce ARDS incidence. Chest 2015.148: 1396–1404.
9. Biscotti M, Sonett J, Bacchetta M. ECMO as bridge to lung transplant. Thorac Surg Clin 2015.25: 17–25.
10. Maury G, Langer D, Verleden G, et al. Skeletal muscle force and functional exercise tolerance before and after lung transplantation: A cohort study. Am J Transplant 2008.8: 1275–1281.
11. Haneya A, Philipp A, Foltan M, et al. First experience with the new portable extracorporeal membrane oxygenation system Cardiohelp for severe respiratory failure in adults. Perfusion 2012.27: 150–155.
12. Palanzo D, Qiu F, Baer L, Clark JB, Myers JL, Undar A. Evolution of the extracorporeal life support circuitry. Artif Organs 2010.34: 869–873.
13. Wang D, Zhou X, Liu X, Sidor B, Lynch J, Zwischenberger JB. Wang-Zwische double lumen cannula-toward a percutaneous and ambulatory paracorporeal artificial lung. ASAIO J 2008.54: 606–611.
14. MacLaren G, Combes A, Bartlett RH. Contemporary extracorporeal membrane oxygenation for adult respiratory failure: Life support in the new era. Intensive Care Med 2012.38: 210–220.
15. Garcia JP, Iacono A, Kon ZN, Griffith BP. Ambulatory extracorporeal membrane oxygenation: A new approach for bridge-to-lung transplantation. J Thorac Cardiovasc Surg 2010.139: e137–e139.
16. Perme CS, Southard RE, Joyce DL, Noon GP, Loebe M. Early mobilization of LVAD recipients who require prolonged mechanical ventilation. Tex Heart Inst J 2006.33: 130–133.
17. Pruijsten R, van Thiel R, Hool S, Saeijs M, Verbiest M, Reis Miranda D. Mobilization of patients on venovenous extracorporeal membrane oxygenation support using an ECMO helmet. Intensive Care Med 2014.40: 1595–1597.
18. Zhang J, Taskin ME, Koert A, et al. Computational design and in vitro characterization of an integrated maglev pump-oxygenator. Artif Organs 2009.33: 805–817.
19. Schewe RE, Khanafer KM, Arab A, Mitchell JA, Skoog DJ, Cook KE. Design and in vitro assessment of an improved, low-resistance compliant thoracic artificial lung. ASAIO J 2012.58: 583–589.
20. Kawahito S, Maeda T, Yoshikawa M, et al. “Blood trauma induced by clinically accepted oxygenators.” ASAIO J 2001.47: 492–495.
21. Federspiel WJ, Henchir KA. Lung, artificial: Basic principles and current applications, in Encyclopedia of Biomaterials and Biomedical Engineering. 2004, pp. Taylor & Francis, 910–921.
22. Svitek RG, Frankowski BJ, Federspiel WJ. Evaluation of a pumping assist lung that uses a rotating fiber bundle. ASAIO J 2005.51: 773–780.
23. Svitek RG, Federspiel WJ. A mathematical model to predict CO2 removal in hollow fiber membrane oxygenators. Ann Biomed Eng 2008.36: 992–1003.
24. AAMI 7199:2009 Cardiovascular implants and artificial organs – blood-gas exchangers (oxygenators. Association for the Advancement of Medical Instrumentation. 2009.
25. ASTM F1841-97 Standard Practice for Assessment of Hemolysis in Continuous. Flow Blood Pumps. ASTM Book Of Standards, Volume 2013.13.01.
26. Koller T Jr, Hawrylenko A. Contribution to the in vitro testing of pumps for extracorporeal circulation. J Thorac Cardiovasc Surg 1967.54: 22–29.
27. Strueber M. Artificial lungs: Are we there yet? Thorac Surg Clin 2015.25: 107–113.
28. Maquet HLS Set Advanced Flyer. Available at: https://www.maquet.com/int/products/hls-set-advanced/. Accessed January 19, 2017.
29. Jeffries RG, Frankowski BJ, Burgreen GW, Federspiel WJ. Effect of impeller design and spacing on gas exchange in a percutaneous respiratory assist catheter. Artif Organs 2014.38: 1007–1017.
30. Wu ZJ, Gartner M, Litwak KN, Griffith BP. Progress toward an ambulatory pump-lung. J Thorac Cardiovasc Surg 2005.130: 973–978.
31. Makarewicz AJ, Mockros LF, Anderson RW. A pumping intravascular artificial lung with active mixing. ASAIO J 1993.39: M466–M469.
32. Leverett LB, Hellums JD, Alfrey CP, Lynch EC. Red blood cell damage by shear stress. Biophys J 1972.12: 257–273.
33. De Bartolo C, Nigro A, Fragomeni G, et al. Numerical and experimental flow analysis of the Wang-Zwische double-lumen cannula. ASAIO J 2011.57: 318–327.
34. Stehouwer MC, de Vroege R, Kelder JC, Hofman FN, Bastian A, Bruins P. Effect of oxygenator size on air removal characteristics: a clinical evaluation. ASAIO J 2016.62: 421–426.
35. Ye SH, Arazawa DT, Zhu Y, et al. Hollow fiber membrane modification with functional zwitterionic macromolecules for improved thromboresistance in artificial lungs. Langmuir 2015.31: 2463–2471.

artificial lung; hollow fiber membrane; oxygenator design

Copyright © 2017 by the American Society for Artificial Internal Organs