Microchannel technologies for artificial lungs: (2) screen-filled wide rectangular channels.

Artificial lungs with blood-side channels on a 10-40 microm scale would be characterized, similar to the natural lungs, by tens of thousands to hundreds of millions parallel blood channels, short blood paths, low pressure drops, and low blood primes. A major challenge for developing such devices is the requirement that the multitude of channels must be uniform from channel to channel and along each channel. One possible strategy for developing microchannel artificial lungs is to fill broad rectangular channels with micro scale screens that can provide uniform support and stability. The present work explores the effectiveness of 40 microm screen-filled blood-side channels and, as a comparison, 82 microm screen-filled channels. Small concept-devices, consisting of a single 69 mm wide and 3 or 6 mm long channel, were tested using 30% hematocrit blood and oxygen or air on the gas side. The measured oxygen fluxes in the devices were in the range of 4 to 9 x 10(-7) moles/(min x cm(2)), with the latter close to the theoretical membrane limit. The pressure drop was in the range of 1-6 mm Hg. Extrapolating the data to a device designed to process 4 L/min suggests a required blood prime of only 35 ml.

Microchannel technologies for artificial lungs: (1) theory.

The feasibility of developing micro channel artificial lungs is calculated for eight possible strategies: 12 and 25 microm circular channels imbedded in gas-permeable sheets, 12 and 25 microm high open rectangular channels with gas-permeable walls, 12 and 25 microm high broad open channels with support posts and gas-permeable walls, and two 40 microm high screen-filled rectangular channels with gas-permeable walls. Each strategy is considered by imposing a pressure drop maximum of 10 mm Hg and limiting the possibility of shear-induced blood trauma. The pressure drop limit determines the acceptable channel length and required size to oxygenate 4 L/min of venous blood. Circular channels imbedded in open-pore, gas-permeable materials are especially attractive. With 12 microm channels, such a device would require 140 million, 0.8 mm long channels, but the total size of the gas-exchange section would be only 57 ml and a blood prime of only 13 ml. Also attractive are 12 mum high broad open channels with support posts and 40 mum screen-filled rectangular channels. The total size of the former would be 250 ml with a blood prime of 13 ml, and the total size of the latter would be 270 ml with a blood prime of 27 ml.

Microchannel technologies for artificial lungs: (3) open rectangular channels.

Lithographic techniques were used to develop patterned silicone rubber membranes that provide 15 microm high microchannels for artificial lungs. Two types of devices were fabricated as a proof-of-concept: one has a series of parallel, straight, open rectangular channels that are each 300 microm wide, separated by 200-microm walls, and 3-mm long and the other is a wide rectangular channel with support posts, also 3- mm long. Experiments with 30% hematocrit, venous, bovine blood showed average oxygen fluxes ranging from 11 x 10(-7) moles/(min x cm(2)) at a residence time of 0.04 sec to 6.5 x 10(-7) moles/(min x cm(2)) at a residence time of 0.20 sec. The average oxygen flux vs. residence time, which is due to transverse molecular diffusion, follows the same relation for all membranes tested. The corresponding increase in hemoglobin saturation ranged from 9% at the residence time of 0.04 sec to 24% at the residence time of 0.20 sec. The support-post channel membranes are attractive for designers because they can be arbitrarily wide and would be less prone to blockage.

Development of an implantable artificial lung: challenges and progress.

Unlike dialysis, which functions as a bridge to renal transplantation, or a ventricular assist device, which serves as a bridge to cardiac transplantation, no suitable bridge to lung transplantation exists. Our goal is to design and build an ambulatory artificial lung that can be perfused entirely by the right ventricle and completely support the metabolic O2 and CO2 requirements of an adult. Such a device could realize a substantial clinical impact as a bridge to lung transplantation, as a support device immediately post-lung transplant, and as a rescue and/or supplement to mechanical ventilation during the treatment of severe respiratory failure. Research on the artificial lung has focused on the design, mode of attachment to the pulmonary circulation, and intracorporeal versus paracorporeal placement of the device.

Hemodynamic effects of attachment modes and device design of a thoracic artificial lung.

A thoracic artificial lung (TAL) was designed to treat respiratory insufficiency, acting as a temporary assist device in acute cases or as a bridge to transplant in chronic cases. We developed a computational model of the pulmonary circulatory system with the TAL inserted. The model was employed to investigate the effects of parameter values and flow distributions on power generated by the right ventricle, pulsatility in the pulmonary system, inlet flow to the left atrium, and input impedance. The ratio of right ventricle (RV) power to cardiac output ranges between 0.05 and 0.10 W/(L/min) from implantation configurations of low impedance to those of high impedance, with a control value of 0.04 W/(L/min). Addition of an inlet compliance to the TAL reduces right heart power (RHP) and impedance. A compliant TAL housing reduces flow pulsatility in the fiber bundle, thus affecting oxygen transfer rates. An elevated bundle resistance reduces flow pulsatility in the bundle, but at the expense of increased right heart power. The hybrid implantation mode, with inflow to the TAL from the proximal pulmonary artery (PA), outflow branches to the distal PA and the left atrium (LA), a band around the PA between the two anastomoses, and a band around the outlet graft to the LA, is the best compromise between hemodynamic performance and preservation of some portion of the nonpulmonary functions of the natural lungs.

Partial prevention of monocyte and granulocyte activation in experimental vein grafts by using a biomechanical engineering approach.

Leukocytes interact with endothelial cells and contribute to the development of vascular diseases such as thrombosis and atherosclerosis. These processes are possibly influenced by mechanical factors. This study focused on the role of mechanical stretch in the activation of monocytes and granulocytes in experimental vein grafts. Two models were created by using rats: a nonengineered vein graft with increased tensile stress, which was created by grafting a jugular vein into the abdominal aorta, and an engineered vein graft with reduced tensile stress, which was created by restricting the vein graft into a cylindrical sheath constructed by using fixative-treated intestinal tissue. The density of activated monocytes and granulocytes, which attached to the endothelium, and the distribution of the intercellular adhesion molecule (ICAM)-1 in endothelial cells were examined using immunohistological assays. It was found that, in nonengineered vein grafts, the density of activated monocytes and granulocytes increased significantly compared to that in normal jugular veins at day 1, 5, 10 and 20. At each observation time, the cell density in the proximal region of the nonengineered vein grafts was significantly higher than that in the middle and distal regions, and the cell density in the distal region was significantly higher than that in the middle region. These changes were associated with ICAM-1 clustering at day 1 and 5 and focal ICAM-1 un-regulation at day 10 and 20. In engineered vein grafts, the density of activated monocytes and granulocytes decreased significantly compared to that in nonengineered vein grafts at all observation times, although it was significantly higher than that in normal jugular veins. At each observation time, the cell density in the proximal and distal regions was significantly higher than that in the middle region, but no significant difference was found between the proximal and distal regions. ICAM-1 clustering along endothelial cell borders was found at day 1 and 5, but no apparent focal ICAM-1 up-regulation was found at day 10 and 20. These results suggested that mechanical stretch due to exposure to increased tensile stress contributed to the activation of monocytes and granulocytes in experimental vein grafts, and this event could be partially prevented by reducing tensile stress using a biomechanical engineering approach.

Influence of a natural and a synthetic inhibitor of factor XIIIa on fibrin clot rheology.

We investigated the origins of greater clot rigidity associated with FXIIIa-dependent cross-linking. Fibrin clots were examined in which cross-linking was controlled through the use of two inhibitors: a highly specific active-center-directed synthetic inhibitor of FXIIIa, 1,3-dimethyl-4,5-diphenyl-2[2(oxopropyl)thio]imidazolium trifluoromethylsulfonate, and a patient-derived immunoglobulin directed mainly against the thrombin-activated catalytic A subunits of thrombin-activated FXIII. Cross-linked fibrin chains were identified and quantified by one- and two-dimensional gel electrophoresis and immunostaining with antibodies specific for the alpha- and gamma-chains of fibrin. Gamma-dimers, gamma-multimers, alpha(n)-polymers, and alpha(p)gamma(q)-hybrids were detected. The synthetic inhibitor was highly effective in preventing the production of all cross-linked species. In contrast, the autoimmune antibody of the patient caused primarily an inhibition of alpha-chain cross-linking. Clot rigidities (storage moduli, G') were measured with a cone and plate rheometer and correlated with the distributions of the various cross-linked species found in the clots. Our findings indicate that the FXIIIa-induced dimeric cross-linking of gamma-chains by itself is not sufficient to stiffen the fibrin networks. Instead, the augmentation of clot rigidity was more strongly correlated with the formation of gamma-multimers, alpha(n)-polymers, and alpha(p)gamma(q)-hybrid cross-links. A mechanism is proposed to explain how these cross-linked species may enhance clot rigidity.

Structural origins of fibrin clot rheology.

The origins of clot rheological behavior associated with network morphology and factor XIIIa-induced cross-linking were studied in fibrin clots. Network morphology was manipulated by varying the concentrations of fibrinogen, thrombin, and calcium ion, and cross-linking was controlled by a synthetic, active-center inhibitor of FXIIIa. Quantitative measurements of network features (fiber lengths, fiber diameters, and fiber and branching densities) were made by analyzing computerized three-dimensional models constructed from stereo pairs of scanning electron micrographs. Large fiber diameters and lengths were established only when branching was minimal, and increases in fiber length were generally associated with increases in fiber diameter. Junctions at which three fibers joined were the dominant branchpoint type. Viscoelastic properties of the clots were measured with a rheometer and were correlated with structural features of the networks. At constant fibrinogen but varying thrombin and calcium concentrations, maximal rigidities were established in samples (both cross-linked and noncross-linked) which displayed a balance between large fiber sizes and great branching. Clot rigidity was also enhanced by increasing fiber and branchpoint densities at greater fibrinogen concentrations. Network morphology is only minimally altered by the FXIIIa-catalyzed cross-linking reaction, which seems to augment clot rigidity most likely by the stiffening of existing fibers.

A hypothesis regarding vascular acoustic emission accompanying arterial injury induced by balloon angioplasty.

Stress-induced structural damage is often accompanied by sound release. This behavior is known as acoustic emission (AE). We hypothesize that vascular injury such as that produced by balloon angioplasty is associated with AE. Postmortem human peripheral arterial specimens were randomly partitioned into test (n = 10) and control segments (n = 10). Test segments were inserted into a pressurization circuit and subjected to two consecutive hydrostatic pressurizations. Amplitude, frequency, and energy content of the AE signals released during pressurization were quantified. Test and matched control segments subsequently underwent identical histological processing. Pressure-induced tissue trauma was estimated via computerized histomorphometric analysis of the resulting slides (n = 100). Vascular acoustic emission (VAE) signals exhibited an amplitude range of +/- 5.0 mu bars and were observed to occur during periods of increasing intraluminal pressure. The VAE signal power within the monitored bandwidth was concentrated below 350 Hz. More than 25 times as much VAE energy was released during the first pressurization as during the second: 1,855 +/- 513.8 mJ vs. 73 +/- 44.9 mJ (mean +/- SEM, p < 0.006). Estimates of circumferential intimal wall stress at AE onset averaged 170 kPa, slightly below reported values of arterial tissue rupture strength. Histomorphometric estimates of tissue trauma was greater for the test than their matched control segments (p < 0.0001). These preliminary data suggest that detectable acoustic energy is released by vascular tissue subjected to therapeutic stress levels. Histological analysis suggest that the underlying source of sound energy may be related to tissue trauma, independent of histological preparation artifacts. From this preliminary work, we conclude that VAE may be a fundamental property accompanying vascular tissue trauma, which may have applications to improving balloon angioplasty outcomes.

In vitro identification of angioplasty-induced injury by use of vascular acoustic emissions.

BACKGROUND: We have developed a novel method of diagnosing stress-induced vascular injury. This approach uses the sound energy released from atherosclerotic arterial tissue during in vitro balloon angioplasty to characterize type and severity of induced trauma. METHODS AND RESULTS: Thirty-two postmortem human peripheral arterial specimens 1.0 cm long were subjected to in vitro balloon angioplasty with simultaneous acoustic emission monitoring. Specimens were examined before and after angioplasty to ascertain the extent of angioplasty-induced injury. Gross observation was used to identify dissection. A three-dimensional intravascular ultrasound reconstruction technique was used to estimate the luminal surface area of the specimen. Change in luminal surface area (postangioplasty minus preangioplasty) was used to quantify induced injury. The energy content and spectral distribution of the digitally acquired vascular acoustic emission (VAE) signals were computed. Comparisons of angioplasty-induced trauma with VAE signal characteristics were made. Dissection (mural laceration of variable depth) was observed in 15 of 32 specimens. Eleven showed no evidence of induced dissection, and 6 had preexisting intimal disruptions. The energy content of the VAE signals collected from specimens with dissection was greater than that obtained from those in which dissection was absent: 845 +/- 89.4 mJ (mean +/- SEM; n = 15) versus 128 +/- 40.8 mJ (n = 1 l; P < .001). Comparison of induced trauma and VAE signal energy demonstrated a proportional relationship (r = .87, P < .001, n = 32). CONCLUSIONS: VAE signals contain information characterizing type and severity of angioplasty-induced arterial injury. Because vascular injury is related to adverse procedural outcome, development of VAE technology as an adjunct to conventional diagnostic modalities may facilitate optimal balloon angioplasty delivery and postprocedural care.

Testing of an intrathoracic artificial lung in a pig model.

A low input impedance, intrathoracic artificial lung is being developed for use in acute respiratory failure or as a bridge to transplantation. The device uses microporous, hollow fibers in a 0.74 void fraction, 1.83 m2 surface area bundle. The bundle is placed within a thermoformed polyethylene terephthalate glucose modified housing with a gross volume of 800 cm3. The blood inlet and outlet are 18 mm inner diameter vascular grafts. Between the inlet graft and the device is a 1 inch inner diameter, thin-walled, latex tubing compliance chamber. These devices were implanted in Yorkshire pigs via median sternotomy with an end to side anastomosis to the pulmonary artery and left atrium. The distal pulmonary artery was occluded to divert the right ventricular output to the device. Pigs 1 and 2 were supported fully for 24 hrs and then killed. Pig 3 was supported partially for 20 hrs and died from bleeding complications. The first implant, in a 55 kg male pig, transferred an average of 176 ml/min +/- 42.4 of O2 and 190 ml/min +/- 39.7 of CO2 with an average blood flow rate of 2.71/min +/- 0.46. The normalized average right ventricular output power, Pn, was 0.062 W/(L/min) +/- 0.0082, and the average device resistance, R, was 3.5 mmHg/(L/min) +/- 0.62. The second implant, in a 60 kg male pig, transferred an average of 204 ml/min +/- 22.5 of O2 and 242 ml/min +/- 17.2 of CO2 with an average blood flow rate of 3.7 L/min +/- 0.45, Pn of 0.064 W/(L/min) +/- 0.0067, and R of 4.3 mmHg/(L/min) +/- 0.89. The third implant, in an 89 kg male pig, transferred an average of 156 ml/min +/- 39.6 of O2 and 187 ml/min +/- 21.4 of CO2 with an average blood flow rate of 2.5 L/min +/- 0.49, Pn of 0.052 W/(l/min) +/- 0.0067, and R of 3.4 mmHg/(L/min) +/- 0.74. These experiments suggest that such an artificial lung can temporarily support the gas transfer requirements of adult humans without over-loading the right ventricle.

New design for a pumping artificial lung.

A new prototype of a pumping artificial lung (PAL) has been designed and tested. The device performs the functions of both the pump and oxygenator components of an extracorporeal perfusion circuit. Previous prototypes that the authors developed (Type A) had gas exchanging microporous fibers formed into propeller-like vanes that, upon rotation, pump the blood. The design of the new PAL prototypes (Type B) uses the rotation of an annular bank of fibers to drive flow. The fiber bank, including sealed gas manifolds, lies within the housing of a modified Bio-Medicus BMP-50 pump head (Bio-Medicus, Eden Prairie, MN). Rotation of the fiber bank is driven through a magnetic coupling. Inlet and outlet gas lines enter the pump head through a sealed bearing. The Type A PAL suffered from insufficient pumping rates and gas exchange, necessitating redesign. The authors have constructed two PAL-B prototypes with a priming volume of only 140 ml and gas exchange surface areas of 0.16 and 0.60 m2. During in vitro saline testing, these prototypes showed significant pump performance, pumping 7.0 L/min against zero head at 3,500 rpm. The larger prototype had exchange rates in saline of up to 71 ml O2/min and 75 ml CO2/min. Gas exchange fluxes (O2 = 119 ml/[min.m2] and CO2 = 125 ml/[min.m2]) for the PAL-B are significantly higher than that of commercially available oxygenators in saline. Future prototypes will have increased surface area and fibers smaller than the 0.038 cm outside diameter fibers used in the present prototypes. A primary concern in using microporous fibers to push the blood was the durability of the fibers at high pump speeds. High speeds exhibited no negative effects on gas exchange abilities or fiber durability.

Mixing of two solutions combined by gravity drainage.

A variety of medical therapies require the mixing of solutions from two separate bags before use. One scenario for the mixing is to drain the solution from one bag into the other by gravity through a short connecting tube. The degree of mixing in the lower bag depends on the relative densities of the two solutions, the geometry of the two bags and the connecting tube, and the placement of the connecting tube. Solutions with densities differing by as much as 12% were mixed by draining the solution from an upper bag into a lower bag for a particular geometric configuration. The two solutions had different electrical conductivities, and the conductivity of the combined solution as it exited from the lower bag was used as a measure of the effectiveness of mixing. When the more dense solution was drained from the upper bag into the less dense solution in a lower bag, mixing was very effective. The incoming jet of high density solution entrained the low density solution. Flow visualization indicated that the incoming jet penetrated to the bottom of the lower bag, and resulting large vortical structures enhanced mixing. When the less dense solution was drained from the upper bag into the more dense solution in the lower bag mixing was less effective. The buoyancy force reduced the momentum of the incoming jet such that it did not penetrate to the bottom of the lower bag, resulting in stratification of the solutions.

Computer-assisted design of an implantable, intrathoracic artificial lung.

A semiempirical mathematical model of convective oxygen transport is used to design a new, low pressure loss, implantable artificial lung that could be used as a bridge to lung transplantation in patients with advanced respiratory failure. The mass transfer and flow friction relations pertinent to the design of a cross-flow hollow fiber membrane lung are described. The artificial lung is designed to transfer over 200 ml/min of oxygen at blood flow rates up to 5 L/min. A compact design and a blood-side pressure loss of < 15 mm Hg allows the device to be implanted in the left chest without the need for a prosthetic blood pump. Surgical implantation of the artificial lung would require the creation of inflow and outflow anastomoses. Oxygen would be supplied via an external source. Blood properties, operating conditions, and empirically determined mass transfer and flow properties are all specified and input into a computer program that numerically solves the design equations. Computer-generated values for the device frontal area, blood path length, and fiber surface area are thereby obtained. The use of this computer-assisted design minimizes the need for extensive trial-and-error testing of prototype devices. Results from in vitro tests of a prototype implantable lung indicate that the mathematical model we describe is an accurate and useful tool in the design of hollow fiber artificial lungs.

Use of a mathematical model to predict oxygen transfer rates in hollow fiber membrane oxygenators.

A semi-empirical theoretical model of oxygen transfer is used to predict the rates of oxygen transfer to blood in hollow fiber membrane oxygenators over a wide range of inlet conditions. The predicted oxygen transfer rates are based on performance of the devices with water, which is more cost effective and easier to handle than blood for in vitro evaluations. Water experiments were conducted at three different flow rates to evaluate oxygen transfer performance in three commercially available membrane oxygenators. Data obtained from these experiments were used in a computer model to predict the rate of oxygen transfer to bovine blood at specified inlet conditions. Blood experiments were conducted at three different flow rates at a wide variety of inlet conditions, including different pH levels, hemoglobin concentrations, and oxyhemoglobin saturations for the three types of oxygenators. The measured and predicted oxygen transfer rates are closely correlated, which suggests that we have an accurate, reliable method for predicting oxygen transfer in hollow fiber membrane lungs.

A pumping artificial lung.

The authors have developed a single device that performs the functions of a centrifugal pump and a membrane artificial lung. Unlike other systems that combine pre-existing components, our device is constructed so that the vanes of an impeller pump are made up of gas exchanging microporous fibers. The device has a variety of applications: in an easily primed emergency cardiopulmonary bypass circuit, as a low surface area component of extracorporeal life support (ECLS) circuits, and as a low volume, high perfusion rate bridge to transplant. Five prototype devices, with gas exchange surface areas ranging from 0.09 to 0.35 m2, have been tested in vitro to characterize the gas exchange and pumping capabilities of the device. These small devices pump fluid effectively. The larger device could pump 2.7 l/min with 91 mmHg pressure difference from inlet to outlet. These preliminary devices transferred to only 33 ml/min of oxygen (O2) and 35 ml/min of carbon dioxide (CO2), however. The combination of pumping ability and gas exchange is encouraging, but it is apparent that larger surface areas and less blood shunting around the gas exchanging impellers are needed for sufficient gas exchange. Somewhat higher surface areas are feasible within the pump-head casing used for these preliminary prototypes; larger casings could be used for still higher surface areas.

Design and evaluation of a new, low pressure loss, implantable artificial lung.

The authors designed and tested an artificial lung intended for intrathoracic implantation as a bridge to lung transplantation in chronic pulmonary insufficiency or as an alternative in the treatment of advanced acute respiratory failure. The prototype devices are comprised of 380 microns outer diameter polypropylene matted fibers with a blood path length of 3.5 cm, frontal area of 128 cm2, void fraction (porosity) of 0.53, and surface area of approximately 2.2 m2. Blood flow is external and approximately perpendicular to the fiber bundle, which fits in an extruded, flexible polyethylene terephthalate housing. Inflow and outflow anastomoses are made to the pulmonary artery and the left atrium, respectively, thereby avoiding a prosthetic blood pump. Inlet and outlet gas lines exit through the chest wall. Nine in vitro experiments of oxygen (O2) transfer performance by the device, with water, initially were done. Our previously described semiempirical mathematical model of convective O2 transfer in cross-flow, hollow fiber membrane lungs was applied to the results from the water tests to predict the transfer rates at any set of blood conditions. Five in vitro blood tests were conducted using a single-pass technique to evaluate O2 and carbon dioxide (CO2) transfer rates, measure pressure losses, and compare predicted and measured O2 transfer rates. O2 transfer rates of 150-200 ml/min, and CO2 transfer rates exceeding 200 ml/min, could be achieved at blood flow rates as great as 4 l/min. Pressure drops of approximately 10-20 mmHg were observed at blood flow rates of 2-4 l/min. Preliminary results of device implantation in two pigs indicate the feasibility of achieving clinically significant O2 and CO2 transfer rates with a low blood-side pressure loss.

A dynamic intravascular artificial lung.

Intravascular lung assist devices (ILADs) must transfer sufficient amounts of oxygen and carbon dioxide to and from limited surface areas. It has become apparent that passive devices, i.e., those without an active means for enhancing transfer, cannot achieve sufficient transfer within the space available. High speed rotation or oscillation of fiber sheets can increase transfer rates up to 800% over the rates achieved by a stationary device, judiciously configured fiber sheets cause an additional benefit when rotated: reduced resistance to blood flow across the device. The authors have developed a series of device prototypes based on these principles of transfer augmentation and minimization of flow resistance. The prototypes are small enough to fit inside the vena cava, with transfer surface areas ranging from 0.1 m2 to 0.5 m2. Transfer rates of O2 up to 53 ml/min and CO2 up to 51 ml/min and fluxes of 208 ml (min/m2) for O2 and 310 ml (min/m2) for CO2 have been achieved.

A pumping intravascular artificial lung with active mixing.

Intravascular, as well as extracorporeal, artificial lungs need to be effective and efficient in transferring both oxygen and carbon dioxide. This paper describes the preliminary development of a device that not only is efficient in gas transfer, but also can reduce any pressure loss by providing its own pumping action. The exchange surfaces of the device consist of many short, microporous, hollow fibers arranged in layers like the threads of a screw and placed in a cross-flow configuration. Rotation of the device greatly increases gas transfer efficiency, by increasing the relative velocity between the blood and the fiber surfaces, and pushes the blood along a path similar to that of an Archime-dean screw. In vitro water tests of prototype devices indicate that the rotation can enhance the gas transfer rates by as much as a factor of six. In vitro blood studies indicate moderate blood pumping against zero pressure head, a simulation of veno-venous bypass.

Rapid platelet separation and collection for postoperative autologous transfusion.

The authors describe a technique for rapid, on-line, precardiopulmonary bypass separation and collection of 10-20% of a patient's platelets for storage and postoperative reinfusion. In a three stage process, blood is diverted through a hollow fiber hemofilter and platelets are concentrated near the fiber walls; the core is flushed of whole blood and the concentrated platelets are collected. Process mechanics were investigated in vitro with bovine blood; efficacy was studied with in vivo ovine experiments; and platelet function and viability were examined with in vitro human blood tests. Using commercially available hemofilters with 13,000 polyacrylonitrile fibers, the authors collected 1 x 10(11) to 2 x 10(11) platelets in 250 ml in 3.5 min. With fresh human blood, collected platelets exhibited normal morphology and aggregation response to adenosine diphosphate, collagen, and thrombin for 4 hours after collection.