A Novel Macroscale Acoustic Device for Blood Filtration

Abstract

Retransfusion of a patient's own shed blood during cardiac surgery is attractive since it reduces the need for allogeneic transfusion, minimizes cost, and decreases transfusion related morbidity. Evidence suggests that lipid micro-emboli associated with the retransfusion of the shed blood are the predominant causes of the neurocognitive disorders. We have developed a novel acoustophoretic filtration system that can remove lipids from blood at clinically relevant flow rates. Unlike other acoustophoretic separation systems, this ultrasound technology works at the macroscale, and is therefore able to process larger flow rates than typical micro-electromechanical system (MEMS) scale acoustophoretic separation devices. In this work, we have first demonstrated the systematic design of the acoustic device and its optimization, followed by examining the feasibility of the device to filter lipids from the system. Then, we demonstrate the effects of the acoustic waves on the shed blood; examining hemolysis using both haptoglobin formation and lactate dehydrogenase release, as well as the potential of platelet aggregation or inflammatory cascade activation. Finally, in a porcine surgical model, we determined the potential viability of acoustic trapping as a blood filtration technology, as the animal responded to redelivered blood by increasing both systemic and mean arterial blood pressure.

Introduction

Cardiopulmonary bypass (CPB) is a common technique in coronary bypass heart surgery and other cardiac/respiratory procedures. During this procedure, the cardiotomy suction portion of the CPB circuit aspirates and returns shed blood back to the patient, limiting the need for additional blood transfusions during the perioperative period. The autologous transfusion limits complications associated with heterologous transfusions, such as transfusion-associated lung injury, transfusion-associated immunomodulation, and cellular hypoxia after red blood cell (RBC) transfusion [1]. However, even with allogenous transfusions, short- and long-term morbidities persist, with neurological complications being most prevalent. Short- and long-term cognitive deficits have been observed in up to 40–60% of patients, while the incidence of acute postoperative stroke can reach up to 5% [2,3]. The shed blood that is collected and recirculated to the patient often contains lipid particles that can agglomerate and potentially form micro-emboli, which is thought to be the leading cause of the neurological complications from CPB [4,5]. Additionally, the shed lipid particles have also been implicated in the damage of other organs, including the lungs, kidneys, and heart [4].

To reduce the embolic load present in the shed blood, methods have been developed to reduce the lipid particles present in transfused blood. Current methods utilized for blood salvage use either filters or centrifugal techniques to separate blood, resulting in moderate levels of RBC separation and return. Typical filters have a pore size of 25 or 40 μm and their removal efficacy is limited to 30–40% [3,5,6], which is consistent with the finding that most lipid particles are less than 20 μm [7]. Filters also become saturated over time, require replacement, and have the potential to disperse larger particles into smaller ones. While centrifugation is more effective at removing lipids than the standard 20–40 μm filter found in the cardiotomy suction circuit, reportedly 50–84% of lipids [8,9], they suffer from limitations of potential fragmentation and deformation of RBCs, and the potential to activate clotting or inflammatory cascades [3]. Additionally, the time required to process the blood in these devices, which work in a “batch” mode, prevents the immediate transfusion of blood back to the patient [8]. Thus, it would be beneficial if a method existed that could more efficiently separate lipids from whole blood, while working in a continuous mode to transfer needed blood components immediately back to the patient.

To improve upon the current available technology, we have designed an acoustophoresis device which separates particles based on acoustic and mechanical properties. Lipid particles have very different acoustic and mechanical properties, such as speed of sound, density, and compressibility compared to the cells found in whole blood. In this way, when the shed blood is processed through a tuned acoustic field, the device could acoustically filter lipids from the major components of blood. Prior work in this space has shown that acoustophoresis may be an effective cell concentrator and lipid removal technology [7]. The feasibility of this approach has been demonstrated on the microscale in micro-electromechanical system (MEMS) work by using acoustophoresis to sort lipid and blood cells into different outlets [2], but hundreds of the MEMS chips operating in parallel would be required to meet the volume and flow rates required for real world application.

In this work, we describe the creation of a macroscale acoustophoresis device that is able to achieve flow rates of up to 2 L/h, which is an approximate 100-fold increase in the flow rate of previous acoustic devices [2,10]. The achievement of these flow rates allows for a clinically relevant volume of blood to be purified. In this device, acoustic energy is applied to a chamber via a piezoelectric transducer, creating a standing wave that captures and separates particles of various sizes. Once the blood is processed, it can be collected and reinfused at a later time, or could potentially be reinfused continuously as the device is filtering a flowing sample. The following experiments explored the effectiveness of the device to filter lipids from blood, the potential of hemolysis, and other negative effects such as platelet aggregation, as well as the success of reinfusion of filtered blood in a porcine surgical model.

Materials and Methods

Reagents.

All experiments using fresh porcine blood were collected according to IACUC approved protocols and donated to the study. Blood was collected in standard hemotology evacuated tubes treated with 7.2 mg ethylenediaminetetraacetic acid purchased from ThermoFisher Life sciences (Agawam, MA) and was diluted by a factor of 10 using phosphate buffered saline, also purchased from ThermoFisher, so that visual observations of particle capture could be conducted, unless otherwise indicated.

Blood/lipid isolation experiments were conducted using the diluted porcine blood and a 0.75% safflower oil emulsion. The mixture consisted of 25 mL emulsion, 25 mL of whole porcine blood, and 200 mL of phosphate buffered saline resulting in an end RBC concentration of approximately 4%. After gentle mixing of the solution to ensure homogeneity, the acoustophoretic system was prepared for the 20 min test. The mixture was pumped through the system using a peristaltic pump that passed the blood/lipid mixture from the inflow reservoir at a rate of 16 mL/min. The cell collector draw was set to 1 mL/min, resulting in a flow of 15 mL/min into the outflow reservoir. Visual observations were made using a Proscope HR USB camera (Bodelin, Lake Oswego, OR) to document lipid aggregation on the top polycarbonate window.

Lipid Analysis.

Similar to the experiments concentrating RBCs, the effectiveness of collecting and concentrating lipid particles was performed first to separate from blood to evaluate the experimental measurement. A 0.2% lipid suspension in saline was created using pork belly fat (chosen as the best simulant for fat collected from a surgical field) that was procured from a local butcher. The lipid suspension was then passed through a LipiGuard^® SB transfusion filter (Haemonetics, Braintree, MA) with 40 μm size cutoff to simulate the actual procedure used in a comparative cell saver protocol, and the final mixture was sent into the acoustic resonator. Once collected, the lipids were dried and weighed.

Blood Analysis.

Complete blood counts, both before and after the acoustic filtration, were obtained using a VetScan HMS hematology analyzer (Abaxis, Co., Union City, CA).

Platelet Aggregation.

Fresh porcine blood was collected and run through the device before being reconstituted. To assess the viability of platelets after sonic capture, blood was treated with adenosine diphosphate after 55 s. Percent of platelet aggregation was measured with a Bio Data Platelet PAP 8-E Aggregation Profiler.

ELISA.

Five ELISA kits, Porcine Haptoglobin ELISA Kit (GenWay Biotech, San Diego, CA), Porcine lactate dehydrogenase (LDH) ELISA Kit (Novateinbio, Woburn, MA), Porcine D-Dimer ELISA Kit (U.S. Biological, Salem, MA), Porcine IL-6 ELISA Kit (Sigma Aldrich, St. Louis, MO), and Porcine TNF-α ELISA Kit (Fisher Scientific, Waltham, MA) were used for this study according to the respective manufacturer's instructions, which in the case of porcine haptoglobin, required a 1/10,000-fold dilution of the sample.

Porcine Surgery.

This study was approved by the Institutional Animal Care and Use Committee (IACUC) of Baystate Medical Center and all methods were performed in accordance with the relevant guidelines and regulations. Two Yorkshire female pigs weighing ∼20 kg were utilized for this study. The day of the procedure they were given a single pre-anesthetic dose of Telazol 5 mg/kg, Ketamine 2.5 mg/kg, and Xylazine 2.5 mg/kg (TKZ) at 1 cc/20 kg. The animals were then endotracheally intubated and maintained under general anesthesia with continuous Isoflurane. Once anesthetized, carotid artery cutdown was performed to allow for invasive hemodynamic monitoring. After adequate hemodynamic monitoring was set up and baseline measurements obtained, the animals were drained of 250 mL of whole blood, which was mixed with ethylenediaminetetraacetic acid to prevent clotting. The amount of blood removed was no more than ∼15% of the animal's total blood volume thereby preventing shock and hemodynamic instability. Of the drained volume, the acoustic wave separation processing occurred in two steps to get fluid back to the animal as quickly as possible. First, 125 mL of whole blood was diluted with 875 mL of normal saline to create a 7:1 saline to blood dilution; this was then processed through the acoustics. Once the first batch was completed, the collected 200 mL blood product was transfused back to the animal, while the second portion of the blood diluted at the same ratio was processed. A total of 400 mL of blood product was transfused back into the animals. Multiple hemodynamic measurements were obtained during the procedure including temperature, systolic blood pressure (SBP), and mean arterial pressure (MAP) and compared to baseline measurements. Multiple blood samples were collected at 30-min intervals after the initial blood collection as well as 5 min after each transfusion to determine the hemoglobin/hematocrit. Three hours after blood transfusion, the animals were euthanized using a single dose of euthanasia solution (Fatal Plus^®, Vortech, Dearborn, MI).

Theory

System Design.

The underlying principle of the acoustic separation is based on the nonuniform acoustic pressure field in the fluid established by an acoustic standing wave. The introduction of a particle in this acoustic pressure field leads to a scattering of the acoustic pressure. The acoustic pressure acting on the surface of the particle then consists of the sum of the incident acoustic standing wave and the scattered wave. The net time averaged force on the particle is found by integrating the acoustic pressure on the surface of the particle and is called the ARF [11]. The ARF is defined as a function of a field potential U, $F_A=-\nabla (U)$

where the field potential U is defined as

$$U=V_0\left[\frac{⟨p^2⟩}{2{\rho }_f c_f^2}{f}_1-\frac{3{\rho }_f⟨u^2⟩}{4}{f}_2\right]$$

and f₁ and f₂ are the monopole and dipole contributions, respectively, defined by

$$f_1=1-\frac{1}{\Lambda {\sigma }^2},\hspace{1em}{f}_2=\frac{2\left(\Lambda -1\right)}{2\Lambda +1}$$

where p is the acoustic pressure, u is the fluid particle velocity, Λ is the ratio of particle density ρ_p to fluid density ρ_f, σ is the ratio of particle sound speed c_p to fluid sound speed c_f, and V_o is the volume of the particle. This expression is valid in the large wavelength approximation, i.e., when the particle radius is significantly smaller than the wavelength of the sound field. Compressibility β is related to the speed of sound by $\beta =1/\rho c^2$. For a one-dimensional standing wave, where the acoustic pressure is expressed as

$$p=A\hspace{0.17em}\text{cos}\hspace{0.17em}(kz)\text{cos}\hspace{0.17em}(\omega t)$$

where A is the acoustic pressure amplitude, k is the wavenumber, ω is the angular frequency, z is the spatial coordinate aligned with the direction of the standing wave, and t is the time. In this case, there is only the axial component of the acoustic radiation force F_ARF, which is found to be

$$F_A(z)=(X\pi R_p^{3}k{P}_{ac}^2/{\rho }_f{c}_f^2)\text{sin}\hspace{0.17em}(2kz)$$

where X is the acoustic contrast factor given by

$$X=\left(\frac{5\Lambda -2}{1+2\Lambda }-\frac{1}{{\sigma }^2\Lambda }\right)$$

Particles with a positive contrast factor will be driven to the pressure nodal planes, and particles with a negative contrast factor will be driven to the pressure anti-nodal planes.

In addition to the axial ARF component, a three-dimensional acoustic wave also exerts lateral forces on the suspended particle, orthogonal to the axis. The axial component of the ARF directs particles to collect in planes at the pressure nodes or antinodes every half wavelength, determined by a positive or negative acoustic contrast factor, respectively. The lateral component of the ARF collects the particles within the planes to local clusters, where they grow in collective size until they reach critical mass and the gravity/buoyancy force causes the particle to sink or rise out of suspension, respectively, as demonstrated in Fig. 1.

Fig. 1.

Schematic of the process. (a) Top view of system shows RBCs (dark circles) and shed blood lipid particles (light circles) entering the system and flowing horizontally. A standing wave establishes pressure node (solid) and antinode (dashed) planes in the center. (b) The axial component of the acoustic radiation force (ARF) aligns particles to nodes/antinodes based on positive/negative contrast factor. (c) The lateral component of the ARF clumps the particles within the planes to create striated columns. (d) Looking at the cross section of a nodal plane, RBCs sink together as clumps to a collector on the bottom. In an antinodal plane, lipid particles do the same and rise out.

Particles dissimilar to a fluid medium can be flown through a standing wave, retained, and then collected. Acoustic-based separation of lipid particles and RBCs is especially attractive because of the differences in density and compressibility of each, as shown in Table 1. RBCs have a positive contrast factor, causing them to deflect to pressure nodes, and are denser than the suspension media (i.e., plasma or saline), causing them to sink when collected. Conversely, lipid particles have a negative acoustic contrast factor, causing them to collect at pressure antinodes, and are less dense than plasma, causing them to rise up when collected, as shown in Fig. 1.

Table 1.

Material properties and acoustic contrast factors of RBCs and lipids

Material	Diameter (μm)	Density (kg/m3)	Compressibility (1/Pa)	Contrast factor	${\rho }_p/{\rho }_{{\mathrm{H}}_2\mathrm{O}}$	${\beta }_p/{\beta }_{{\mathrm{H}}_2\mathrm{O}}$
Red blood cells	6	1092	3.48 × 10−10	3.22 × 10−10	1.09	0.76
Lipids	10–60	921	5.17 × 10−10	−2.19 × 10−10	0.92	1.14

We first examined the effectiveness of the system at separating red blood cells from a diluted supply. Diluted blood was chosen because the acoustical attenuation is lower than in whole blood, and therefore, would require less acoustical energy to operate. The RBC separation tests began by analyzing the many drive parameters. Optimization included identification of the transducer power range, of the fluid flows through the system, and tests of the performance of the system at different dilutions of the blood sample. When the power is too low, the acoustic forces are too weak to collect the particles and to hold them against the flow. However, when the power is too high, the acoustic force interferes with the gravitational precipitation and the transducer may heat the resonator to temperatures too high for live cell treatment. An optimal voltage of 22.5 V was determined and used for the remaining tests.

System Filtration.

The filtration system balances the magnitude of the acoustic radiation forces with the drag forces of particles in the flow and gravitational forces on the particles to settle or buoy out of solution. Before evaluating the system ability to remove lipid particles from blood, it was desired to measure the effectiveness of concentrating RBCs out of suspension. To evaluate the performance of the system, the optimal power level for the acoustic standing wave was determined. Then, with a fixed power input, the drag force amplitude was modulated by running a matrix of experiments at various flow rates. Two flow rates were varied in the experiments, the flow rate into the acoustic resonator and the flow rate at which the concentrated cells were removed from the resonator. Two performance measures were determined: first, the percent of RBCs that entered the system that were collected, and second, the factor by which the cells were concentrated to in the collected volume.

The acoustic field for this device was generated by a custom transducer built in house. The piezoelectric element of the transducer was a 1 in × 1 in PZT-8 (American Piezo, Pennsylvania Company, Mackeyville, PA) plate with a center frequency of 2 MHz. The transducer was driven by a continuous sinusoidal wave generated with a function generator that was amplified with a linear amplifier. The standing wave spanned a distance of 1 in and was terminated by a steel reflector. The transducer was operated in a multimode, i.e., at an excitation frequency close to the anti-resonance frequency of the transducer that excited an eigenfrequency of the transducer (typically a 3 × 3 mode), which then set up a multidimensional standing wave in the fluid with strong axial and lateral acoustic radiation forces [12].

Results

The lipid particles were seen to capture in the standing wave field, aggregate to critical size, and buoy to the top where a “lipid trap” had been designed to remove them from the remaining flow. The results can be seen in Fig. 2. Quantitation of the lipid content gives 60 ± 5, 65 ± 15, and 150 ± 25 mg in the fraction extracted from the lipid collection port, from the resonator volume, and from the filtrate, respectively. In total, this constitutes 275 ± 45 mg of lipids, which coincides within the error with the spiked load of 250 ± 20 mg. Thus, minimal to no lipid load was detected in the fraction collected through the RBC port either by direct measurement or calculation.

Fig. 2.

The lipid suspension was analyzed under a microscope with 40× magnification during multiple steps of the process: (a) lipid particles in saline immediately after spike (scale in image B applies to A–D as well), (b) lipid particles in saline after filtration through LipiGuard^® SB filter, (c) lipid particles in saline after acoustic filtration in filtrate, (d) lipid particles in saline in lipid-collecting port, and (e) top view on the lipid-collecting port at the end of filtration

Due to the dynamic performance of the system, the system performance was characterized in wide range of flow rates to select the optimal combination of in and outflows. The data matrices, presented in Fig. 3, show the system performance at different combinations of the feed flow rate (${\dot{V}}_{\text{in}}$) and of the ratio of the flow rates of the filtrate (${\dot{V}}_f$) and collector (${\dot{V}}_c$) channels (a.k.a., plasma and RBC channels, respectively). These measurements have been performed with 500 mL of ten-fold diluted porcine blood. The selected range of feed flows was centered at 30 mL/min or 1.8 L/h. At flow rates of 10 or 20 ml/min, 92.5% (±8.5%) of the RBCs were captured, and the system was optimized for both RBC capture and packing with higher flow rates of the filtrate in comparison to the collector (Fig. 3).

Fig. 3.

Performance of the acoustic system processing 500 mL of a ten-fold diluted porcine blood at different inflow rates and ratios of outflow rates. The percent of cells collected and concentration factor are presented in the middle and right images, respectively.

We additionally tested to determine if the acoustic device was also capturing white blood cells and platelets while we were collecting the red blood cells. Using a flow rate of 10 ml/min, we observed separation of 91% ( ± 8%) of white blood cells and 86% ( ± 5%) of platelets from the diluted whole blood sample (Table 2). Furthermore, the 10× diluted blood was reconcentrated 6.2–7.4×, returning the blood close to its physiological concentration.

Table 2.

Separation efficiency, concentration factor, and population percentages of cells after filtration

	RBC	WBC	PLT
Separation efficiency (%)	91	91	86
Concentration factor	7.64	6.35	6.20

To examine any effects to the blood after separating it with the acoustic standing waves, potential RBC hemolysis, activation of platelet aggregation and activation of the complement pathway were examined. First, lactate dehydrogenase and haptoglobin ELISAs were performed on recombined treated blood and compared to nontreated blood. There was a slight raise of 15% in haptoglobin levels while LDH levels decreased by 25% (Fig. 4), but it is unlikely that either of these changes is significant. Thus, neither measurement demonstrated hemolysis was occurring after acoustic processing, which was consistent with a lack of hemolysis upon microscopic inspection of the RBCs (data not shown).

Fig. 4.

The hemolysis measurements obtained after acoustic processing of blood by haptoglobin (on the left) and LDH (on the right), n = 3–4

Separated blood was reconstituted and examined to determine if the device induced platelet aggregation in nonstimulated platelets (Fig. 5(a)) as well as if platelets could still be activated after acoustic separation if treated with a platelet agonist (Fig. 5(b)) or if fibrin was formed, as measured by its degradation (Fig. 5(c)). No discernable activation of aggregation was found but the platelets could still be stimulated to aggregate. The ELISA measurement demonstrated no significant change in D-Dimer levels, suggesting that the acoustic filtration did not include arterial or venous thrombus formation. Finally, potential activation of inflammation was examined by exploring TNF-α and IL-6 release. In comparison to the nonacoustically processed blood, there was a decrease of 79% in the IL-6 concentration and 87% in the TNF-α concentration, signifying no inflammatory activation occurred due to the acoustic capture of the cells (Fig. 5(d)).

Fig. 5.

Examination of potential negative effects of blood processing. The potential of arteriole clot formation was examined by (a) platelet aggregation and (b) the ability of platelets to activate was subsequently verified. (c) The potential of venous clot formation was measured by examining for fibrin degradation products. Finally, the potential for the activation of inflammatory pathways (d) was measured using an ELISA to measure interleukin 6 (on the left) and TNF—a levels (on the right).

Finally, we examined the effect of acoustophoretic filtered blood that was autotransfused into a porcine undergoing surgery. The purpose of this test was to determine the physiological effect of processing through the acoustic filtering device on erythrocytes and observe any change in the subject's vitals after receiving a transfusion of processed blood. The pigs were anesthetized and baseline SBP and MAP levels were recorded as well as baseline hemoglobin/hematocrit measurements. SBP and MAP were noted to decrease soon after exsanguination in both animals (Figs. 6(a) and 6(b)). A rise in both is seen prior to the transfusions, which is secondary to the animals' own compensatory drive. However, after each transfusion, there is a noted increase in these values suggestive that the animal does respond to the blood being administered. Remaining variations in the physiology of the animal during the course of the experiment were within the normal range for an animal under anesthesia.

Fig. 6.

Examination of the effects of retransfusion of acoustically processed blood on both SBP and MAP. Data from both of the pigs (799 and 800) are shown.

Discussion

The use of acoustic waves to purify blood cells from lipids, bacteria, and even from other blood cells has been demonstrated for a decade using flow rates of mL/h [13–15]. In general, the use of acoustics for separating cells is more efficient and less damaging to the cells, with separation efficiencies approaching 97% with minimal to no hemolysis [2,15]. However, the technology was limited by these flow rates, which were hundreds of fold lower than what would be considered clinically relevant for surgical procedures or even blood banking.

In this study, we have designed a new generation acoustophoretic device that can separate blood cells at flow rates as high as 2 L/h, which could be further increased by using multiple systems. Thus, for the first time, we can test acoustic filtering and processing of shed blood in surgical procedures at levels that would be appropriate to model surgical conditions. We found that acoustic separation has significant advantages over devices that rely on centrifugation and filtration. Acoustic filtration RBCs approached 90% (Fig. 3) compared to 80% with common centrifugation techniques. Acoustics filtration can be considered gentle on the shed blood, as no hemolysis or RBC fragmentation was observed after acoustics (Fig. 4), which can occur after centrifugation [16]. This finding is consistent with the previously published data examining acoustophoresis on the MEMS scale [2]. Acoustic capture also retains white blood cells and platelets, without leading to activation of platelet aggregation, fibrin generation or inflammatory cascades, including IL-6 and TNF-a (Fig. 5).

The observation that shed blood can be collected, processed, and redelivered with the animal responding by increasing blood pressure demonstrates the potential viability of acoustic trapping as a potential lipid removal technology (Fig. 6). Applications of acoustic filtration to blood purifications have been pursued for a long time, but suffered from an inability to handle flow rates of greater than 60 ml/h. By overcoming these flow rate limitations, we have successfully shown for the first time that acoustic filtration of shed blood can occur at clinically relevant flow levels. Our results demonstrate an improved technology that can capture a higher percentage of RBCs, induce less hemolysis and RBC damage, and retain white blood cells and platelets yet avoid activating coagulation or inflammation cascades.

The successful use of acoustophoresis for filtering and purifying blood has many other potential applications that would be of interest in future studies. The filtering of bacteria, such as E. coli, from the blood could be useful in the treatment of sepsis, as recent studies have suggested that one potential solution is to focus on developing strategies to filter bacteria out of the bloodstream and into areas of the body such as the liver where the body's natural immune system is better equipped to eliminate them or ideally, out of the body all together [17–19]. There is also great interest in isolating circulating tumor cells from the blood stream to use as a diagnostic and prognostic marker of cancer progression. Most of the currently available technologies are unable to isolate the cells in sufficient number to analyze them or require alterations to the cells for capture that alters their physiology [20]. Using acoustics, previous studies have been able to isolate cancer cells in a noncontact, label-free method [21]. Finally, the use of acoustics to separate blood may be of significant interest in transfusion medicine, as cells captured in a standing waved can be washed. Washing of cells can help eliminate some common causes of RBC storage lesions, such as microparticles, free hemoglobin, and adenosine triphosphate.

Funding Data

• National Heart, Lung, and Blood Institute (Grant No. HL118833).