Microfluidic-Based Biosensor for Blood Viscosity and Erythrocyte Sedimentation Rate Using Disposable Fluid Delivery System

Abstract

To quantify the variation of red blood cells (RBCs) or plasma proteins in blood samples effectively, it is necessary to measure blood viscosity and erythrocyte sedimentation rate (ESR) simultaneously. Conventional microfluidic measurement methods require two syringe pumps to control flow rates of both fluids. In this study, instead of two syringe pumps, two air-compressed syringes (ACSs) are newly adopted for delivering blood samples and reference fluid into a T-shaped microfluidic channel. Under fluid delivery with two ACS, the flow rate of each fluid is not specified over time. To obtain velocity fields of reference fluid consistently, RBCs suspended in 40% glycerin solution (hematocrit = 7%) as the reference fluid is newly selected for avoiding RBCs sedimentation in ACS. A calibration curve is obtained by evaluating the relationship between averaged velocity obtained with micro-particle image velocimetry (μPIV) and flow rate of a syringe pump with respect to blood samples and reference fluid. By installing the ACSs horizontally, ESR is obtained by monitoring the image intensity of the blood sample. The averaged velocities of the blood sample and reference fluid (<U_B>, <U_R>) and the interfacial location in both fluids (α_B) are obtained with μPIV and digital image processing, respectively. Blood viscosity is then measured by using a parallel co-flowing method with a correction factor. The ESR is quantified as two indices (t_ESR, I_ESR) from image intensity of blood sample (<I_B>) over time. As a demonstration, the proposed method is employed to quantify contributions of hematocrit (Hct = 30%, 40%, and 50%), base solution (1× phosphate-buffered saline [PBS], plasma, and dextran solution), and hardened RBCs to blood viscosity and ESR, respectively. Experimental Results of the present method were comparable with those of the previous method. In conclusion, the proposed method has the ability to measure blood viscosity and ESR consistently, under fluid delivery of two ACSs.

1. Introduction

Microcirculation plays a substantial role in regulating blood flows and exchanging substances (gases, nutrients, and waste) between blood samples and peripheral tissues. Impaired microcirculation commonly leads to organ failures or mortality [1]. There is a need for comprehensive research that offers an insight that intrinsic properties and flow characteristics of blood samples share with microcirculatory disorders such as hypertension, sickle cell anemia, and diabetes [2]. The previous study has reported that biophysical properties of blood samples (hematocrit (Hct), viscosity, and erythrocyte sedimentation rate (ESR)) are strongly correlated with coronary heart diseases [3]. Thereafter, the biophysical properties of the blood sample have been studied extensively for the effective monitoring of circulatory disorders [4,5,6,7,8,9].

Under normal physiological conditions, red blood cells (RBCs) occupy 40–50% of blood volume. As RBCs are the most abundant cells in the blood sample, the biophysical properties of the blood sample are determined dominantly by properties of RBCs. The characteristics of RBCs, including morphology, membrane viscoelasticity, and RBCs count, are evaluated by quantifying several biophysical properties of blood samples, including viscoelasticity (or viscosity), deformability, and hematocrit. In that regard, plasma proteins in blood samples induce RBC aggregation, which occurs at an extremely low shear rate (i.e., $\dot{\gamma }$ = 1~10 s⁻¹) [10] or stasis. Among the biophysical properties of blood samples, blood viscosity is determined by several factors, including hematocrit, plasma viscosity, RBCs deformability, and RBCs aggregation. Thus, their properties of blood samples are employed to monitor variations in the characteristics of blood samples. At lower shear rates, RBC aggregation causes to increase blood viscosity. At high shear rates, the deformation and alignment of RBCs lead to a decrease in blood viscosity. In other words, blood viscosity provides information on aggregation and deformability simultaneously. However, at extremely low shear rates, a syringe pump (SP) exhibits fluidic instability and RBC sedimentation continuously occurs. A microfluidics-based viscometer does not provide consistent values of blood viscosity. Conventionally, blood viscosity has been measured at sufficiently high shear rates (i.e., $\dot{\gamma }$ > 10 s⁻¹ [11,12] or $\dot{\gamma }$ > 50–100 s⁻¹ [13,14]), especially in microfluidic environments. For the reason, blood viscosity obtained with a microfluidic device does not give sufficient information on the contributions of plasma proteins to RBC aggregation. To evaluate variations in plasma proteins consistently, it is additionally necessary to quantify RBCs aggregation or ESR.

A microfluidic device has several advantages, including small volume consumption, fast measurement, easy sample handling, high sensitivity, and disposability. Thus, it has been widely used to measure various biophysical properties of blood samples (i.e., blood viscosity [15], RBCs aggregation [16], RBCs deformability [17,18], and hematocrit [19]).

The previous methods for measuring blood viscosity are conveniently divided into three categories (i.e., driving sources, devices, and quantification techniques). First, extrinsic driving sources such as SPs [20], pressure sources, and hand-held pipettes [13] have been suggested for delivering a blood sample into a specific device. Additionally, intrinsic driving sources such as capillary force (or surface tension) [21,22] and gravity force [23] have been applied to supply blood samples into a device. Second, various devices such as a microelectromechanical system (MEMS)-based microfluidic device, a 3D-printed microfluidic device [13,24], and a paper-based device [25] have been suggested for inducing blood flow in a specifically constrained direction. Third, quantification techniques such as advancing meniscus (i.e., variations of a blood column over time) [15,22,26,27], the falling time of a metal sphere in a tube [28], electric impedances (i.e., resistance, capacitance) [29,30], droplet length [31], digital flow compartment with a microfluidic channel array [11,12], interface detection in co-flowing streams [32,33], and reversal flow switching in a Wheatstone bridge analog of a fluidic circuit [14] have been suggested to measure blood viscosity.

To measure RBCs aggregation in microfluidic environments, a blood sample is placed into a microfluidic channel. By applying shear stress to the blood sample with external driving systems (i.e., an SP [34], pinch valve [16], or stirring motor [35]), the RBCs in the blood sample are aggregated or disaggregated, depending on the shear rate. Several quantification methods, such as light intensity (i.e., transmission, and back-scattering) [16], electrical conductivity [36,37], microscopic RBC images [38,39,40], ultrasonic images [41], and optical tweezers [42] have been suggested for obtaining temporal variations of RBCs aggregation. As another approach, RBC aggregation can be quantified by measuring the sedimentation distances of RBCs in a blood sample during a specific duration (i.e., ESR). Unlike the conventional Westergren ESR method, a microfluidic-based ESR measurement is quantified by measuring the conductivity of the blood sample in a PDMS chamber with a square cross-section (i.e., each side = 4 mm, depth = 5 mm) [43]. Owing to the continuous ESR in the driving syringe, RBC-free regions (or depleted regions) expand from the top layer with an elapse of time. The blood sample is supplied into a microfluidic device from the top layer of the driving syringe. To monitor blood flows in the microfluidic channel, microscopic images are sequentially captured with a high speed camera. Image intensity of each microscopic image is calculated over time by conducting digital image processing. The ESR is then evaluated by quantifying temporal variations of the image intensity [44].

To measure blood viscosity and RBC aggregation inexpensively, two SPs should be replaced with an inexpensive and disposable delivery system. To remove the syringe pump, single ACS is suggested to infuse the blood sample into a microfluidic device for measuring pressure and RBCs aggregation over continuously varying flow rates [45]. In this study, the ultimate goal of this study is to measure blood viscosity and RBC aggregation (or ESR), without two SPs.

In this study, a simple method for measuring blood viscosity and ESR is proposed. It involves the quantification of the interfacial location in a co-flowing channel and microscopic image intensity of blood sample flowing in a microfluidic device. Two air-compressed syringes (ACSs) are employed to simultaneously deliver the blood sample and reference fluid. Based on an ACS for delivering blood samples as suggested in previous studies [46,47], two ACSs are suggested to deliver blood samples and reference fluid simultaneously. Since the flow rates of both fluids are not specified under fluid delivery with the ACSs, it is necessary to quantify them with a time-resolved micro-particle image velocimetry (µ-PIV) technique. Based on a parallel co-flowing method with a correction factor [32], the blood viscosity is measured by monitoring the interfacial location in a co-flowing channel. Unlike the previous studies [46,47], two ACSs are installed horizontally to measure ESR effectively. Continuous sedimentation in the ACS causes an expansion of an RBC-free layer from the top layer. When blood samples are delivered to the blood channel from the ACS, the populations of RBCs (or hematocrit) are reduced over time. Since a continuous ESR contributes to increasing microscopic image intensity of blood flows, the ESR can be quantified by monitoring the image intensity of the blood sample.

When compared to previous methods that have the ability to measure blood viscosity under fluid delivery with syringe pumps, two syringe pumps are replaced by two ACSs as a novelty of this method. Here, a 40% glycerin solution is newly selected as the reference fluid. RBCs as fluid tracers are added into reference fluid. Velocity fields of both fluids are obtained consistently over time by conducting a time-resolved micro-PIV technique.

By installing the ACSs horizontally, the continuous ESR inside the ACS is filled with the blood sample causing it to expand RBC-free regions. As RBCs aggregation tends to increase substantially at lower hematocrit or lower velocity, it contributes to increasing the image intensity of blood samples. Thus, it is possible to evaluate the ESR by monitoring the microscopic image intensity of the blood sample.

2. Materials and Methods

2.1. Fabrication of Microfluidic Device and Experimental Procedure

A microfluidic device for measuring blood viscosity and ESR consisted of two inlets (a, b), one outlet (a), and a T-shaped channel (width = 250 μm, depth = 20 μm), as shown in Figure 1A-a. The T-shaped channel was composed of a blood channel, a reference channel, and a co-flowing channel. When analyzing the velocity fields of each fluid, the T-shaped channel does not require to align each microscopic image in the horizontal direction. Conventional micro-electromechanical-system techniques, such as photolithography and deep reactive ion etching (DRIE), were employed to fabricate 4-inch silicon mold. To peel off PDMS block from the master mold easily, plasma surface treatment was conducted after the DRIE process [48]. PDMS elastomer (Sylgard 184, Dow Corning, Midland, MI, USA) was mixed with a curing agent at a ratio of 10:1. After positioning the mold on a petri dish, the PDMS mixture was poured into the mold. Air bubbles dissolved in the PDMS were removed by operating a vacuum pump (WOB-L Pump, Welch, Gardner Denver, Milwaukee, WI, and USA) for 1 h. After curing the PDMS in a convective oven at 70 °C for 1 h, a PDMS block was peeled off from the mold. It cut with a razor blade. Two inlets and outlets were punched with a biopsy punch (outer diameter = 1.0 mm). After treating the surfaces of the PDMS block and a glass slide with an oxygen plasma system (CUTE-MPR, Femto Science Co., Gyeonggi-do, Korea), the PDMS block was bonded on a glass substrate. A microfluidic device was finally prepared by placing it on a hotplate at 120 °C for 10 min.

As shown in Figure 1A-b, two polyethylene tubes (L₁) (length = 300 mm, inner diameter = 500 μm, and thickness = 500 μm) were tightly fitted into two inlets (a, b). The end of each tube was connected to the individual syringe needle of the ACS. The outlet of each ACS was clamped with a pinch valve. The other tube (L₂) (length = 200 mm, inner diameter = 500 μm, and thickness = 500 μm) was tightly fitted into outlet (a). The end of the tube (L₂) was connected to a waste dish. To remove air bubbles and avoid non-specific binding of plasma proteins to the inner surface of the channels, the channel was filled with bovine serum albumin (BSA) solution (C_BSA = 2 mg/mL) through outlet (a). After an elapse of 10 min, the microfluidic channel was newly filled with 1× PBS.

Based on the concept of ACS as reported in a previous study [45], two ACSs were employed to deliver the blood sample and reference fluid into the microfluidic device. Figure A1A (Appendix A) showed two ACSs filled with the blood sample (Hct = 50%) and reference fluid (RBCs suspended in 40% glycerin solution (Hct = 7%)). Each ACS was composed of a disposable syringe (~ 1 mL), a fixture, and a pinch valve. Two pinch valves were used to stop or allow the fluid flow of each fluid. Each ACS was placed horizontally on the stage of the optical microscope and fixed with an adhesive tape. Here, an angle of inclination of the ACS only depended on an individual fixture. It was certain that the installation angle of the ACS remained identical because the same fixture of the ACS was used for all experiments.

As shown in Figure A1B (Appendix A), the operation of each ACS was classified into five steps: (1) piston movement at the lowest position forward at t = t₁, (2) air suction by moving the piston to 0.7 mL backward at t = t₂, (3) blood suction by moving the piston to 0.3 mL backward at t = t₃, (4) air compression by moving the piston to 0.3 mL forward at t = t₄, and (5) blood delivery by removing the pinch valve at t = t₅. As the air cavity inside the ACS was compressed to 0.3 mL, internal pressure increased substantially above atmospheric pressure. Similarly, the reference fluid was sucked into the syringe. The remainder of the procedure was the same as blood delivery. By removing two pinch valves, blood sample and reference fluid were delivered to the corresponding inlets because pressure difference increased inside the ACS.

The microfluidic device was positioned on an optical microscope (BX51, Olympus, Tokyo, Japan) equipped with a 20× objective lens (NA = 0.4). As shown in Figure 1A-c, a high-speed camera (FASTCAM MINI, Photron, Tokyo, Japan) was used to obtain sequential microscopic images of the blood sample and reference fluid flowing in the microfluidic channels. The camera offered a spatial resolution of 1280 × 1000 pixels. Each pixel corresponded to 10 μm physically. A function generator (WF1944B, NF Corporation, Yokohama, Japan) triggered the high-speed camera at an interval of 1 s. Then, two microscopic images were captured at a frame rate of 5 kHz.

To minimize the effect of temperature on blood viscosity, all experiments were conducted at a room temperature of 25 °C. Contributions of two factors (i.e., humidity, and atmospheric pressure) to the present method were neglected. After the blood sample was injected into an ACS, the blood sample did not contact with environment air. Additionally, the ACS was operated by pressure difference (ΔP) between pressure inside ACS (P_ACS) and atmosphere pressure (P_atm) (i.e., ΔP = P_ACS − P_atm) [47]. The pressure difference depended on air volume inside the ACS (i.e., gauge pressure), rather than atmospheric pressure.

2.2. Quantification of Microscopic Image Intensity, Blood Velocity Fields, and Interfacial Location

First, blood viscosity was obtained by quantifying the velocity fields of blood sample flowing in the blood channel, the velocity fields of reference fluid flowing in the reference channel, and the interface location between two fluids flowing in the co-flowing channel.

RBCs as fluid tracers were added into reference fluid to obtain the velocity fields of the reference fluid. To measure velocity fields of reference fluid consistently, RBCs should be distributed uniformly in reference fluid during experiments. According to previous studies [1,2], when reference fluid was prepared by adding RBCs into 1× PBS and filled into the ACS, sedimentation of RBCs in ACS occurred continuously over time. RBCs in reference fluid did not flow uniformly over time. After a certain lapse of time, there were no fluid tracers in reference fluid. It was then impossible to obtain the velocity fields of the reference fluid. To resolve the critical issue, a 40% glycerin solution was carefully selected as a base solution in reference fluid. Additionally, to minimize contributions of RBCs to velocity fields and viscosity, hematocrit of RBCs added into reference fluid was fixed at Hct = 7%.

As shown in Figure 1B, an ROI (300 × 300 pixels) was selected to obtain the velocity fields of the blood sample flowing in the blood channel. Another ROI with 300 x 300 pixels was selected to obtain the velocity fields of the reference fluid flowing in the reference channel. By conducting a time-resolved µPIV technique, the velocity fields of the blood sample (U_B) across the blood channel width were obtained over time. Additionally, velocity fields of the reference fluid flow (U_R) across the reference channel width were obtained over time. The size of the interrogation window was selected as 64 × 64 pixels. The window overlap was set to 75%. The velocity fields were validated and corrected with a median filter. The averaged velocities (<U_B>, <U_R>) of both fluids were calculated as an arithmetic average over the specific ROI. To obtain the interface (i.e., blood sample-filled width) in the co-flowing channel (α_B), an ROI with 300 × 400 pixels was selected in the co-flowing channel. A gray-scale microscopic image was converted into a binary image by adopting Otsu’s method [49]. By conducting an arithmetic average over the ROI, variations of the interfacial location in the co-flowing channel (α_B) were obtained over a period of time.

Second, the ESR was evaluated by quantifying the microscopic image intensity of the blood sample flowing in the blood channel. To evaluate the microscopic image intensity of blood flows, and ROI with 300 x 300 pixels was selected in the blood channel. The image intensity of the blood sample flowing in the blood channel was obtained by conducting digital image processing with a commercial software package (Matlab 2019, Mathworks, Natick, MA, USA). An averaged value of microscopic image intensity (<I_B>) was obtained by performing an arithmetic average of image intensity over the specific ROI.

2.3. Quantification of Blood Viscosity and ESR

As a preliminary demonstration, blood samples (normal RBCs suspended in specific dextran solution (10 mg/mL), Hct = 50%) and reference fluid were delivered to the corresponding inlets (a, b) under the fluid delivery with two ACSs. To visualize the velocity fields of the reference fluid flowing in the reference channel, the reference fluid was prepared by adding normal RBCs (Hct = 7%) into a 40% glycerin solution.

Figure 1C-a showed microscopic images captured at specific times (t) (t = 50, 300, 600, 900, 1200, and 1500 s). Above t = 600 s, the populations of RBCs flowing in the blood channel decreased substantially over time. As shown in Figure 1C-b, temporal variations of U_B and U_R were obtained by conducting the µPIV technique. As the pressure difference between the inner pressure and atmospheric pressure tended to decrease over time in the ACS, the averaged velocity of the reference fluid (<U_R>) tended to decrease stably over time. However, owing to the continuous ESR inside the ACS, an RBC-free liquid was observed in a tube, as shown in Figure A1C (Appendix A). The averaged velocity of the blood sample (<U_B>) varied unstably above t = 400 s. Figure 1C-c showed the temporal variations in the image intensity of the blood sample flowing in the blood channel (<I_B>), and the interface between the two fluids in the co-flowing channel (α_B). Similar to U_B, the continuous ESR inside the ACS led to unstable behaviors in <I_B> and α_B. In this study, the separation time when unstable behavior began was denoted as T_st. At t < T_st, three factors (<U_B>, <U_R>, and <α_B>) exhibited stable variations over time. Thus, the blood viscosity was quantified from the three factors (<U_B>, <U_R>, and α_B). For a rectangular channel with an extremely low aspect ratio, an approximate formula of fluidic resistance was derived approximately as $R=\frac{12 {\mu }_B L }{w h^3 }$. A co-flowing channel was filled with a blood sample and reference fluid, respectively. For simple mathematical representation, both streams were represented as two fluidic resistances connected in parallel. The corresponding fluidic resistance for each fluid was derived as R_B =$\frac{12 {\mu }_B L}{W{\alpha }_B{h}^3}$ for a blood sample, and R_R =$\frac{12 {\mu }_R L}{W(1-{\alpha }_B) h^3}$ for reference fluid. Here, μ_R meant the viscosity of the reference fluid. As both fluids had the same pressure drop (i.e., ΔP = R_B·Q_B = R_R·Q_R), blood viscosity formula (μ_B) was derived as ${\mu }_B={\mu }_{R }\times \left(\frac{{\alpha }_B}{1-{\alpha }_B}\right)\times \left(\frac{Q_R}{Q_B}\right)$. Here, Q_B and Q_R represented the flow rate of the blood sample and reference fluid, respectively. The simple mathematical model did not account for real boundary conditions in co-flowing flows. Thus, to compensate for the deviation from the real boundary condition, the previous study included a correction factor in the analytical formula of blood viscosity. According to the blood viscosity formula reported in a parallel co-flowing method with a correction factor [32], the blood viscosity formula (μ_B) was modified as ${\mu }_B=C_f\times {\mu }_{R }\times \left(\frac{{\alpha }_B}{1-{\alpha }_B}\right)\times \left(\frac{Q_R}{Q_B}\right)$. Since the correction factor (C_f) was varied depending on the channel size, a numerical simulation was conducted to determine the correction factor. Based on a procedure discussed in a previous study [32], a numerical simulation using commercial computational fluid dynamics (CFD) software (CFD ACE+, ESI Group, Paris, France) for a rectangular channel (width = 250 μm, depth = 20 μm) was conducted to obtain the viscosity of the test fluid with respect to the interface. For convenience, it was assumed that the reference fluid and test fluid behaved as Newtonian fluids. Both fluids had the same value, as μ_ref = μ_test = 1 cP. The interface between both fluids was relocated by varying the flow rate ratio of the reference fluid to test fluid. As shown in Figure A2A (Appendix A), when the interface moved from center line (α_x = 0.5) to each wall (i.e., α_x = 0 or 1), the blood viscosity without the correction factor (i.e., μ_n) showed a large deviation when compared with the viscosity of the test fluid (μ_test = 1 cP). Considering that the viscosity of the test fluid should have a constant value of μ_test = 1 cP with respect to the interface, the correction factor (C_f) could be obtained by reciprocating μ_n with respect to α_x (i.e., C_f = 1/μ_n). By conducting a regression analysis, the variations of the correction factor with respect to the interface were obtained as $C_f= 9.7212{{\alpha }_x}^4-19.442{{\alpha }_x}^3+ 15.687{{\alpha }_x}^2-5.9659{\alpha }_x+ 1.8992$ (R² = 0.9968). As shown in Figure A2B (Appendix A), to validate C_f, the viscosities of the test fluid were given as (a) μ_test = 1 cP and (b) μ_test = 4.08 cP. The flow rates of both fluids were the same, at 1 mL/h (i.e., Q_ref = Q_test = 1 mL/h). By applying the correction factor, the viscosities of the test fluids were determined as 1 cP and 4.14 cP, respectively. From the results, it was found that the parallel co-flowing method with the correction factor had the ability to measure the viscosity of a test fluid within 1.4% of a normalized difference. However, at t > T_st, the continuous ESR inside the ACS caused unstable behaviors in blood flows. To quantify the ESR, two indices (i.e., I_ESR, T_ESR) were suggested from <I_B> and T_st, as shown in Figure 5. Based on previous studies [45,50], one ESR index (I_ESR) was suggested simply by integrating <I_B> from t = T_st to t = T_end (i.e., $I_{ESR}={\int }_{t=T_{st}}^{t=T_{end}}<I_B>dt$). Here, T_end represented the end time of each experiment. Additionally, T_ESR = T_st − T_i. T_i indicated the initial time when the blood sample started to fill the blood channel.

From the preliminary demonstration, four factors (<U_B>, <U_R>, α_B, and <I_B>) could be effectively employed to obtain the blood viscosity and ESR when two ACSs were employed to deliver the blood sample and reference fluid into a microfluidic device.

2.4. Selection of Base Solution in Reference Fluid

To visualize the velocity fields of the reference fluid, RBCs were added into the reference fluid as fluid tracers. Glycerin solution was suggested as a reference fluid to minimize the sedimentation of RBCs inside the ACS. According to a previous study [51], the density (ρ) and viscosity (μ) increased at higher concentrations of glycerin solution as shown in Figure A3A (Appendix A). To evaluate the sedimentation rate of the RBCs added into the reference fluid, a simple ESR tester was prepared by using a disposable syringe (~1 mL) as shown in Figure A3C (Appendix A). The disposable syringe was fitted vertically into a hole (outer diameter = 4 mm) of the PDMS block. The outlet of the hole was closed with 3M adhesive tape. The syringe was filled with a specific concentration of glycerin solution (~0.5 mL). A 50 μL RBCs droplet was dropped into a specific concentration of glycerin solution. To monitor the sedimentation rate of the RBCs droplet in the simple ESR tester, snapshots were captured at an interval of 1 s with a smartphone camera (Galaxy A5, Samsung, Korea). As shown in Figure A3B (Appendix A), temporal variations of sedimentation height (H) were obtained by varying the concentration of the glycerin solution (C_glycerin) (C_glycerin = 5%, 10%, 20%, 30%, and 40%). Figure A3C (Appendix A) showed sedimentation of the RBCs droplet in 30% glycerin solution over time (t) (t = 0, 156, 192, 249, 259, 270, and 275 s). From the results, the RBCs droplet in the 40% glycerin solution remained nearly identical at the upper position, even without sedimentation. Furthermore, considering that the densities of normal RBCs range from 1090 kg/m³ to 1106 kg/m³ [52], the reference fluid was selected as a 40% glycerin solution (C_glycerin = 40%), because its density was greater than that of the RBCs.

2.5. Statistical Analysis

The statistical significance was evaluated by conducting statistical analyses with a commercial software package (Statistical Package for the Social Sciences (SPSS) Statistics version 24, IBM Corp., Armonk, NY, USA). Two ESR indices (I_ESR, T_ESR) and blood viscosity (<μ_B>) obtained by the present method were compared with results reported in a previous study (i.e., blood viscosity: μ_B, ESR index: S_EAI). An analysis of variance (ANOVA) test was applied to verify significant differences between comparative results. A linear regression analysis was conducted to verify the correlations between two parameters. All experimental results were expressed as mean ± standard deviation. If the p-value was less than 0.05, the experimental results exhibited significant differences within a 95% confidence interval.

3. Results and Discussion

3.1. Contribution of RBCs Added into Reference Fluid to Viscosity and Velocity Fields

To evaluate the effects of the RBCs added into the reference fluid on fluid viscosity, the viscosity of the reference fluid was measured by varying the volume of the RBCs added to the reference fluid (i.e., hematocrit [Hct]). In that regard, a 1× PBS was delivered to the blood channel (i.e., left-side channel) at a constant flow rate of Q_PBS = 1 mL/h with a syringe pump (SP) (neMESYS, Centoni Gmbh, Germany). The hematocrit (Hct) of the reference fluid was adjusted to Hct = 3%, 5%, 7%, and 9% by adding normal RBCs into the 40% glycerin solution. The reference fluid was delivered to the reference channel (i.e., right-side channel) at a constant flow rate of Q_glycerin = 1 mL/h with an SP. Figure 2A showed microscopic images for evaluating the interfacial location in the co-flowing channel with respect to Hct ((a) Hct = 0%, (b) Hct = 3%, (c) Hct = 5%, (d) Hct = 7%, and (e) Hct = 9%). To verify the contribution of the hematocrit in the reference fluid to the velocity fields (U_R), the velocity fields of the reference fluid were obtained across the reference channel width with respect to Hct. As shown in Figure 2B-a, a variation of the velocity profile (U_R) was obtained across the reference channel width with respect to Hct. The inset showed the microscopic image and velocity profile of the reference fluid with Hct = 3%. From the results, the velocity profile did not show a distinctive difference depending on the hematocrit. Figure 2B-b showed variations of the averaged velocity of the reference fluid (<U_R>) with respect to Hct. The hematocrit in the reference fluid did not contribute to varying <U_R> significantly. As shown in Figure 2C-a, the variations of the interface in the co-flowing channel (α_R) and the viscosity (μ_R) were obtained with respect to Hct. The interface and viscosity remained constant as α_R = 0.771 ± 0.003 and μ_R = 3.868 ± 0.068 cP for the reference fluid, with up to 9% hematocrit. From the results, it could be observed that providing up to a 9% hematocrit in the reference fluid did not significantly contribute to increasing the viscosity of the reference fluid. In addition, as shown in Figure A3A (Appendix A), an empirical formula [51] indicated that a 40% glycerin solution without any RBCs had a viscosity value of 4.07 cP. Based on the parallel co-flowing method with the correction factor [32], the viscosity of the reference fluid was measured consistently within a 5% difference when compared with the empirical formula. Furthermore, a previous flow-switching method [14] was employed to measure the viscosity of the reference fluid with respect to hematocrit. The inset of Figure 2C-b showed reversal flow-switching in the junction channel for the reference fluid with Hct = 7%. By increasing the flow rate of the 1× PBS (Q_PBS) from Q_PBS = 1 mL/h to Q_PBS = 3.1 mL/h, the hydrodynamic balancing in both side channels caused to reverse flow direction from left direction to right direction (i.e., reversal flow-switching phenomena) [14]. In other words, the junction channel was filled with blood at Q_PBS = 1 mL/h. However, it was filled with 1× PBS at Q_PBS = 3.1 mL/h. Based on the viscosity formula suggested in the flow-switching method, the viscosity of the reference fluid was quantified as μ_R = 3.1 ± 0.05 cP. As shown in Figure 2C-b, the viscosity obtained by both methods remained stable, with respect to hematocrit. Similar to the case in the parallel co-flowing method with the correction factor, the results of the flow-switching method indicated that the RBCs added into the reference fluid did not contribute to varying the viscosity within 9% hematocrit. The viscosity obtained by the flow-switching method was underestimated by approximately 20% when compared with that obtained by the parallel co-flowing method with the correction factor. From these results, in this study, the reference fluid was prepared by adding normal RBCs (Hct = 7%) into a 40% glycerin solution throughout all experiments.

3.2. Relationship between Flow Rate of Syringe Pump and Averaged Velocity Obtained by μPIV

To obtain blood viscosity, the flow rates of the blood sample and reference fluid should be measured from the averaged velocity obtained by conducting the µPIV technique. In other words, there was a need to obtain the relationship between the flow rate delivered by the SP (Q_sp) and the averaged velocity obtained by conducting the µPIV technique (<U>).

The hematocrit of the blood sample was adjusted to Hct = 30%, 40%, and 50% by adding normal RBCs into the base solution (1× PBS, plasma). Using two SPs, the flow rate of each fluid decreased stepwise from Q_sp = 1.5 mL/h to Q_sp = 0.1 mL/h, at an interval of 0.2 mL/h. With respect to each flow rate, the SP had been operated for 8 min. The blood sample was prepared by adding normal RBCs into plasma. As shown in Figure 3A-a, temporal variations of the averaged velocity (<U_B>) and the flow rate of the SP (Q_sp) were obtained by varying the hematocrit. At a higher flow rate of Q_sp, the hematocrit contributed to decreasing <U_B>. At a lower flow rate of Q_sp, <U_B> remained constant, without contribution from the hematocrit. By changing the base solution from plasma to a 1× PBS, temporal variations of <U_B> and Q_sp were obtained with respect to Hct. As shown in Figure 3A-b, the variations of <U_B>, with respect to hematocrit, were very similar to those of a blood sample composed of plasma. By averaging <U_B> with respect to Q_sp, <U_B> was quantified as mean ± standard deviation with respect to Q_sp. To determine the relationship between <U_B> and Q_sp, a scatter plot was used to plot <U_B> on a vertical axis and Q_sp on a horizontal axis. Figure 3A-c showed variations of <U_B> with respect to Q_sp and Hct in a blood sample composed of plasma. For example, <U_B> was estimated as about 30 mm/s for Q_sp =1.3 mL/h. Based on formula of flow rate (i.e., Q_μP = <U_B>A_c, A_c = w x h), flow rate obtained by μPIV was estimated as Q_μPIV = 0.54 mL/h. When compared with Q_sp =1.3 mL/h, the normalized difference between Q_sp and Q_μPIV was estimated as 59%. In this study, instead of the flow rate formula, the flow rate of the blood sample or reference fluid was estimated from the calibration formula obtained in advance. Thus, it was necessary to determine the relationship between velocity (<U_B>) and Q_sp. A regression analysis was conducted by assuming the regression formula as a quadratic model. Regression formulas between <U_B> and Q_sp with respect to Hct were obtained, as shown inside of Figure 3A-c. The regression formulas for each hematocrit were obtained as <U_B> = −5.027 Q_sp² + 30.279 Q_sp (R² = 0.998) for Hct = 30%, <U_B> = −6.262 Q_sp² + 30.660 Q_sp (R² = 0.999) for Hct = 40%, and <U_B> = −5.916 Q_sp² + 29.137 Q_sp (R² = 0.999) for Hct = 50%. Figure 3A-d showed variations of <U_B> with respect to the Q_sp and Hct in a blood sample composed of 1× PBS. From the regression analysis, as shown inside Figure 3A-d, the regression formulas for each hematocrit were obtained as <U_B> = −4.850 Q_sp² + 30.791 Q_sp (R² = 0.998) for Hct = 30%, <U_B> = −7.897 Q_sp² + 33.519 Q_sp (R² = 1.000) for Hct = 40%, and <U_B> = −5.717 Q_sp² + 29.286 Q_sp (R² = 0.999) for Hct = 50%. For the same hematocrit, the base solution (i.e., plasma or 1× PBS) did not contribute to varying the coefficients of the quadratic formula (i.e., normalized difference < 4% except Hct = 40%). However, for the same base solution, the coefficients of a quadratic model varied significantly with respect to hematocrit.

A regression formula between Q_sp and <U_R> for the reference fluid (i.e., 40% glycerin solution) with RBCs (Hct = 7%) was obtained by using a similar procedure to that used for the blood sample. Figure 3B-a showed the temporal variations of Q_sp and <U_R> for the reference fluid. <U_R> was obtained as a mean ± standard deviation for a corresponding Q_sp. When compared with the blood sample, <U_R> increased substantially, owing to the lower value of the hematocrit. As shown in Figure 3B-b, variations of <U_R> with respect to Q_sp were represented by a scatter plot. From a regression analysis, the regression formula between <U_R> and Q_sp was obtained as <U_R> = −7.770 Q_sp² + 37.127 Q_sp (R² = 0.9875).

From the results, the coefficients of the quadratic formula were varied significantly with respect to hematocrit. However, the base solution (i.e., 1× PBS, or plasma) did not contribute to changing the coefficients of the regression formula. Using regression formulae between Q_sp and <U_R> (or <U_B>) obtained in advance, the <U_R> or <U_B> obtained by conducting the µPIV technique was converted into a flow rate (i.e., Q_B, Q_R, respectively).

3.3. Quantitative Comparison of Blood Viscosity with Respect to Fluid Delivery System (ACS, SP)

Since the <U_R> and <U_B> obtained from the µPIV technique were converted into flow rates (Q_R and Q_B) from regression formulae obtained in advance, the blood viscosity could be measured by monitoring the interface (α_B) in the co-flowing channel, under fluid delivery with an ACS. The blood viscosity obtained by the ACS was quantitatively compared with one obtained by an SP. Blood samples (Hct = 30%, 40%, and 50%) were prepared by adding normal RBCs into the base solution (1× PBS, plasma).

Figure 4A-a showed the temporal variations of Q_R, Q_B, and α_B for the blood sample (normal RBCs in 1× PBS, Hct = 50%). In addition, Figure 4A-b depicted the temporal variations of Q_R, Q_B, and α_B for the blood sample (normal RBCs in plasma, Hct = 50%). Using the blood viscosity formula, the blood viscosity was obtained by using the temporal variations of Q_R, Q_B, and α_B. Here, the viscosity of the reference fluid was given as μ_R = 4.08 cP by using measurement results reported in previous studies [14,51]. For a rectangular channel (width = w, depth = h) with a lower aspect ratio [32], the formula of shear rate ($\dot{\gamma }$) was given as approximately $\dot{\gamma }=\frac{6Q_B}{w{h}^2}$. The corresponding shear rate of the blood viscosity obtained at a specific blood flow rate (Q_B) was estimated by using the shear rate formula. A scatter plot was employed to plot μ_B on a vertical axis, and $\dot{\gamma }$ on a horizontal axis. As shown in Figure 4B-a, variations of μ_B were obtained with respect to the shear rate under fluid delivery with the two ACSs. Here, the blood sample (Hct = 50%) was prepared by adding normal RBCs into plasma or 1× PBS. The blood sample has behaved as a Newtonian fluid at sufficiently higher shear rates ($\dot{\gamma }>{10}^3 s^{-1}$). From the experimental results, μ_B remained constant with respect to the shear rate. By conducting an arithmetic average of μ_B over specific shear rates, the blood viscosity was expressed as <μ_B> = mean ± standard deviation. The viscosity of the blood sample composed of plasma (<μ_{B, plasma}> = 2.381 ± 0.042 cP) was significantly higher than that of the blood sample composed of 1× PBS (<μ_{B, PBS}> = 1.845 ± 0.0573 cP). To compare with the blood viscosity obtained under fluid delivery with two ACSs, the same blood samples were employed to measure the blood viscosity under fluid delivery with two SPs. Two fluids (blood sample, reference fluid) were delivered to each inlet of the microfluidic device, at the same flow rate (Q_B = Q_R).

As represented in Figure 3A-a and Figure 3A-b, the flow rate of the SP (Q_sp) decreased stepwise from Q_sp = 1.5 mL/h to Q_sp = 0.1 mL/h at an interval of 0.2 mL/h. Each flow rate had been maintained for 8 min. As shown in Figure 4B-b, variations in μ_B of the blood samples (normal RBCs in plasma and 1× PBS, Hct = 50%) were obtained with respect to the shear rate. The blood viscosity remained constant with respect to the shear rate. The viscosity of the blood sample composed of plasma (<μ_{B, plasma}> = 2.728 ± 0.0918 cP) was higher than that of the blood sample composed of 1× PBS (<μ_{B, PBS}> = 2.109 ± 0.0429 cP). When compared with the blood viscosity obtained by the ACS, blood viscosity obtained by the SP increased by approximately 12.5%. To determine the effects of hematocrit on blood viscosity, variations of blood viscosity were obtained by varying the hematocrit (Hct = 30%, 40%, and 50%), base solution (1× PBS, plasma), and fluid delivery system (ACS, SP). Figure 4B-c showed the variations of <μ_B> with respect to Hct, base solution, and the fluid delivery system. Under fluid delivery with an ACS, <μ_B> tended to increase with respect to Hct. Under fluid delivery with an SP, there was no significant difference between Hct = 30% and Hct = 40%. The blood viscosity increased at Hct = 50% when compared with Hct = 30% or 40%. To determine the correlation between the blood viscosity obtained by the ACS (<μ_{B, ACS}>) and the blood viscosity obtained by the SP (<μ_{B, SP}>), a scatterplot was used to plot <μ_{B, ACS}> on a vertical axis, and <μ_{B, SP}> on a horizontal axis, as shown in Figure 4B-d. According to a linear regression analysis, <μ_{B, ACS}>was expressed as <μ_{B, ACS}>= 0.5524 <μ_{B, SP}> + 0.7248 (R² = 0.7037, p-value = 0.037). Here, p-value = 0.037 indicated that a linear regression showed sufficient relationship between two viscosity values (i.e., <μ_{B, ACS}> vs. <μ_{B, SP}>). In addition, R² was obtained as a high value of R² = 0.7037. Although two SPs were effectively used to deliver two fluids during measurement of blood viscosity, the arrangement included challenges, such as a bulky size and a high cost. From the correlation between <μ_{B, ACS}> and <μ_{B, SP}>, it was found that the ACS can be effectively employed to deliver two fluids in the measurement of blood viscosity. Thus, the blood viscosity can be measured consistently under fluid delivery with two ACSs.

3.4. Quantitative Measurement of ESR with Respect to base Solution and Hematocrit

The ESR of the blood sample was evaluated by quantifying the microscopic image intensity of the blood sample (<I_B>) flowing in the blood channel. Two ESR indices (t_ESR, I_ESR) were suggested by quantifying the temporal variations of <I_B>. The blood samples (Hct = 30%, 40%, and 50%) were prepared by adding normal RBCs into a base solution (1× PBS, plasma).

As shown in Figure 5A-a, variations of <I_B> for the blood sample (normal RBCs in plasma) were obtained with respect to Hct. <I_B> tended to decrease at higher values of Hct. In addition, T_st tended to be shorter at lower values of the hematocrit. To exclude the contribution of plasma protein to the ESR, the plasma was replaced with the 1× PBS. As shown in Figure 5A-b, temporal variations of <I_B> for the blood sample (normal RBCs in 1× PBS) were obtained by varying Hct. <I_B> tended to decrease at higher values of Hct. With a certain elapse of time, <I_B> remained constant. There was no existence of separation time within 2000 s (i.e., T_st > 2000 s). The results indicated that the 1× PBS did not sufficiently contribute to enhancing ESR when compared with plasma.

To quantify the ESR of the blood sample (normal RBCs in plasma) from <I_B> as shown in Figure 5A-a, two ESR indices (t_ESR, I_ESR) were obtained with respect to the hematocrit. Figure 5B-a showed variations of t_ESR and I_ESR with respect to Hct. According to the results, t_ESR tended to increase significantly with respect to hematocrit (p-value = 0.0004). I_ESR tended to decrease substantially with respect to hematocrit (p-value = 0.001). Under blood delivery with the ACS, the RBCs tended to fall down continuously inside the ACS, which was installed horizontally. Owing to the continuous ESR inside the ACS, the populations of RBCs delivered to the blood channel decreased over time. Thus, <I_B> increased gradually over time, as shown in Figure 5A-a. However, when the plasma was replaced with the 1× PBS, the blood sample did not exhibit an ESR inside the ACS. For this reason, after a certain period of time, <I_B> remained constant over time, as shown in Figure 5A-b. To quantitatively compare with results reported in a previous study, two indices (t_ESR, I_ESR) and S_EAI (previous ESR index) [45] were plotted on a vertical axis and horizontal axis, respectively. S_EAI exhibited larger scatters than t_ESR or I_ESR. From the regression analysis, the linear regression exhibited higher values of R² = 0.7474~0.7755. The results indicated that the two ESR indices exhibited consistent variations with respect to hematocrit when compared with S_EAI. Thus, the two ESR indices (t_ESR, I_ESR) can be effectively used to evaluate the variation of ESR with respect to hematocrit.

3.5. Variations of Blood Viscosity and ESR for Blood Samples Composed of Specific Dextran Solutions

A specific dextran solution as a base solution was prepared to enhance the ESR of the blood sample. To exclude the contributions of hematocrit to ESR, the hematocrit of the blood sample was adjusted to Hct = 50%. The blood samples were prepared by adding normal RBCs into specific concentrations of dextran solution (i.e., C_dex = 0, 5, 10, 15, and 20 mg/mL). C_dex = 0 meant 1× PBS as control. As shown in Figure 6A-a, temporal variations of Q_B were obtained with respect to C_dex. From the results, the blood sample composed of dextran solution (C_dex <= 5 mg/mL) exhibited stable variations of Q_B over time. However, above C_dex >= 10 mg/mL, the separation time (T_st) tended to reduce at higher concentrations of the dextran solution. Figure 6A-b showed temporal variations of α_B with respect to C_dex. Similar to the case with Q_B, α_B varied unstably above C_dex >= 10 mg/mL. T_st tended to be shorter at higher concentrations of the dextran solution. By using stable variations of Q_B and α_B obtained at t < T_st, variations of μ_B were obtained with respect to the shear rate. As shown in Figure 6A-c, blood viscosities were obtained at sufficiently higher shear rates ($\dot{\gamma }>{10}^3 s^{-1}$). They remained constant with respect to the shear rate. Additionally, the blood viscosity tended to increase at higher concentrations of the dextran solution. By averaging the μ_B values obtained at shear rates, the blood viscosity was expressed as <μ_B> = mean ± standard deviation. Figure 6A-d showed variations of <μ_B> with respect to C_dex and the fluid delivery system (ACS, SP). When compared with the results reported in a previous study [45], the present results exhibited sufficiently consistent variations of <μ_B> with respect to C_dex. In addition, there was no significant difference between the ACS and SP. As shown in Figure 6A-e, to determine the correlation between <μ_B> obtained by the proposed method (two ACSs) and μ_B obtained by the previous method (two SPs) [45], a scatter plot was used to plot <μ_B> (proposed method) on a vertical axis and μ_B (previous method) on a horizontal axis. According to linear regression analysis, the high value of R² = 0.9767 indicated that the proposed method could give comparable results when compared with the previous method. Thus, ACSs could be effectively employed to deliver fluid samples. After measuring the blood viscosity with respect to C_dex, the contributions of the dextran solution to the ESR were evaluated by quantifying the image intensity of the blood sample flowing in the blood channel (<I_B>). As shown in Figure 6B-a, temporal variations of <I_B> were obtained with respect to C_dex. As a result, T_st was reduced at higher concentrations of the dextran solution. Using <I_B> with respect to C_dex, two ESR indices (t_ESR, I_ESR) were obtained with respect to C_dex. As shown in Figure 6B-b, the ESR indices remained constant up to C_dex = 5 mg/mL. Above C_dex >= 10 mg/mL, t_ESR tended to decrease substantially with respect to C_dex (p-value = 0.0001). I_ESR increased largely at higher concentrations of dextran solution (p-value = 0.0001). To compare with the results reported in a previous study [45], a scatterplot was used to plot t_ESR and I_ESR (i.e., proposed method) on a vertical axis, and S_EAI (i.e., previous method: periodic on-off control with an SP) on a horizontal axis. As shown in Figure 6B-c, a linear regression analysis was conducted to determine the correlation between the proposed method and the previous method. The higher value of R² = 0.8202–0.8548 indicated that the two ESR indices (t_ESR, I_ESR) gave comparable results when compared with the previous method. Thus, the ESR indices can be effectively used to quantify the ESRs of blood samples.

3.6. Variations of Blood Viscosity and ESR for Blood Samples Composed of Hardened RBCs

As the last demonstration, the proposed method was applied to evaluate the contribution of hardened RBCs to the ESR. As shown in Figure 5A-a, a blood sample (Hct = 50%) composed of plasma did not contribute to variations in the ESR. To stimulate the ESR of a blood sample with a high value of Hct = 50%, the plasma as a base solution was replaced with a specific concentration of dextran solution (i.e., C_dex = 15 mg/mL). Hardened RBCs were prepared by sufficiently exposing normal RBCs to specific concentrations of GA solution (C_GA) (i.e., C_GA = 0, 5, 10, and 15 μL/mL). C_GA = 0 indicated normal RBCs as control. The blood samples (Hct = 50%) were then prepared by adding hardened RBCs into the specific dextran solutions.

Figure 7A showed the temporal variations of Q_B with respect to C_GA. T_st tended to increase at higher concentrations of the GA solution. At C_GA = 15 μL/mL, Q_B tended to decrease stably over time. Figure 7B showed the temporal variations of α_B with respect to C_GA. The variations of α_B were very similar to those of Q_B. At C_GA = 15 μL/mL, α_B remained constant after a certain period of time. Figure 7C showed the temporal variations of <I_B> with respect to C_GA. Except for at C_GA = 15 μL/mL, <I_B> tended to increase stably over time. Additionally, T_st tended to increase at higher concentrations of the GA solution. By measuring three factors (Q_B, α_B, and <I_B>) simultaneously, the hardened blood sample composed of hardened RBCs fixed with C_GA = 15 μL/mL did not exhibit an ESR inside the ACS. Thus, there were no significant variations of Q_B and α_B. As shown in Figure 7D, variations of <μ_B> were obtained with respect to C_GA. <μ_B> tended to increase considerably with respect to C_GA. As the GA solution contributed to stiffening the RBCs’ membranes, it was reasonable that the blood viscosity increased at higher concentrations of the GA solution. Figure 7E showed variations of the two ESR indices (t_ESR, I_ESR) with respect to C_GA. t_ESR tended to increase with respect to C_GA. I_ESR tended to decrease with respect to C_GA.

From the experimental results, it was found that the GA solution caused an increase in blood viscosity. Furthermore, the proposed method had the ability to consistently measure blood viscosity and ESR, under simultaneously fluid delivery from two ACSs.

4. Conclusions

In this study, a simple method of measuring blood viscosity and ESR was demonstrated by quantifying averaged velocities of a blood sample and reference fluid, where the blood sample and reference fluid were delivered to a microfluidic device with two ACSs. According to the experimental results, a 40% glycerin solution with RBCs (Hct = 7%) was selected as the reference fluid to obtain velocity fields and avoid sedimentation of RBCs in the ACS. Using a calibration formulae between the flow rate of an SP (Q_sp) and the averaged velocity obtained by the µPIV technique (<U_B>) in advance, <U_B> or <U_R> was converted into Q_B or Q_R, respectively. As a demonstration, the proposed method was employed to evaluate the contributions of the hematocrit (Hct = 30%, 40%, and 50%), base solution (1× PBS, plasma, dextran solution), and hardened RBCs to the blood viscosity and ESR, respectively. The results of the proposed method were comparable with those reported in previous studies that used two SPs. From the experimental results, it could be concluded that the proposed method had the ability to consistently measure blood viscosity and ESR under simultaneous fluid delivery with two ACSs. However, image acquisition for quantifying blood flows in microfluidic channels was demonstrated from the optical microscope and a high-speed camera. To resolve the issue, the proposed method should be improved substantially by adopting a portable image acquisition system in the near future.