1 Introduction

Blood, containing various cellular components and biological molecules, is the most important body fluids in clinical practice. The abnormalities of the physiological conditions as well as the quantity of cells or macromolecular biomarkers in blood could provide vast information for medical diagnostics. For example, the well-known sickle-cell anemia (SCA) is caused by the mutation of red blood cells (RBCs) from discoid-shaped ones to abnormal sickle-shaped ones (Stuart and Nagel 2004). The disorders (i.e., leukopenia or leukocytosis) of white blood cells (WBCs) commonly indicate immune disease, infection, or even blood cancer (Craddock et al. 1960). Therefore, blood tests ranging from simple glucose tests to cellular evaluations (e.g., complete blood count) or molecular tests (e.g., liver function test) are one of the most useful medical tests for disease diagnosis (Mach et al. 2013). For efficiently performing blood tests, the first and the most important pretreatment step is the rapid separation of blood components (Yu et al. 2014). In current research or clinical use, centrifugation, flow cytometry sorting and chemical lysis are still the “gold standard” for blood sample pretreatment (Heikali and Di Carlo 2010). However, these widely employed methods may have some problems for future point-of-care testing (POCT), such as expensive, time-consuming, labour-intensive, and large sample consumption.

The advent of microfluidics provides a potential approach for the rapid and low cost pretreatment of blood samples, and poses a new concept “Blood on a Chip (BOC)” which has attached increasing research interests in recent years (Toner and Irimia 2005). Up to now, a large number of novel prototype devices have been reported to expand the functions or application areas of BOC. However, the issue of low-throughput (typically less than 100 μl/min) existing in most of these devices prevents them from applying to continuous-flow POCT devices. Inertial microfluidics, which employs the inherent inertial effects (e.g., inertial migration and cross-sectional secondary flow) of microfluids, offers a significant throughput advantage over the most reported active or passive devices (Amini et al. 2014). Therefore, inertial microfluidic devices promise to be an ideal alternation for high-throughput and continuous blood processing. For example, the simplest straight channel devices have been applied to the removal of pathogens (e.g., Escherichia coli) from blood according to their different focusing modes (Mach and Di Carlo 2010; Wei Hou et al. 2012) or the sheathless ordering of blood cells for microflow cytometers (Hur et al. 2010). However, due to the monotonous equilibrium positions (i.e., four positions for square channels or two positions for rectangular channels), the straight channel is not suited for separating particles/cells of different sizes. Through introducing microcavities into the straight channel design, the planar vortex or cross-sectional secondary flow induced via microcavities has been used for isolating rare circulating tumor cells (CTCs) (Lee et al. 2013) or blood plasma (Lee et al. 2014) from human whole blood. However, in these devices, an additional sheath flow is required to pinch the sample flow towards the channel side wall for successful isolation, which makes the flow control more complex and further limits its processing throughput to only ml/hour level.

Besides the introduction of microcavities, another simple approach for inducing secondary flow is the bending of channels (Berger et al. 1983). The coupling of inertial migration effects and cross-sectional secondary flow in curved inertial microfluidics enables the manipulation modes to be more flexible and powerful, and further reduces the number of final equilibrium positions. Up to now, various curved channel geometries (e.g., simple arc, sinusoid, and spiral) have been reported for the processing of blood samples (Martel and Toner 2014). Among these channel geometries, the spiral channel has attracted increasing research interests due to its compact footprint and abundant particle migration modes. For example, a Dean Flow Fractionation (DFF) chip with two inlets and two outlets has been developed to realize the isolation of rare CTCs from blood (Hou et al. 2013), and it has been commercialized by the Clearbridge BioMedics for the early detection of cancer metastasis. To enhance the isolation performance, several efforts have been made to optimize the channel geometry from single spiral to double spiral (Sun et al. 2012) or optimize the channel cross-section from rectangular to trapezoidal (Warkiani et al. 2014). In addition to CTCs isolation, the spiral channel device has also been employed for the separation of leukocytes (Wu et al. 2012; Nivedita and Papautsky 2013) or bacteria (Hou et al. 2015) from blood. Although great success has been achieved in the applications of curved inertial microfluidics for blood processing, the focusing dynamics of complex blood cells in spiral channels are still not well understood. Whichever the final application is, the underlying physics of blood cell focusing are always the most basic and most important issue. Improved understanding of the inertial focusing of blood cells would be valuable for guiding the design and operation of inertial microfluidics.

Herein, we investigated the migration dynamics of human blood cells in a spiral inertial microfluidic device in which the blood cells would be focused to form a narrow cell band under the complex coupling of inertial lift force and Dean drag force. The cell focusing dynamics under various flow rates were first investigated to uncover the competition between the effects of the two dominant forces. Then, the migration dynamics of blood cells with different concentrations were explored to study the steric crowding effects on cell focusing. Given the above understandings of blood cell focusing, the efficient isolation of plasma from 20-fold diluted blood was successfully performed under a throughput of 700 μl/min. The isolation performance was then compared with the previously-reported devices to embody the throughput advantage of our device.

2 Materials and methods

2.1 Blood sample preparation

Human whole blood employed in this work was draw from a healthy consenting volunteer, and collected via a 4 ml vacutainer collection tube (BD Biosciences) containing anticoagulant K2EDTA. To ensure the viability of the cellular components, the blood sample was stored at 4 °C, and processed within 6 h after collection. The initial hematocrit (HCT) which indicates the volume percentage of RBCs in blood was determined to be 43.8 % through an automated analyzer in Zhongda Hospital (Nanjing, China). The blood sample was gently rocked back and forth before use, and was then diluted with different volumes of 0.9 % sodium chloride solution (Baxter) to prepare samples of different dilution ratios (different HCT levels).

2.2 Microdevice design and fabrication

The main channel in our spiral inertial microfluidic device was designed to be a five-loop Archimedean spiral with one inlet and two outlets on the ends. The aspect ratio (AR=height/width=h/w) of channel cross-section was determined to be about 1:3 for easy fabrication using the lithography process. The channel height was set to be 50 μm which enables the cellular components of blood to satisfy the focusing criterion (a p/h ≥0.07 (Kuntaegowdanahalli et al. 2009), where a p is the particle diameter). The initial radius of the spiral channel was fixed at 3.5 mm, and the distance between adjacent channels was 500 μm. Although the total length of our spiral channel was calculated to be about 16 cm, the footprint of the whole device was only 2 cm2. To realize efficient isolation of blood plasma under relatively high HCTs, a classic Y-shaped outlet system was adopted. The angle between the two bifurcated channels of 200 μm wide was set to be 85°. An expansion channel (2.4 mm long), which gradually increases the channel width from 150 to 300 μm, was used to connect the main spiral channel with the Y-shaped outlet system.

The microdevice was fabricated according to the procedure of standard soft lithography (Whitesides et al. 2001). First, a layer of 50 μm thick negative photoresist (SU-8 2050, MicroChem Corp.) was spun coating onto a clean 4” silicon wafer. Then, the resist film was selectively exposed with UV light via a mask aligner (SÜSS MicroTec AG) to transfer the spiral channel pattern in photomask onto the resist. Following the SU-8 developing step for washing the uncured resist, the obtained master mold was fixed in a 12 cm petri dish (Corning Inc.), and casted with the degassed liquid PDMS (Sylgard 184, Dow Corning Corp.) with a mixing weight ratios (10:1) of the base to curing agent. After having cured at 80 °C for 2 h, the PDMS block containing microchannels was carefully peeled off from the master mold, and cut into acceptable pieces. The orifices near the inlet and outlet were punched using a simple biopsy micropunch (0.75 mm, Harris Uni-Core). The oxygen plasma treatment (Harrick Plasma) was used to assist in irreversibly bonding the PDMS block with a clean glass slide.

2.3 Experimental set-up and operation

The fabricated device was mounted onto the observation platform of an invert microscope (IX71, Olympus). The prepared blood cell suspension was loaded into a 10 ml disposable syringe which was then equipped onto a programmable syringe pump (Legato 270, KD Scientific, Inc.) for providing the sample with a stable and precise flow rate. Three pieces of PEEK tubing (outer diameter of 1/32”, Upchurch Scientific) were used to connect the device orifices with the syringe or collecting tubes for sample introduction or exportation. Before pumping through the blood sample, the microchannel was pre-wetted with the sodium chloride solution to remove microbubbles. During the experiments, the syringe was periodically rocked to prevent blood cells from undesirable sedimentation. Fast-moving blood cells were visualized through an equipped 10× objective (NA = 0.3, Olympus) and a high-speed CCD camera (Exi Blue, QImaging). The real-time motions of individual cells under an extremely low exposure time of 1 μs were then captured using the supporting IMAGE-PRO Express software, and were saved in multimedia files such as videos or image sequences. To statistically illustrate the distributions of flowing blood cells over a certain period, over 300 discrete images were stacked in Z direction using free IMAGEJ software (http://rsb.info.nih.gov/ij/), creating a composite image of which each pixel contains the maximum value over all images in the stack at that particular pixel location.

3 Basic physics and principle

In common wisdom, the flows in microfluidic channels of micrometer dimensions are assumed to occur at very low channel Reynolds numbers ranging from ο(10−6) to ο(10), and often incorrectly approximated as the Stokes flow through neglecting the inertia element in Navier–Stokes (N-S) equations (Squires and Quake 2005). The channel Reynolds number (Rec) describes the magnitude ratio of inertial force and viscous force, and can be calculated as Rec = ρUD h/μ, where the ρ is the density of fluids, U is the average velocity of fluids, D h is the hydraulic diameter of the channel cross-section, and μ is the dynamic viscosity of fluids. In the case of finite Rec channel flows, the inertia effects of flowing fluids cannot be neglected, and some interesting hydrodynamics phenomena are observed (Di Carlo 2009). The first is the lateral inertial migration of particles across the streamline, which is first observed by Segre and Silberberg (1961) in their macroscopic hydrodynamic experiments of particle migration in a cylindrical tube. After its first discovery in 1960s, plenty of theoretical investigations (Saffman 1965; Cox and Brenner 1968; Ho and Leal 1974; Schonberg and Hinch 1989; Asmolov 1999; Matas et al. 2004; Miura et al. 2014) were reported to uncover the underlying physics of this interesting particle lateral migration. By now it is well understood that this phenomenon is caused by the inertial lift force (F L∝ρU 2 a p 4 /D h 2), which is actually a force balance of shear-induced lift force (F LS) and wall-induced lift force (F LW) (Asmolov 1999; Matas et al. 2004). The parabolic velocity profile of channel flows along the channel width and height would induce a F LS which acts down the shear gradient, and forces the particle to move towards the channel walls. Instead, the particle-wall interaction would induce a F LW to push the particle away from the channel wall. The cross-sectional points where these two forces balance each other is the so-call equilibrium positions. In circular channels, the particles migrate to form a regular particle annulus, as observed in the macroscopic condition while in rectangular cross-section channels, the particles equilibrate at two positions near the center of long channel faces (Hur et al. 2010).

The second is the secondary flow (also named as Dean flow) which drags the flowing particle in the two counter-rotating vortices located at the top and bottom half of channel cross-section (Dean 1928). The strength of Dean vortices can be evaluated via another dimensionless parameter, Dean number (De = Rec(D h/2R)1/2, where R is the channel radius) (Berger et al. 1983). An additional Dean drag force would be acted on the particles, and its math expression can be described as F D∝ρU 2 a p D h 2 R −1 (Di Carlo et al. 2007). Therefore, the particles flowing in curved channels would be stable at the new equilibrium positions close to the inner channel wall under the complex coupling of F L and F D (Gossett and Di Carlo 2009; Amini et al. 2014).

4 Results and discussions

4.1 Flow-rate effects on blood cell focusing

To understand the migration behaviors of individual blood cells flowing in spiral inertial microfluidic devices, the cell-cell interactions (e.g., attractive or repulsive interactions) or steric crowding effects have to be avoided whenever possible. Therefore, the blood sample was advisedly diluted 1000-fold, and then brought into homogeneous suspensions with a HCT of 0.044 % by shaking gently. The diluted blood suspension was made to run through the fabricated device at the flow rates of 100~700 μl/min. The corresponding Dean numbers (Des) calculated using an initial channel radius of R = 3.5 mm are De = 1.73~12.08. Snapshots of blood cell migration near the Y-shaped outlet under different Des were continuously recorded and stacked using the above-mentioned methods. The obtained composite images illustrating the statistical distribution of cells across the channel width were then grouped in Fig. 1(a). From this figure, it is found that at the lowest tested flow rate of De = 1.73, blood cells are unable to reach the stable equilibrium positions due to the weak lateral forces acting on cells, and thus no obvious cell focusing was observed except a small cell-free region near the inner wall. The reason for the generation of cell-free region is that the relatively stronger Dean dragging effect near the inner wall drags the cells into the circulation regions of Dean vortices (Gossett and Di Carlo 2009). With further increasing De, blood cells near the outer wall tend to migrate fast into the inner channel region due to the enhanced Dean dragging and inertial lift effects at high flow rates (Di Carlo et al. 2007; Zhang et al. 2014). When the De is increased to be larger than 5.18, nearly all the flowing blood cells are focused at the equilibrium positions, forming a narrow cell band close to the inner wall, as illustrated in the images of Fig. 1(a).

Fig. 1
figure 1

a Composite images illustrating the distributions of blood cells near the Y-shaped outlet system under different Des. b Variations of cell distribution width and lateral focusing position as a function of Des

Full size image

To illustrate the flow-rate effects on the migration dynamics of complex blood cells, the distribution width and lateral focusing position of blood cells across the channel width under different Des were quantitatively measured. Form the results illustrated in Fig. 1(b), it is found that with the increase of flow rate, the lateral focusing position of blood cells first shifts towards the inner wall until a well-focused cell band is formed, and then slowly shifts back towards the channel centerline. Meanwhile, the width of cell bands was also observed to vary accordingly. Specifically, it decreases sharply during the focusing process, achieves a constant value after all cells are well-focused, but slightly increases at the later flow rate stage due to the Dean mixing effect which is commonly regarded as a beneficial effect for sample mixing or reaction under the laminar flow condition (Sudarsan and Ugaz 2006). These bidirectional changes of the width (decrease-increase) and the position (inward-outward) of cell focusing band can be well-explained using the five-stage process model concluded from experiments on the migration dynamics of rigid microspheres (Xiang et al. 2013). The rapid-narrowing and inward-shifting of the cell band in the first stage result from the dominant inertial lift force directing towards the inner wall. With De increasing to a specific threshold value, the Dean drag force finally exceeds that of the inertial lift force in strength (Kuntaegowdanahalli et al. 2009), and dominates the lateral migration of cells, resulting in the widening and outward-shifting of cell bands at higher flow rates. It is worth noting that the fine turning of cell focusing quality or position can be realized through careful regulation of the flow rate after the focusing process is completed. The best focusing performance occurs in the transition stage (at a middle flow rate of De = 6.90) where the flowing cells are exposed to the complex coupling of both inertial lift force and Dean drag force. The indicator for achieving best focusing is that the centerline of the cell band reaches the position closest to the inner wall. However, the perfect single-line particle array observed in previous studies (Bhagat et al. 2008; Kuntaegowdanahalli et al. 2009; Russom et al. 2009; Sun et al. 2012; Zhou and Papautsky 2013; Liu et al. 2015) regarding inertial migration of rigid microspheres was not achieved probably due to the non-spherical and polydisperse-sized features of blood cells. The rotation of discoid RBCs and the differential equilibriums of polydisperse-sized blood cells of various types make the focusing band of blood cells obviously wider than that of rigid particles (the smallest cell-band width is measured to be ~25 μm which is about 3~4 times the average diameter of blood cells).

The finding that nearly all the flowing blood cells migrate to occupy their lateral positions in a relative narrow cell band located within the inner channel region provides a probability of using our spiral inertial microfluidic device to realize blood plasma separation. Although the high flow rate would cause the cell focusing performance to be worse, an alluring high processing throughput can be offered by increasing the flow rate. Considering the importance of high-throughput in dealing with large-volume bodily fluids for future commercial applications, the processing flow rate for the following experiments was selected to be the highest safe value of 700 μl/min. Further increasing the flow rate would cause obvious deformation of the PDMS channel or even leakage near the inlet. The possible method for addressing this problem is to fabricate our devices using other rigid materials (e.g., PMMA, glass and so on) which can resist high-pressure injections (Sollier et al. 2011). Another important consideration is whether the over-high shear stress induced at high flow rates would have negative effects on the viability of tested blood cells. For example, the RBCs may lyse and release the hemoglobin into the blood plasma (called “hemolysis”) when flowing under sufficiently high flow rates. To calculate the cell loss caused by the high flow rate, the numbers of collected samples were counted and compared with those of initial samples. The result shows that over 98 % blood cells are recovered after running under the maximum driving flow rate of 700 μl/min, which indicates that the negative effects of shear stress on cell viability could be neglected. In addition, the flow condition (Rec = ~100, the main velocity of flowing blood cells U = ~1.5 m/s) under the highest tested flow rate has been proved to be safe for processing blood cells via a number of previously-reported biological tests (Lokhandwalla et al. 2001; Mach and Di Carlo 2010).

4.2 Variation of focusing statuses of RBCs along the channel height

Another miraculous finding is the dependence of cell focusing status on channel height. Figure 2(a) is a snapshot of the focused-blood-cell array near the Y-shaped outlet system at the selected flow rate of 700 μl/min. A closer inspection of this cell array shows that the aligned blood cells appear as two images with different pixel intensities (illustrated in insets (i) and (ii) of Fig. 2(a)). One possible explanation for that is the variety of blood cell types (i.e., RBCs of red color and WBCs of white color). However, this possibility could be safely eliminated since these two image modes were found to have nearly equal numbers of cells while the ratio of RBCs to WBCs in healthy human blood is over 500:1. A more reasonable explanation is that these two images indicate two focusing statuses of the discoid RBCs, that is, rotational focusing of these abundant RBCs was observed with the disk plane parallel to the bottom channel wall (flat focusing status (i) in Fig. 2(a), the bright one) or perpendicular to the bottom channel wall (vertical focusing status (ii) in Fig. 2(a), the dark one), which is also confirmed by the observed cell shapes (sphere or ellipse) in the enlarged snapshot illustrated in Fig. 2(a). Hence, the difference in the shading of flowing blood cells between the two focusing statuses is attributed to the difference in the thickness of the aligned RBCs along the channel height (T vertical = ~6 μm vs. T flat = ~3 μm). To further characterize the focusing statuses of the RBCs along the channel height, we carefully turned the focus plane of the microscope from the bottom channel wall to the top one while continuously capturing image at various channel heights. Selected snapshots at different channel heights are listed in Fig. 2(b), in which blood cells go from out-of-focus, to in-focus and to out-of-focus again. From this figure, it is interesting to find that the focusing status of the aligned RBCs changes from the vertical attitude to the flat attitude. A transition stage with a proportion of RBCs being in the flat attitude and the rest being in the vertical attitude was found in a position close to the channel middle plane. The dependence of focusing status on channel height is also confirmed at other flow rates tested in our experiments (data not show). The detailed mechanism for this specific phenomenon needs to be further investigated.

Fig. 2
figure 2

a Snapshot of focused-blood-cell array near the Y-shaped outlets at the flow rate of 700 μl/min. Two focusing statuses of aligned RBCs were observed with the disk plane parallel to the bottom channel wall (flat focusing status (i)) or perpendicular to the bottom channel wall (vertical focusing status (ii)). b Snapshots of blood cell focusing along the channel height. The scale bar is 100 μm

Full size image

4.3 Concentration effects on blood cell focusing

In the above experiments, an extremely low cell concentration was applied to study the migration dynamics of blood cells. However, the problem of large processing volume brought about by high dilution ratio flings down a challenge to the practical applications of the device. It seems inconceivable to process or even simply transport such a huge volume of blood sample after 1000-fold dilution (e.g., a typical volume of 1~10 ml sampled in clinic would dilute to a final unimaginable volume of 1000~10,000 ml). At the other extreme, the significant interactions between adjacent cells in low-dilution conditions would degrade the focusing performance, which may become a problem for cell focusing or counting applications requiring high-quality or even single-line focusing (Xuan et al. 2010). To better guide the operation of our device, we investigated the concentration effect on focusing dynamics of blood cells through pumping blood samples with different dilution ratios (10-fold~1000-fold) into the fabricated device. Figure 3(a) illustrates the distribution of blood cells just before the Y-shaped outlet at the maximum flow rate of 700 μl/min. From this figure, it is found that the width of cell bands sharply decreases with the increase of dilution ratio (or the decrease of HCTs). Blood cells migrate to form a narrow band only at a dilution ratio of larger than 500-fold. At low dilution ratios, cells are poorly focused because of the steric crowding effects and cell-cell interactions. Excessive blood cells per unit volume would compete for the limited axial equilibrium positions in the focusing line, that is, a part of blood cells first occupy the equilibrium positions under the coupling of inertial lift force and Dean drag force while other non-equilibrated cells tend to migrate towards the equilibrium positions and unavoidably collide with the equilibrated cells or other cells on the way. The bounced cells would repeat this process until an equilibrium status is reached. Therefore, the cell band is actually the maximum possible distribution region of the bounced cells. In addition, we found that at the lowest dilution ratio of 10-fold, two small cell-free regions were generated near both inner and outer channel walls. Different from those observed in the above investigation on flow-rate effects (see Fig. 2), the cell-free regions are generated in this experiment due to the dominate repulsion effects in near-wall regions. With the increase of dilution ratio, these two cell-free regions (especially the outer cell-free region) get widened gradually, resulting in the narrowing of the cell band. Meanwhile, the equilibrium positions of blood cells would not shift inward or outward as discussed in the experiment on flow rate effects.

Fig. 3
figure 3

a Composite images illustrating cell distributions across the channel width at different dilution ratios or hematocrits. b Widths (W) of the focused-cell-band as a function of dilution ratio

Full size image

To quantitatively probe into the concentration effect on blood cell migration, we measured the width (W) of the focused-cell band under different dilution ratios using calibrated IMAGEJ software, and plotted the measured data as a function of dilution ratio in Fig. 3(b). From this figure, it is found that the relationship between W and dilution ratio exhibits a non-linear variation, that is, W sharply decreases with the increase of dilution ratio under low dilution ratios (10-fold~100-fold) but gradually decreases under high dilution ratios (100-fold~1000-fold). This finding indicates that the collision probability or the steric crowding effect is nonlinearly dependent on the blood cell concentration. After the blood cell concentration is diluted to below a certain threshold, the cell-interactions can be ignored owing to the adequate spacing between two adjacent cells in the main flow streams, and the W would keep constant regardless of the flow rate. Therefore, the use of high ratio dilution can often be found in studies on particle or cell migration mechanisms (Martel and Toner 2012; Xiang et al. 2013) but may be problematic for real-world applications.

4.4 High-throughput isolation of plasma from human whole blood

Given the above understandings on the flow-rate and concentration effects on blood cell migration, efficient isolation of blood plasma from human whole blood was performed using our spiral inertial microfluidic devices under the optimal flow rate and dilution ratio conditions. It is commonly known that plasma, a water-based (up to 91 %) biological buffer, constitutes ~55 % of the total human blood volume and contains a rich mine of biomarkers (e.g., proteins, metabolites, circulating nucleic acids and so on) whose contents or physiological states are especially valuable for clinical diagnosis (Kersaudy-Kerhoas and Sollier 2013). However, blood plasma isolation is always performed manually and off-chip with the assistance of high-speed centrifugation or membrane filtration which seem impossible to be integrated into the next generation on-chip analyses. Therefore, microfluidic devices or schemes to efficiently realize microscale plasma separation are highly demanded for building low-cost diagnostic instruments. To efficiently design or evaluate a microscale plasma separator, four critical parameters need to be taken into consideration: (1) operating throughput (or sample processing time); (2) dilution ratio; (3) purity of the isolated plasma, calculated as 1-c p/c b (where c p is the number of blood cells in the isolated plasma and c b is the number of blood cells in the initial blood sample); (4) yield of plasma, calculated as the percentage of the isolated plasma volume over the total volume of the injected blood sample. The tradeoff between these four technical parameters needs to be carefully balanced in order to meet the demand for efficient blood plasma isolation in future commercial applications.

To embody the throughput advantage of our spiral inertial microfluidic device, the maximum safe flow rate of 700 μl/min was selected to perform the blood plasma isolation. A classical and universal Y-shaped outlet system with two equal parts was adopted to export the isolated plasma, which can further be expanded for other cell/particle processing applications. Although using this type of outlet system would make the isolation yields of plasma lower, it can significantly reduce the required dilution ratio of the initial blood sample, thus improving the processing efficiency of our device. On the basis of the cell band widths measured under different dilution ratios (see Fig. 3(b)), the optimal dilution ratio for our device was determined to be 20-fold under which the cell band width was measured to be ~170 μm (slightly larger than the half channel width (150 μm) of the Y-shaped outlet). The 20-fold diluted blood sample was then pumped through the spiral channel at the flow rate of 700 μl/min. As prospected in the working principle for isolating blood plasma, blood cells randomly distributed near the inlet migrate towards the inner channel region, form a cell band of nearly half the channel width, and are finally exported through the inner outlet while cell-free blood plasma is collected through the outer outlet. The blood cell distributions across the channel width near the inlet and outlets are illustrated in the stacked composite images of Fig. 4(a), which well validates the above demonstrated plasma isolation principle. Almost the entire cell band of ~170 μm wide before branches is exported out through the inner outlet due to the splitting effect. Pictures of the samples collected from inner and outer outlets are shown in Fig. 4(b). It is found that the sample collected from the inner outlet (Tube one) appears red while the sample collected from the outer outlet (Tube two) appears pale-yellow which is the natural color of blood plasma. To ascertain their compositions, the collected liquids were sampled several times (N = 12) using a microscope (80i, Nikon) under dark-field illumination (one group of the captured images is listed in Fig. 4(b)). The blood cells (mainly RBCs) were found to be highly enriched in the sample collected from the inner outlet (Tube one) while there are few blood cells observed in the sample from the other outlet (Tube two). The cell numbers per unit volume of the samples were counted several times using a hemocytometer. Under the optimal conditions, the purity of the blood plasma isolated using our spiral inertial microfluidic device was measured to be almost 100 %, and the yield of plasma was calculated to be ~38.5 %.

Fig. 4
figure 4

a Schematic diagram of the working principle for isolating blood plasma from diluted human whole blood using our spiral inertial microfluidic device with a universal Y-shaped outlet system. The blood cells randomly distributed near the inlet migrate towards the inner channel region, and are finally exported through the inner outlet while the cell-free blood plasma is collected through the outer outlet. The insets show the composite images illustrating cell distributions near the inlet and outlets at the flow rate of 700 μl/min under a 20-fold dilution. b Pictures of samples collected from the inner outlet (Tube one) and the outer outlet (Tube two), respectively. The dark-field microscope images were captured to show the detailed compositions of the collected samples

Full size image

4.5 Discussions on the isolation performances of our device

The technical performances of our spiral inertial microfluidic device were then compared with those of previously-reported blood plasma separators. Specifically, the most popular active external-field (e.g., acoustic, electrical and so on) based devices (Lenshof et al. 2009; Hai et al. 2011; Mohammadi et al. 2015) can offer a very high isolation accuracy because the active forces acting on blood cells can be finely tuned by adjusting the parameters of the field-generators. As a result, the purity of the plasma collected using these devices can be quite close to 100 %. However, due to the lagged response of cells in external fields, the throughput of active devices is usually very low. Our spiral inertial microfluidic device can also offer a near 100 % isolation purity as well as a much higher throughput (~ml/min) as compared with the high-resolution active devices. Meanwhile, the expensive and cumbersome field generators can be avoided for the only power required in our device is the power to drive the sample at specific flow rates, and it can be provided by the popular low-cost integrated micropumps or by manually pushing the syringes.

Another widely-investigated category is the devices developed from the macroscale “gold standard” plasma separation methods (high-speed centrifugation or membrane filtration). Although these centrifugation or membrane based devices (Haeberle et al. 2006; Tripathi et al. 2015) can directly process blood samples without pre-dilution, they inherit or even magnify the shortcomings of macroscale methods, such as labour-intensive and easily-clogged. In comparison to these devices, our device requires the sample to be diluted 20-fold, that is, for a volume of 7.5 ml sampled from the vein (e.g., for the CellSearch system to enumerate enough rare CTCs), a final volume of 150 ml needs to be processed timely. Owing to the high-throughput feature of our device (700 μl/min), this huge volume sample (150 ml) can be handled within 215 min. Although this performance is remarkably superior to those of other techniques, it is still unbearable that the processing of one sample costs more than 3 h. However, the primary advantage of microfluidics is to deal with samples of extremely low volume (even in nl or μl scales), which indicates that our device may be a good tool for processing small-volume blood sampled from fingersticks. In addition, multi-devices can be staked in the vertical direction to increase overall throughputs of our device.

Following the great success in paper-based pregnancy testing and glucose measurement, the newly-born paper microfluidic devices (Nilghaz and Shen 2015; Chen et al. 2015) propose a new concept of zero-cost diagnostics and have also been developed for isolating blood plasma. However, their applicability in multi-functional instruments is limited since it is difficult to recover the isolated plasma for downstream analyses. Instead, our spiral inertial microfluidic device can be either used as a single unit for realizing relatively simple functions or integrated with other Lab-on-a-chip (LOC) components to achieve multi-parameter detections.

In addition, our spiral inertial microfluidic device was compared with other similar passive hydrodynamic devices. It is found that the throughput of our device is still several times higher than those of inertial microfluidics with straight channels (Mach and Di Carlo 2010) or channels with microcavities (Lee et al. 2014). Recently, multi-stage devices which integrate two or more methods (including both active methods, e.g., dielectrophoresis and magnetophoresis, and passive methods, e.g., inertial focusing, deterministic lateral displacement and hydrophoresis) have been applied for plasma or blood cell separation (Yan et al. 2014; Karabacak et al. 2014).These devices inherit both the high resolution feature of active methods and the high throughout feature of passive methods, and thus have untapped potentials to be unearthed. In all, the spiral inertial microfluidic device developed in our work provides a balanced overall performance, and has overriding advantages in processing throughput and operating efficiency.

5 Conclusion

In this work, the focusing dynamics of blood cells in a spiral inertial microfluidic device were systematically investigated. First, the flow-rate effects on cell migration dynamics were characterized to obtain an improved understanding of the force competition acting on flowing blood cells. The blood cells were found to get focused and form a relatively narrow cell band instead of a perfect single-particle array observed in the experiments with rigid polymer microspheres due to the irregular shape and polydisperse size of blood cells. It was firstly discovered in our experiments that the focusing status of discoid red blood cells (RBCs) is dependent on the channel height. The focusing status of the aligned RBCs was found to change from the vertical attitude near the bottom channel wall to the flat attitude near the top channel wall. Then, the concentration effects on cell migration dynamics were explored to understand the interactions and steric crowding effects of flowing blood cells. It was found that the width of the cell bands sharply decreases with the increase of the dilution ratio. These experimental results would provide valuable insights for guiding the operation of spiral inertial microfluidic devices. Finally, the isolation of plasma from human whole blood with a dilution ratio of 20-fold was successfully performed at a throughput up to 700 μl/min. The purity of the isolated blood plasma was close to 100 %, and the yield was calculated to be ~38.5 %. The isolation performances of our device were then compared with those of the reported devices. It is concluded that our spiral inertial microfluidic device provides a balanced overall performance and has overriding advantages in terms of processing throughput and operating efficiency.