A Consecutive Separation Strategy Using a 3D‐Printed Microfluidic Chip to Achieve High‐Purity White Blood Cells From Blood

ABSTRACT

Label‐free and direct separation of white blood cells (WBCs) remains one of the major challenges in the field of efficient leukocyte separation. To address this challenge, we propose a WBC sorting strategy based on inertial microfluidics and 3D‐printing technology, which employs a single inertial microfluidic chip to perform four consecutive high‐throughput separation steps between red blood cells (RBCs) and WBCs. Specifically, the chip features a trapezoidal cross‐section spiral channel to achieve size‐dependent, stable spatial separation of cells along the inner and outer channels. After the conceptual design, the UV cured 3D printing technology is employed to fabricate the inertial chip and its fixture. Subsequently, mixed polystyrene microparticles are used as test objects to verify the sorting performance of the inertial chip, and the results indicate that the optimal separation flow rate is ∼1700 $\mu \mathrm{L}/\min$. This corresponds to a maximum WBC throughput of ∼1.12 × 10⁷ cells/h during a single‐pass separation under the current 1:100 dilution condition, assuming a whole‐blood WBC concentration of 1.1 × 10⁷ cells/mL. Ultimately, the diluted whole‐blood samples are used to examine the four‐cycle consecutive sorting strategy, and the results demonstrated that this strategy, proposed in this work, could directly obtain WBCs from blood samples with a recovery efficiency of ∼94.9% and a purity of ∼85.5%. Related experiments indicate that this strategy which is centered on the 3D‐printed chip offers high reliability and sensitivity for label‐free WBC sorting from blood cells. With the aid of state‐of‐art additive manufacturing technology, it is promising that our strategy could be widely applied on fields including cell analysis and disease diagnosis.

A four‐cycle consecutive separation strategy using a 3D‐printed inertial microfluidic chip for high‐purity WBC isolation from blood.

1. Introduction

Although white blood cells (WBCs) represent merely 0.1% of the total blood cell population, they are highly informative for blood‐based diagnostics [1, 2]. Their quantity is recognized as a standard reference in clinical diagnosis of inflammation and infections. As an indicator reflecting the balance of peripheral leukocyte subpopulations, the neutrophil‐to‐lymphocyte ratio (NLR) is significantly associated with the clinical prognosis of spontaneous intracerebral hemorrhage (ICH) [3, 4]. In addition, tumor immunology studies have found that highly diverse populations of tumor‐infiltrating WBCs can exert strikingly antagonistic effects against different types of tumors [5, 6]. In the field of biomedicine, WBCs play an indispensable role as primary immune cells [7]. However, at present, due to the high proportion of red blood cells (RBCs), achieving high‐purity separation of WBCs and RBCs from whole blood is difficult, with numerous challenges including low separation efficiency, low WBC purity, and high cost [8, 9].

The traditional method for isolating WBCs is centrifugation. Its principle is to exploit differences in the density, size, and morphology of blood components so that the sample stratifies under a defined centrifugal force and duration, allowing the desired fraction to be selectively collected [10, 11, 12]. This approach enables WBC separation at relatively low cost and is among the most widely used methods at present [13]. However, existing centrifugation methods can damage cells; under external centrifugal forces, the integrity of the cell membrane is readily compromised [14]. In addition, the lysis buffer used during centrifugation not only slightly reduces cell viability but also introduces exogenous contaminants, thereby compromising subsequent assays and experiments [15, 16].

Beyond traditional centrifugation, label‐based approaches are now also widely used to separate RBCs and WBCs [17, 18]. For example, one of the most extensively employed method is flow cytometry, which uses specific fluorescently labeled antibodies to bind to biomarker molecules on target cell surfaces and achieves WBC separation through laser interrogation and electrostatic deflection [19, 20, 21]. In addition, in 2016, Sajay et al. developed a cell marker‐based immunomagnetic bead separation method. By designing CD45‐conjugated magnetic particles, they bound specifically to WBCs, and efficient WBC separation was achieved via the combined effects of external magnetic deflection and fluid dynamics [22]. However, these cell marker‐based separation methods still have limitations. First, the labeling process requires the selection of specific antibodies for target cells, but there are batch‐to‐batch variations in antibody selectivity and specificity, which can lead to variations in experimental results [23, 24]. In addition, false identification may arise when some leukocytes upregulate specific markers upon activation [25]. More importantly, some studies have shown that targeting specific markers on cells may interfere with the normal function of the cell membrane and may introduce impurities into subsequent assays, compromising the stability of the measured results [26, 27]. More broadly, recent reviews have highlighted that microfluidic cell sorting is increasingly developing toward clinically relevant workflows that reduce sample preparation, avoid unnecessary labeling, and facilitate downstream analysis of heterogeneous blood samples [28, 29].

In recent years, with advances in technology, WBC separation methods have increasingly been classified, according to their working principles, into active and passive technologies [30, 31]. Among them, active technologies rely on externally applied physical fields to achieve separation [32]. For example, Amir Shamloo et al. invented a method for separating target cells using an acoustic device in 2017, achieving separation of different types of cells by altering the amplitude of the acoustic field [33]. In addition, methods for separating target cells by altering external field conditions such as electrical, magnetic, and optical fields have also been progressively developed [34, 35, 36]. However, active separation strategies still have notable limitations. These methods require the application of external physical fields and typically rely on cell‐by‐cell coupling–response mechanisms, which constrain the number of cells that can be processed per unit time and result in low overall throughput [37, 38]. For the vast population of blood cells, such active approaches yield inefficient separation and cannot meet the current, rapidly growing demand for large quantities of WBCs [39, 40].

Compared to active technologies, passive technologies rely on the motion characteristics of particles in fluids, achieving efficient sorting by utilizing the physical properties of cells, such as size and density [41, 42, 43]. Inertial microfluidics offers several advantages over other passive sorting technologies, including low cost, high throughput, simple channel design, and label‐free operation [44]. It can adjust the design of the internal channels to utilize the inertial lift and Dean vortex forces generated during fluid flow, gathering particles of different sizes to different equilibrium positions [45, 46]. This allows for cell separation while maximizing the preservation of cell viability. Therefore, the geometry of the internal channels in inertial microfluidics has become one of the key areas of research in the field of microfluidics [47]. For example, Bhagat et al. and Kuntaegowdanahalli et al. proposed a sheath‐less rectangular spiral channel to separate microparticles of different diameters [48]. However, with the advancement of technology, Wu et al. designed a trapezoidal spiral inertial microfluidic sorter and proved that the trapezoidal spiral channel offers better separation resolution compared to traditional rectangular spiral channels, especially for biological particles of similar sizes [49]. As a result, due to the advantages of the trapezoidal spiral channel in separation performance, it has been widely applied to the separation of other cells or micro‐swimmers for various biomedical applications [50, 51, 52, 53, 54]. Despite these advances, direct WBC enrichment from whole blood still faces a practical trade‐off among purity, recovery, throughput, and device manufacturability. Recent studies based on whole‐blood acoustic sorting, modified spiral geometries, and cascaded inertial layouts further suggest that improving purity while preserving a simple and scalable device architecture remains an open challenge [55, 56, 57].

To address the above problems, this article presents a label‐free WBC sorting process based on a 3D‐printed chip. Specifically, precise WBC separation is achieved by designing a trapezoidal spiral channel in the chip and exploiting inertial microfluidic effects. In addition, the sorting accuracy and cell purity are further improved through a strategy of repeated sorting. However, the rationale of this strategy is not simply to repeat an existing channel design. Because RBCs overwhelmingly outnumber WBCs in whole blood, a single‐pass separation, although effective for initial spatial segregation, still tends to leave residual RBC contamination at the target outlet, thereby limiting the attainable WBC purity. Therefore, we introduced a consecutive multi‐round sorting workflow, in which the inner–outlet fraction is repeatedly reprocessed to progressively reduce residual RBC carryover while maintaining a high WBC recovery. After the process design was completed, microparticles and cell samples were used to validate its performance. Experimental validation shows that the process achieves effective WBC enrichment in terms of recovery and purity. Under the current operating conditions (1700 $\mu \mathrm{L}/\min$) and four consecutive sorting cycles, it was found that the vast majority of cells were not stained after sorting by trypan blue staining, indicating that the vast majority of WBCs after sorting still remained active (seen in Figure S2). The core of our process is an advanced 3D‐printing technology, which enables integrated, rapid fabrication. It can complete the printing of complex structures in a shorter time, providing significant cost‐effectiveness and manufacturing flexibility. Therefore, this process has the potential to serve as an automated and rapid method for extracting WBCs from whole blood samples in future biomedical applications.

2. Methods and Materials

2.1. Workflow Design for WBC Separation

To realize label‐free separation of WBCs from whole blood with high throughput and high purity, a special designed workflow with four main steps was employed. As illustrated in Figure 1a, we first collected whole‐blood samples from healthy donors, defined as anticoagulated blood with no prior separation (whole blood). After a series of preprocessing steps, the samples were infused into the device we designed at a specific flow rate to differentially separate WBCs from RBCs. Due to the extreme imbalance in the ratio of RBCs to WBCs, the boundary between RBCs and WBCs at the bifurcation was not sufficiently sharp in the first separation, a small number of RBCs may enter the inner outlet. Therefore, the inner outlet fraction was reprocessed in subsequent cycles to progressively improve purity. Following four sorting cycles, the vast majority of WBCs were recovered and their purity increased to a high level. Our workflow's core device consists of two components: a chip fixture and a 3D‐printed microfluidic sorting chip. The fixture is used to ensure quick connection between the external liquid pipeline and the internal flow channel of the chip. The chip features one inlet (1) and two outlets (2 and 3), with a spiral channel of trapezoidal cross‐section as its principal internal structure. The total length of the microchannel was approximately 285 mm. The main geometrical parameters of the trapezoidal spiral channel, including the channel width, inner‐wall height, outer‐wall height, and number of spiral turns, are summarized here, whereas the detailed dimensional schematic is provided in Figure S3. Specifically, the channel width was 500 $\mu \mathrm{m}$, the inner‐wall height was 50 $\mu \mathrm{m}$, the outer‐wall height was 150 $\mu \mathrm{m}$, and the number of spiral turns was 5. The separation principle is illustrated in Figure 1b,c. The sample inlet (1) of the cube device (containing the 3D‐printed microfluidic chip) was located in the center of the cube device, and the two outlets (2 and 3) coincided with the outlet of the trapezoidal spiral channel. In our process for WBC extraction, whole blood samples were first collected from human veins, followed by the addition of phosphate buffered saline (PBS) and dilution at a ratio of 1:100 to form a cell suspension. As a result, high‐purity WBC samples were collected from the inner outlet 2. The bifurcation ratio at the outlet was inner outlet:outer outlet = 1:2. In practice, the collected outlet volumes may slightly deviate from the ideal 1:2 split ratio due to unavoidable dead volume and sample retention in the microchannel (seen in Figure S4). We adopted four sequential sorting cycles to optimize the trade‐off between RBC depletion and WBC retention. In preliminary experiments, four cycles yielded higher WBC purity than three, while achieving higher WBC recovery with reduced WBC loss compared with five. Accordingly, the four‐cycle method was confirmed as the standard sorting protocol for the formal experiments owing to its superior overall separation performance.

FIGURE 1.

(a) Workflow of the designed process for isolating WBCs from whole blood. (b) Schematic diagram illustrating the conceptual working principle of the microfluidic chip. The whole blood was pumped in via inlet 1. The WBCs and RBCs samples are collected from the two outlets 2 and 3, respectively. (c) Principle of using the inertial microfluidic separation for WBCs extraction within the chip. RBCs, red blood cells; WBCs, white blood cells.

The main structure of the 3D printed chip was a spiral channel featured with trapezoid cross‐section, which is different from traditional spiral channel with rectangular cross‐section. Specifically, the channel was designed with a smaller height for the inner wall and a greater height for the outer wall. When the fluid flows through the trapezoidal spiral channel at a specific velocity, the cells will both experience the inertial lift force ($F_L$) caused by wall effects, and the Dean drag force ($F_D$) caused by Dean vortex. With the coupling of these two forces, smaller particles (e.g., RBCs) will migrate outward and be captured at the cores of Dean vortex near the outer wall. However, larger particles (e.g., WBCs) will be focused into a narrow band near the inner wall of the channel. Moreover, unlike the traditional spiral channel (rectangular cross‐section) where the cores of Dean vortex were centered at the channel cross‐section, in our spiral channel, the cores of Dean vortex shift toward the outer wall. As a result, the migration distance between smaller and larger particles will be enlarged, yielding a better separation efficiency than the rectangular spiral channel.

2.2. Fabrication Method

In the chip fabrication process, a novel 3D printing technique, namely, projection micro‐stereolithography ($\mathrm{P}\mu \mathrm{SL}$), was innovatively applied. The $\mathrm{P}\mu \mathrm{SL}$ 3D printer consists of three major subsystems, the top photolithography system, the middle material‐forming system, and the bottom motion‐control system, as illustrated in Figure 2a. Briefly, the top photolithography system consists of five components: a 405 nm ultraviolet light source, a digital micromirror device (DMD), a beam splitter, a projection objective lens, and a real‐time monitoring camera (CCD); these components work together to achieve high‐precision full‐area image exposure. The middle material forming system consists of four components: an optical release film, a resin tank, a precision scraper, and the Z1 axis of the printing platform. The main function of the middle material forming system is heating, spreading, and layer‐by‐layer curing of the liquid resin. The bottom motion control system consists of the printing platform, which drives the platform to move precisely along the three axes, enabling layer‐by‐layer stacking and construction of the structure. To enable large‐scale chip fabrication, a commercial projection micro‐stereolithography platform (microArch S240) was employed in this study in combination with our optimized printing process. What's more, in this work, the microfluidic chip was fabricated as an integrated monolithic structure by 3D printing, rather than being assembled from laminated layers.

FIGURE 2.

(a) Schematic structure of $\mathrm{P}\mu \mathrm{SL}$ 3D‐printer. Arranged from top to bottom are three sections. (b) Workflow for the printing process of our chip. (c) Detailed schematic of the photolithography system in section A of the (a). (d) Detailed schematic of step 3 (printing) in the 3D‐printing process shown in the (b). DMD, digital micromirror device.

The detailed manufacturing process of our chip can be summarized in four main steps: modeling, model slicing, printing, and post‐processing, as shown in Figure 2b. First, a digital model of the chip is constructed using the 3D modeling software Solidworks. Second, the chip model is sliced along the Z‐axis to generate a series of continuous 2D images using slicing software Cura. Subsequently, the printing of the chip is then initiated. The resin was selected from commercially available photosensitive resins, and the printing parameters were optimized in‐house. To improve the flowability and spreading performance of the resin, the pre‐treated photosensitive resin was first heated to approximately 50, thereby reducing its viscosity from 450 to 90 cPs. Next, a precision scraper was used to evenly spread the resin over the optical release film surface, forming a consistent layer of liquid resin with uniform thickness. Then, a 405 nm ultraviolet light source in the lithography system was modulated by DMD, focused by an objective lens, and projected onto the resin surface successively, achieving high‐resolution full layer image exposure and curing. The above process was then repeated until the 3D structure of the chip was built layer by layer (seen in Figure 2d). Finally, the printed chip was subjected to post‐processing, during which it was removed from the platform and ultrasonically cleaned in alcohol. After cleaning and drying, the chip was placed in a UV curing box for secondary curing. At last, the chip dimensions were measured under a microscope to determine the processing quality of the chip. Some of the preliminary experiments in this study were conducted using our previously fabricated polymer‐film microfluidic chip. In addition, the detailed fabrication process and performances are described in our earlier publications [58, 59, 60].

2.3. Sample Preparation

To evaluate the inertial aggregation in the spiral channels, two types of polystyrene microparticles were respectively used to simulate WBCs (10 $\mu \mathrm{m}$) and RBCs (4 $\mu \mathrm{m}$). To prepare the microparticle suspensions, all the polystyrene microspheres were diluted in PBS containing 1% Pluronic F127 (Sigma‐Aldrich).

Whole blood samples were collected from healthy volunteers using vacutainer collection tubes containing ethylenediaminetetraacetic acid (EDTA) as an anticoagulant. After collection, the whole blood samples were immediately stored at room temperature (20–25°C) and processed within 4 h to prevent changes in cell morphology and loss of cell viability. In this work, before using, the acquired whole blood was diluted at a ratio of 1:100 in PBS prior to the experiment. To ensure uniform cell distribution in the diluted whole blood samples, the samples were diluted in low‐adhesion centrifuge tubes, and the mixture was gently inverted to achieve thorough mixing.

2.4. Experimental Setup and Data Analysis

To ensure stable and precise infusion of the sample, the prepared particle suspension and cell samples were injected into the cube device at specific flow rates using a syringe pump (Legato 100, KD Scientific). Following a short period of operation, the near‐outlet dynamics of microparticles were continuously imaged with an inverted fluorescence microscope (Olympus IX71) equipped with a Phantom V611 high‐speed camera (Vision Research). Related image frames were taken for later analysis. ImageJ software (NIH) was used to create the composite images illustrating the distribution of particles/cells at specific flow rates. For clarity, the throughput in this work is reported both as the processed suspension flow rate and as the equivalent whole‐blood throughput. Under the optimal operating condition of 1700 $\mu \mathrm{L}/\min$, the maximum WBC throughput was ∼1.12 × 10⁷ cells/h during a single‐pass separation, corresponding to an equivalent whole‐blood WBC throughput of 1.1 × 10⁷ cells/mL under the present 1:100 dilution protocol. To depict particle/cell migration over time, we generated focusing maps by stacking composite images acquired at different flow rates.

In the cell separation experiment, an automatic cell counter (Countess II FL, Thermo Fisher Scientific) was used to distinguish WBCs from RBCs and quantify the concentration of cells in the sample.

The recovery ratio was defined as the number of target cells collected at the target outlet divided by the number of target cells introduced at the inlet. Purity was quantified as the fraction of target cells among all cells present at the target outlet.

3. Results and Discussion

3.1. Performance Characterization of the Chip Using Polystyrene Particles

To further explore the performance of the spiral channel for particle separation, a mixture of polystyrene microspheres with diameters of 4 and 10 $\mu \mathrm{m}$ was selected to simulate RBCs and WBCs in the diluted blood samples, respectively. Figure 3a,b illustrates the distributions of these two particle types at the outlet of the spiral channel. First, the particle suspension was injected into the chip at a flow rate of 800 $\mu \mathrm{L}/\min$ using the syringe pump, and composite trajectory images at the outlet region were observed. At this flow rate, the small particles (4 $\mu \mathrm{m}$) and large particles (10 $\mu \mathrm{m}$) were basically separated, and the clear boundaries could be observed between the two types of particles. Then, the flow rate was gradually increased. According to earlier publications, it can be inferred that the small particles were trapped in the cores of Dean vortices near the outer wall and remain virtually unchanged in the outer‐wall region across all tested flow rates under the combined influence of inertial lift and Dean drag forces, as illustrated in Figure 3a. (Full sets of outlet distribution images at all tested flow rates are provided in Figure S5) Due to the weak inertial effects of the small particles, the Dean drag force was much greater than the inertial lift. As a result, the small particles exhibited a uniform, dispersed distribution near the outer wall. In contrast, relatively stable focusing of the larger microparticles (10 $\mu \mathrm{m}$) was observed near the inner wall at low flow rates due to the dominant inertial lift force. As the flow rate increased, the focusing of larger particles was progressively improved, with its focusing positions being shifted toward the inner wall, and a distinct particle string was formed gradually. Notably, for the suspension containing a mixture of 4 and 10 $\mu \mathrm{m}$ particles, the best particle separation performance was achieved at a flow rate around 1700 $\mu \mathrm{L}/\min$ according to the experimental results.

FIGURE 3.

(a) Images illustrating the distributions of 10 and 4 $\mu \mathrm{m}$ particles at selected flow rates. (b) Image illustrating the best distribution of particles at 1700 $\mu \mathrm{L}/\min$. (c) Quantify of the particle distributions at different flow rates.

To better evaluate the separation performance of particles, the focusing position and distribution of the 10 and 4 $\mu \mathrm{m}$ particles at various flow rates were measured, as illustrated in Figure 4b. Within the flow rate range of 800$-$2000 $\mu \mathrm{L}/\min$, it was found that the focusing behaviors of the two types of particles exhibited distinct differences. Notably, when the flow rate ranged from 1700 to 1900 $\mu \mathrm{L}/\min$, a relatively efficient extraction of the 10 $\mu \mathrm{m}$ particles could be achieved. Besides, although in this flow rate range, the position of the 4 $\mu \mathrm{m}$ particle band slightly extended to the bifurcation point of the two outlets, there is no adverse impact on the subsequent multiple cascading sorting. Moreover, in this flow rate range, the position of the 10 $\mu \mathrm{m}$ particle string significantly tended to aggregate toward the inner wall of the channel, forming a stable spatial separation between the two particle strings. After repeated experimental verification, it was found that the distance between the two particles reached its maximum when the flow rate was 1700 $\mu \mathrm{L}/\min$. The position of 4 $\mu \mathrm{m}$ particle band exhibited minimal relative displacement, whereas the focusing pattern of the 10 $\mu \mathrm{m}$ particle string was clearly defined. Further increasing the flow rate, the distance between these two types of particles gradually decreased. Besides, the focusing position of the 4 $\mu \mathrm{m}$ particle band fluctuated, with a decrease in focusing stability and a deterioration in sorting performance. Therefore, the optimal flow rate for realizing the best sorting performance was determined to be 1700 $\mu \mathrm{L}/\min$.

FIGURE 4.

(a) Photographs of the samples collected from the inner outlet and outer outlet at the specific flow rate. (b) Representative images illustrating the distributions of RBCs and WBCs at the specific flow rate. (c) Representative aliquots of samples collected after different sorting cycles for visual color comparison (d) Representative cell distribution images of samples under a microscope at different sorting cycles. WBCs, white blood cells.

3.2. Verification of the Multiple and Continuous Separation Strategy

After a series of particle experiments, the cube device was applied to multiple cascade cell sorting. To better showcase the utility of the chip for the precise cell separation, the diluted whole blood sample was infused into the cube device for sorting at a flow rate of 1700 $\mu \mathrm{L}/\min$. The first separation effect of our chip is shown in Figure 4a. From the appearance, it is obvious that the sample color has become lighter. The next separation sample is collected through the inner outlet, and the sample waste liquid is collected through the outer outlet. According to Figure 4b, in the first separation, due to the large number of RBCs themselves, the boundary between RBC and WBC is not obvious. But according to subsequent measurements of WBC recovery rate, it was found that most RBCs were removed in the first separation. In the second separation, the boundary between RBCs and WBCs was clearly formed, and the vast majority of WBCs were recovered through the inner channel, whereas RBCs were removed through the outer channel. However, the boundary line still moved up and crossed the bifurcation point of the flow channel, causing a small amount of RBCs to be carried out and affecting the purity of WBCs. It was speculated that this was due to the collision induced lateral diffusion of high concentration RBCs during the flow process, which caused the boundary line to move up and affected the separation of WBCs. In the third separation, the boundary between RBCs and WBCs was further sharpened and moved down to a position close to the bifurcation point of the flow channel. By observing the trajectory diagram, it could be observed that almost all WBCs flow out through the inner channel, whereas RBCs were removed through the outer channel, further improving the separation efficiency. The boundary of the fourth separation coincided with the bifurcation point of the flow channel, and there might still be a very small number of RBCs entering the inner channel, but it has almost no effect on the purity of WBCs. In the fifth separation, it was observed that the boundary line moved down below the bifurcation point. It was speculated that this was due to the high concentration of WBCs in the collected samples after multiple sorting, which caused some WBC loss due to collision and diffusion during the flow process. Therefore, we chose to perform four cycles of separation during the experiment to achieve maximum WBC separation.

As shown in the subgraphs of Figure 4c,d, they correspond one‐to‐one with the results of Figure 4b. (The detailed images of each cycle at the specific flow rate are shown in Figure S8). From Figure 4c, it can be seen that in the first separation, the whole blood sample appears red and the liquid volume is relatively large, with a lower WBC content. After the second separation, the overall color of the sample became lighter, appearing light red, and the volume decreased. As the number of separation times increases, the color of the sample further faded until it became completely transparent, and the liquid volume continued to decrease while the WBC concentration continued to increase. And Figure 4d verified the above results. Through microscopic observation of cells, it was found that in the first separation, the cell density was high and the overall color was red, with a high concentration of RBC. What's more, in the second separation, the number of cells decreased significantly, the concentration of RBCs decreased significantly, and the concentration of WBCs increased. In further separations, the concentration of RBCs continued to decrease while the concentration of WBCs continued to increase, which fully met our experimental expectations.

3.3. Application of the 3D‐Printed Chip for WBC Separation

To further evaluate the cell component changes in samples after processing using the chip, we conducted a cell staining experiment, as illustrated in Figure 5a. First, 100 $\mu \mathrm{L}$ of samples from 1 to 4 steps separation processes were pipetted by a micropipette and then added to the center of a clean glass slide. To promote cell settling at the bottom of the slide, the glass slide was left undisturbed at room temperature for 30–60 min. Subsequently, the cleaning process was conducted. Spilled liquid was gently absorbed using lint‐free wipes, followed by the slow addition of 100 $\mu \mathrm{L}$ of PBS. The slide was then left undisturbed at room temperature for 30 to 60 min to ensure that the cell sedimentation position remained undisturbed. The cleaning process was completed by repeating the above steps three times. Next, the fixation phase was conducted. Methanol that was precooled to −20°C was added dropwise to the glass slide for cell fixation, and the fixation was kept for 10 min. After fixation, the above cleaning steps were repeated to remove any residual fixative solution. Finally, DAPI stain was added for nuclear staining, and a coverslip was placed to complete the cell staining. Figure 5b shows the fluorescence image of the stained cells at 50× magnification. Through fluorescence treatment, WBCs appear bright in the images, distinguishing them from RBCs, which are consistent with the characterization results shown in Figure 4d. In the first separation, there were more RBCs and fewer white bright spots. In the second separation, the overall number of cells decreased, and the number of bright spots in the images increased significantly, but their proportion was still relatively small. In the third separation, the number of cells sharply decreased, the bright spots in the figure were concentrated, and the concentration significantly increased. As the number of sorting cycles increases, the total number of cells gradually decreases and the concentration of bright spots gradually increases.

FIGURE 5.

(a) Schematic flowchart of the DAPI staining process. (b) Fluorescence images of samples at different sorting cycles. (c) Purity comparison between polymer‐film chip and 3D‐printed chip. (d) Recovery rate comparison between the polymer‐film chip and the 3D‐printed chip. PBS, phosphate buffered saline; WBCs, white blood cells.

According to our previous studies and experimental results [57, 58, 59], both the polymer‐film chip and the 3D‐printed chip are capable of achieving precise separation of WBCs. To compare their actual separation performance, comparative experiments were conducted between the two chips. As shown in Figure 5c,d, the leukocyte purity of both types of chips exhibited an increasing trend with the number of sorting cycles and reached its peak after the fourth separation. However, as shown in Figure 5c, the leukocyte purity of the polymer‐film chip was higher than that of the 3D‐printed chip in each sorting cycle, with the difference being particularly pronounced during the first three rounds. It is speculated that this difference in purity primarily arises from the layer resolution and surface quality limitations inherent to the 3D printing process. Layer‐by‐layer printing might create tiny steps or surface irregularities within the channels, which perturb the local flow field and in turn compromise cell focusing and separation performance. Meanwhile, as shown in Figure 5d, the WBC recovery of both chips remains above 90%, but the polymer‐film chip exhibits consistently higher recovery than the 3D‐printed chip, indicating that the polymer‐film chip retains an advantage in balancing purity and recovery. Although the current separation performance of the 3D‐printed chip is slightly inferior to that of the polymer‐film chip, its advantages in ease of fabrication, low cost, and scalability endow it with greater application potential in the future as 3D printing technology continues to improve in precision.

4. Conclusion

In summary, we have developed a consecutive strategy for high‐purity, label‐free sorting of WBC from blood cells, with a 3D‐printed microfluidic chip at its core. After the conceptual design, we examined how channel geometry and flow rate influence particle/cell focusing and stream bifurcation and determined the optimal rounds of sorting cycles through comparative multi‐cycle experiments. The results show that at a flow rate of 1700 $\mu \mathrm{L}/\min$, 10 and 4 $\mu \mathrm{m}$ polystyrene particles can be effectively separated, whereas the optimal number of rounds for sorting high‐purity WBCs from blood cells was four. Finally, we applied the strategy to blood samples and achieved efficient WBC isolation, with recovery and purity of approximately 94.9% and 85.5%, respectively. Clinical evaluations demonstrate that the proposed sorting strategy effectively depletes RBCs and isolates WBCs from whole blood, thereby streamlining the WBC separation process. It provides a low‐cost, high‐throughput, and efficient separation method, and the device offers additional advantages of ease of fabrication and broad accessibility. With continued advances in printing resolution and wider adoption of 3D printing, this strategy holds promise to become a mainstream method for WBC separation.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: elps70113‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.