A high-sensitivity and clogging-free microfluidic impedance flow cytometer based on three-dimensional hydrodynamic focusing

Abstract

Microfluidic impedance flow cytometry has functioned as an enabling instrument in single-cell analysis, which, however, suffers from the limiting tradeoff between high sensitivity and clogging-free operation. In order to address this issue, this study presented a microfluidic impedance flow cytometer based on three-dimensional (3D) hydrodynamic focusing, in which the crossflow of conductive sample fluids and insulating sheath fluids was leveraged to centralize and restrict electric field lines to the sample fluid, thereby achieving high impedance sensitivity of single cells without the concern of channel blockage. Different from conventional impedance flow cytometry, in this study, impedance amplitude dips (rather than pulse singles) generated by single microparticles traveling through the 3D hydrodynamic focusing region were experimentally validated using microbeads. Based on the home-developed microfluidic impedance flow cytometer, high-sensitivity and clogging-free impedance profiles of three leukemia cell lines (K562, Jurkat, and HL-60) and four types of purified leukocytes (neutrophil, eosinophil, monocyte, and lymphocyte) were quantified as −8.01 ± 2.96%, −4.53 ± 1.09%, −6.36 ± 1.54%; −8.11 ± 0.84%, −7.23 ± 1.06%, −9.05 ± 2.00% and −5.68 ± 1.24%, respectively. When a recurrent neural network was adopted for cell-type classification, high classification accuracies of 93.9% for three leukemia cell lines and 87.8% for four types of purified leukocytes were achieved. This study presented a promising impedance flow cytometer that combines high sensitivity with sustainable working capabilities, potentially overcoming the limitations of conventional microfluidic impedance flow cytometry and significantly advancing its commercial development.

---

Introduction

Due to dimensional comparisons of microfluidics with cells, microfluidic platforms have been widely used for single-cell analysis.^1–4 Within this fields, microfluidic impedance flow cytometry, integrating microfabrication with impedance flow cytometry, has become a widely adopted method for quantifying single-cell bioelectrical properties. Compared to conventional counterparts, in microfluidic impedance flow cytometry, the geometrical dimensions of sensitive microchannels and strategic placements of sensitive microelectrodes can be finely controlled, resulting in highly sensitive and reproducible single-cell impedance measurements. Microfluidic impedance flow cytometry has enabled the characterization of diverse cellular phenotypes (e.g., cancer or mammalian cells, human pathogens, yeasts, and miscellaneous biosystems), validating the broad applicability of this type of flow cytometry.^5–9

Although various studies have been conducted in microfluidic impedance flow cytometry, its commercial instrumentation with actual applications in disease diagnosis or scientific studies remains limited, primarily due to the unresolved tradeoff between high sensitivity and clogging-free operation. More specifically, high-sensitivity impedance measurements of single cells (∼15 μm in diameter) required detection microchannels with critical dimensions of ∼20 μm. This design, however was prone to channel blockage, which has hindered commercial adoption. For example, in recent microfluidic impedance flow cytometry studies, detection microchannels with dimensions of 40 μm × 28 μm, 40 μm × 20 μm, 10 μm × 10 μm, 80 μm × 20 μm, 25 μm × 20 μm, 30 μm × 30 μm, and 20 μm × 20 μm were reported by the groups of Morgan,¹⁰ Caselli,¹¹ Huang,¹² Ye,¹³ Wang,¹⁴ Swami,¹⁵ and our own,¹⁶ respectively. These dimensions of microfluid channels can enable high-sensitivity detection at the single-cell level, but were often limited by channel blockage issues, resulting in compromised sustainable working hours.

In order to address the aforementioned bottleneck of microfluidic impedance flow cytometry, a two-dimensional (2D) hydrodynamic focusing method was developed, wherein a conductive sample fluid was sandwiched between two insulating sheath fluids. This configuration ensured that electric field lines were restricted within the sample fluid, resulting in high impedance sensitivity for single cells through enhanced channel dimensions. As pioneers in this field, Morgan et al. characterized micron-sized particles and bacteria using detection microchannels with enlarged dimensions of 200 μm in width and 30 μm in height.¹⁷ Recently, Wang's group optimized the viscosity of sheath fluids to classify three representative tumor cell lines using detection microchannels enlarged to 50 μm in width and 20 μm in height.¹⁸ In addition, based on the same principle of 2D hydrodynamic focusing, by our own group, the dimensions of detection microchannels were enlarged to 50 μm in width and 20 μm in height, realizing high-accuracy leukocyte differentials.¹⁹

Although 2D hydrodynamic focusing can enlarge channel dimensions to an extent, the height of detection microchannels remained constrained to ∼20 μm, making them still susceptible to blockage. This study presented a microfluidic impedance flow cytometer utilizing three-dimensional (3D) hydrodynamic focusing, where a conductive sample fluid (∼15 μm in diameter) was sandwiched cylindrically by insulating sheath fluids with a total diameter of ∼50 μm, enabling high-sensitivity, clogging-free single-cell impedance measurements. In order to maintain high-sensitivity measurements with 3D hydrodynamic focusing, high-resolution 3D printing was adopted for fabrication to precisely define geometrical dimensions of the sandwiched conductive sample fluid. Miscible insulating sheath fluids confined the electric field lines within the conductive sample during 3D hydrodynamic focusing while allowing subsequent release.

Leveraging this home-developed microfluidic impedance flow cytometer with 3D hydrodynamic focusing, significant impedance variations across three leukemia cell lines (K562, Jurkat, and HL-60) and four types of purified leukocytes (neutrophil NEU, eosinophil EOS, monocyte MON, and lymphocyte LYM) were captured without encountering channel blockage. Furthermore, a recurrent neural network was employed to process these impedance profiles, producing high classification accuracy for different cell types and validating the platform's capability for single-cell analysis.

Materials and methodology

Theoretical analysis

Fig. 1(a) illustrates the schematic of the high-sensitivity, clogging-free microfluidic impedance flow cytometer with 3D hydrodynamic focusing. The region of 3D hydrodynamic focusing can be segmented into three distinct functional zones: a focusing zone, a detection zone, and a diffusion zone. In the focusing zone, the sample fluid is hydrodynamically focused into a narrow stream by the surrounding low-conductivity sheath fluid, ensuring precise cell alignment prior to entering the detection zone. In the detection zone, the three-dimensional sheath flow confines electric field lines to a restricted region. When a cell traverses this zone, it obstructs the concentrated electric field lines, resulting in significant impedance variations. Finally, in the diffusion zone, diffusion and mixing between the sample and sheath fluids occur, yielding a homogeneous mixed solution with relatively high conductivity. In summary, this design effectively enhances single-cell impedance measurement sensitivity while avoiding the risk of channel blockage, thereby addressing the key issue in conventional microfluidic impedance flow cytometry.

Fig. 1 Schematic of the high-sensitivity, clogging-free microfluidic impedance flow cytometer with three-dimensional (3D) hydrodynamic focusing (a). A representative impedance profile of a travelling cell can be divided into three zones, with zone (i) and zone (iii) corresponding to cell incoming and leaving stages that blocks electric field lines while maintaining the liquid–liquid interface between the sample and the sheath fluids, thereby producing increased impedance amplitudes. In zone (ii), as the cell travels within the detection region, it expands the liquid–liquid interface while remaining coupled to the confined electric field, generating an impedance amplitude dip (b). A recurrent neural network was utilized to extract features and classify cell types based on impedance profiles (c).

Based on theoretical analysis, a representative impedance profile of a travelling cell can be divided into three zones. Zone (i) and zone (iii) correspond to the incoming and leaving stages of the travelling cell, where it blocks electric field lines without affecting the liquid–liquid interface between sample and sheath fluids, producing increases in impedance amplitudes. In zone (ii), as the cell travels within the detection region, it expands the liquid–liquid interface while remaining coupled to the restricted electric field, generating a dip of impedance amplitude (see Fig. 1(b)). With the unique impedance profiles, hundreds of features were extracted by an optimized recurrent neural network and subsequently utilized to produce high accuracies of cell-type classification, validating the platform's capability for single-cell analysis (see Fig. 1(c)).

Materials and cell preparation

The sample fluid was an electrically conductive 1× PBS solution (PBS) from ThermoFisher (USA), while the sheath fluid comprised an insulating 10% sucrose solution with osmotic pressure matched to that of the cells. This selection of fluids ensured that the electric field was confined to the sample stream, with minimal impact on cell viability. To verify the 3D hydrodynamic focusing system performance, the impedance of polystyrene microbeads (30, 20, and 10 μm diameter) obtained from Sigma Aldrich (USA) was measured.

Three leukemia cell lines (e.g., K562, Jurkat, and HL60) were purchased from Biology-Medicine Cell Resources of China. All cell lines were cultured in RPMI-1640 medium (10% fetal bovine serum) at 37 °C in a 5% CO₂ atmosphere.

Whole blood samples were collected from three healthy volunteers. The purified leukocytes were obtained using a fluorescent flow cytometer (Beckman Coulter of USA), such as NEU based on FITC-CD15 and APC-CD16, EOS based on APC-CD16 and PE-Siglec-8, MON based on APC-CD16 and PE-CD14, and LYM based on APC-CD3.

Experimental operation and data processing

The microfluidic impedance flow cytometry based on 3D hydrodynamic focusing was fabricated using projection micro-stereolithography of 3D printing technology (nanoArch S230, BMF, China) and positioned on an inverted microscope with a 10X objective lens (IX73 Olympus, Japan). Two 3D hydrodynamic focusing configurations with detection zone diameters of 100 μm and 50 μm were fabricated and characterized for comparative analysis. Microbeads and cells were suspended in phosphate buffer saline as the sample fluid, while a 10% sucrose solution was used as the sheath fluid. Both fluids were delivered to the 3D hydrodynamic focusing region via two syringe pumps (PHD Ultra, USA). Impedance variations were measured using a lock-in amplifier (Zurich Instrument, CH) at frequencies of 60, 400, 700, and 990 kHz.¹⁶

Single impedance profiles of four frequencies were extracted from raw impedance data by a home-developed code (MATLAB 2021b, MathWorks). Specifically, a band-stop filter was used to address the noise caused by the lightly jittery focusing interface and a Savitzky–Golay filter was used to address the noises from the raw data. Then, individual microbead impedance profiles were identified and extracted by applying a threshold criterion (compared with baseline impedance) to distinguish the signal from noise. Finally, an additional screening based on impedance amplitude was performed on each processed impedance profile to exclude those exhibiting abnormal variation levels.

The four-frequency impedance profiles were processed through a long short-term memory (LSTM) recurrent neural network (RNN) to extract 128 distinct features, enabling classification of either leukemia cell lines or the four types of purified leukocytes.¹⁹ Please refer to corresponding paragraphs in the SI for more details.

Ethical statement

Informed consents were obtained from human participants of this study.

Results and discussion

Validation

Fig. 2(a) and (b) validate the performance of the high-sensitivity, clogging-free microfluidic impedance flow cytometer incorporated with 3D hydrodynamic focusing, featuring a 100 μm diameter detection zone. More specifically, Fig. 2(a) shows microscopic images of 3D focusing achieved with different flow ratios between the conductive sample fluid and insulating sheath fluid. With increasing ratios of conductive sample fluid to insulating sheath fluid at a constant sample flow rate, a narrower sample stream was formed in the detection zone, resulting in greater impedance variations during bead passage and elevated baseline impedance levels. When the sample fluid flow rate was increased at a fixed sample-to-sheath fluid ratio, a diffusion zone was formed with gradual fluid mixing, leading to elevated baseline impedance. In contrast, lower sample flow rates promote enhanced diffusion between the sample and sheath fluids in the detection zone. Fig. 2(b) shows raw impedance profiles of 20 and 30 μm microbeads traveling through the 3D hydrodynamic focusing region at a 1 : 5 ratio of the conductive sample to insulating sheath fluids. The larger microbeads significantly expended the liquid–liquid interface between the sample and sheath fluids, resulting in greater impedance variations.

Fig. 2 3D hydrodynamic focusing characterization: numerical simulations and microscopic images at a diameter of 100 μm under various conductive sample flow rates and conductive sample-to-insulating sheath fluid ratios (a). Microbead impedance results: raw impedance profiles and size distribution histograms for microbeads of 20 and 30 μm at 5 μL min⁻¹ conductive sample with a 1 : 5 conductive sample/insulating sheath ratio (b). Numerical simulations and microscopic images of 3D hydrodynamic focusing at a diameter of 50 μm (c). Microbead impedance results: raw impedance profiles and size distribution histograms for beads of 10 and 20 μm under identical experimental conditions to the 100 μm configuration (d).

Fig. 2(c) and (d) validate the performance of the high-sensitivity, clogging-free microfluidic impedance flow cytometer incorporated with 3D hydrodynamic focusing, featuring a 50 μm-diameter detection zone. More specifically, Fig. 2(c) shows microscopic images of 3D focusing under a 1 : 5 ratio of the conductive sample and insulating sheath fluids. At a sample flow rate of 5 μL min⁻¹ with a 1 : 5 sample-to-sheath fluid ratio (conductive sample/insulating sheath fluids), the focusing width is approximately 15 μm, consistent to the detection cell dimensions. Fig. 2(d) shows raw impedance profiles of 10 and 20 μm microbeads traveling through the 3D hydrodynamic focusing region at a 1 : 5 sample to sheath fluid ratio (conductive sample/insulating sheath). Compared to 3D hydrodynamic focusing with a diameter of 100 μm, the 50 μm configuration generates larger impedance variations owing to its reduced focusing width.

Application

As a demonstrative application of the home-developed high-sensitivity, clogging-free microfluidic impedance flow cytometer with 3D hydrodynamic focusing, Fig. 3(a) shows preliminary impedance profiles of individual leukemia cell lines (K562, Jurkat and HL-60). The maximal impedance amplitude dips at 60 kHz were quantified as −8.01 ± 2.96% (n_cell = 5152), −4.53 ± 1.09% (n_cell = 6413) and −6.36 ± 1.54% (n_cell = 6613), respectively. These results of impedance variations correlated with the observation where K562 cells were the largest while Jurkat cells were the smallest (see Table S1). Fig. 3(b) and (c) show the training and validation curves of classification accuracy and loss versus iteration, as well as the confusion matrix for differentiating three leukemia cell lines, achieving 93.8% accuracy. Note that the training and validation curves demonstrated no substantial disparities, suggesting no significant overfitting in the deep neural network. Based on the results of cell-type classification, it was observed that the lowest true positive rate of classification was 92.9% of HL-60, attributable to their intermediate impedance amplitude dips between K562 and Jurkat cells, leading to higher misclassified probability.

Fig. 3 Application: (a)–(c) classification of three leukemia cell lines (K562, Jurkat and HL-60), including preliminary impedance profiles (a), RNN based training curves (b) and confusion matrix (c) in classifying leukemia cell lines. (d)–(f) Classification of four types of purified leukocytes (neutrophil NEU, eosinophil EOS, monocyte MON, and lymphocyte LYM), including preliminary impedance profiles (d), RNN based training curves (e) and confusion matrix (f) in classifying the four types of purified leukocytes.

Fig. 3(d) shows preliminary impedance profiles of individual leukocyte subtypes (NEU, EOS, MON and LYM) travelling through the 50 μm-diameter 3D hydrodynamic focusing structure. The maximum impedance amplitude dips at 60 kHz were quantified as −8.11 ± 0.84% (n_cell = 3586), −7.23 ± 1.06% (n_cell = 7213), −9.05 ± 2.00% (n_cell = 4095) and −5.68 ± 1.24% (n_cell = 6706), respectively. These impedance variations correlated with cellular size observation, with MON exhibiting the largest and LYM the smallest diameters (see Table S1). Fig. 3(e) and (f) show training curves and confusion matrix for classifying four-type purified leukocytes, achieving 87.8% accuracy. Since the amplitude dip variances of monocytes were higher than granulocytes and lymphocytes, they had a higher misclassification rate among leukocyte subtypes, which was consistent with the deep neural network classification results. Detailed performance metrics, including precision, recall, and F1 scores, were provided in Table S2.

Conclusions

In conclusion, this study demonstrated a high-performance microfluidic impedance flow cytometer based on 3D hydrodynamic focusing. Due to the confined electric field lines in the 3D hydrodynamic focusing detection domain, high sensitivities of single-cell impedance profiles were obtained based on this flow cytometry. Furthermore, the insulating sheath fluid configuration effectively enhanced the detection microchannel dimensions, enabling long sustainable working hours of this flow cytometry.

This study also demonstrated preliminary applications of this microfluidic impedance flow cytometry in classifying leukemia cell lines and purified leukocyte subtypes with both high sensitivity and long sustainable working hours. Future developments will focus on the classification of benign and malignant leukocyte mixtures based on this approach, with potential to transform clinical leukocyte differential analysis.

Conflicts of interest

The authors declare no conflicts of interest.

Data availability

Supplementary information is available. See DOI: https://doi.org/10.1039/D5LC00571J.

The authors confirm that the data supporting the findings of this study are available within the main article. For further data access, please contact the corresponding author.

Acknowledgements

This work was supported by the National Natural Science Foundation of China with grants 62331025 and 62121003.

References

1. M. Righetto, C. Brandi, R. Reale and F. Caselli, Integrating impedance cytometry with other microfluidic tools towards multifunctional single-cell analysis platforms, Lab Chip, 2025, 25(5), 1316–1341,  10.1039/d4lc00957f.
2. S. Shen, X. Liu, K. Fan, H. Bai, X. Li and H. Li, Stabilizing and accelerating secondary flow in ultralong spiral channel for high-throughput cell manipulation, Anal. Chem., 2024, 96(28), 11412–11421,  DOI:10.1021/acs.analchem.4c01549.
3. S. Shen, F. Zhang, H. Li, Y. Wu, K. Fan, L. Zhao, H. Li and N. Sun, Multi-vortex regulation in a simple semicircular microchannel with ordered micro-obstacles for high-throughput buffer exchange, Anal. Chem., 2025, 97(6), 3661–3670,  DOI:10.1021/acs.analchem.4c06310.
4. T. Tang, H. Zhao, S. Shen, L. Yang and C. T. Lim, Enhancing single-cell encapsulation in droplet microfluidics with fine-tunable on-chip sample enrichment, Microsyst. Nanoeng., 2024, 10, 3,  DOI:10.1038/s41378-023-00631-y.
5. H. Daguerre, M. Solsona, J. Cottet, M. Gauthier, P. Renaud and A. Bolopion, Positional dependence of particles and cells in microfluidic electrical impedance flow cytometry: origin, challenges and opportunities, Lab Chip, 2020, 20(20), 3665–3689,  10.1039/d0lc00616e.
6. C. Honrado, P. Bisegna, N. S. Swami and F. Caselli, Single-cell microfluidic impedance cytometry: from raw signals to cell phenotypes using data analytics, Lab Chip, 2021, 21(1), 22–54,  10.1039/d0lc00840k.
7. S. Zhu, X. Zhang, Z. Zhou, Y. Han, N. Xiang and Z. Ni, Microfluidic impedance cytometry for single-cell sensing: Review on electrode configurations, Talanta, 2021, 233, 122571,  DOI:10.1016/j.talanta.2021.122571.
8. M. Wang, H. Liang, X. Chen, D. Chen, J. Wang, Y. Zhang and J. Chen, Developments of conventional and microfluidic flow cytometry enabling high-throughput characterization of single Cells, Biosensors, 2022, 12(7), 443,  DOI:10.3390/bios12070443.
9. M. A. Mansor, M. R. Ahmad, M. Petru and S. S. Rahimian Koloor, An impedance flow cytometry with integrated dual microneedle for electrical properties characterization of single cell, Artif. Cells, Nanomed., Biotechnol., 2023, 51(1), 371–383,  DOI:10.1080/21691401.2023.2239274.
10. J. Chen, X. Zou, D. C. Spencer and H. Morgan, Single-cell electro-mechanical shear flow deformability cytometry, Microsyst. Nanoeng., 2024, 10(1), 173,  DOI:10.1038/s41378-024-00810-5.
11. C. Troiano, A. De Ninno, B. Casciaro, F. Riccitelli, Y. Park, L. Businaro, R. Massoud, M. L. Mangoni, P. Bisegna, L. Stella and F. Caselli, Rapid assessment of susceptibility of bacteria and erythrocytes to antimicrobial peptides by single-cell impedance cytometry, ACS Sens., 2023, 8(7), 2572–2582,  DOI:10.1021/acssensors.3c00256.
12. X. Luan, P. Liu, D. Huang, H. Zhao, Y. Li, S. Sun, W. Zhang, L. Zhang, M. Li, T. Zhi, Y. Zhao and C. Huang, piRT-IFC: Physics-informed real-time impedance flow cytometry for the characterization of cellular intrinsic electrical properties, Microsyst. Nanoeng., 2023, 9, 77,  DOI:10.1038/s41378-023-00545-9.
13. M. Liang, Q. Tang, J. Zhong and Y. Ai, Machine learning empowered multi-stress level electromechanical phenotyping for high-dimensional single cell analysis, Biosens. Bioelectron., 2023, 225, 115086,  DOI:10.1016/j.bios.2023.115086.
14. Y. Feng, J. Zhu, H. Chai, W. He, L. Huang and W. Wang, Impedance-based multimodal electrical-mechanical intrinsic flow cytometry, Small, 2023, 19(45), 2303416,  DOI:10.1002/smll.202303416.
15. A. Salahi, A. Rane, L. Xiao, C. Honrado, X. Li, L. Jin and N. S. Swami, Single-cell assessment of the modulation of macrophage activation by ex vivo intervertebral discs using impedance cytometry, Biosens. Bioelectron., 2022, 210, 114346,  DOI:10.1016/j.bios.2022.114346.
16. X. Chen, X. Huang, J. Zhang, M. Wang, D. Chen, Y. Li, X. Qin, J. Wang and J. Chen, Leukocyte differential based on an imaging and impedance flow cytometry of microfluidics coupled with deep neural networks, Cytometry, Part A, 2024, 105(5), 315–322,  DOI:10.1002/cyto.a.24823.
17. C. Bernabini, D. Holmes and H. Morgan, Micro-impedance cytometry for detection and analysis of micron-sized particles and bacteria, Lab Chip, 2011, 11(3), 407–412,  10.1039/c0lc00099j.
18. J. Zhu, Y. Feng, H. Chai, F. Liang, Z. Cheng and W. Wang, Performance-enhanced clogging-free viscous sheath constriction impedance flow cytometry, Lab Chip, 2023, 23(11), 2531–2539,  10.1039/d3lc00178d.
19. M. Wang, J. Zhang, X. Chen, Y. Li, X. Huang, J. Wang, Y. Li, X. Huo and J. Chen, Microfluidic impedance flow cytometer leveraging virtual constriction microchannel and its application in leukocyte differential, Microsyst. Nanoeng., 2024, 10(1), 192,  DOI:10.1038/s41378-024-00833-y.

---

Footnote

† Xiao Chen and Tingxuan Fang are co-first authors.

---