Secondary Logo

Share this article on:

A Novel Implementation of Magnetic Levitation to Quantify Leukocyte Size, Morphology, and Magnetic Properties to Identify Patients With Sepsis

Andersen, Mikkel S.*,‡; Lu, Shulin*; Lopez, Gregory J.*; Lassen, Annmarie T.; Shapiro, Nathan I.*; Ghiran, Ionita C.

doi: 10.1097/SHK.0000000000001139
Clinical Science Aspects
Editor's Choice

Background: We have developed a novel, easily implementable methodology using magnetic levitation to quantify circulating leukocyte size, morphology, and magnetic properties, which may help in rapid, bedside screening for sepsis.

Objective: Our objectives were to describe our methodological approach to leukocyte assessment, and to perform a pilot investigation to test the ability of magnetic levitation to identify and quantify changes in leukocyte size, shape, density, and/or paramagnetic properties in healthy controls and septic patients.

Methods: This prospective, observational cohort study was performed in a 56,000/y visit emergency department (ED) and affiliated outpatient phlebotomy laboratory. Inclusion criteria were admittance to the hospital with suspected or confirmed infection for the septic group, and we enrolled the controls from ED/outpatient patients without infection or acute illness. The bench-top experiments were performed using magnetic levitation to visualize the leukocytes. We primary sought to compare septic patients with noninfected controls and secondary to assess the association with sepsis severity. Our covariates were area, length, width, roundness, and standard deviation (SD) of levitation height. We used unpaired t test and area under the curve (AUC) for the assessment of accuracy in distinguishing between septic and control patients.

Results: We enrolled 39 noninfected controls and 22 septic patients. Our analyses of septic patients compared with controls showed: mean cell area in pixels (px) 562 ± 111 vs. 410 ± 45, P < 0.0001, AUC = 0.89 (0.80–0.98); length (px), 29 ± 2.5 vs. 25 ± 1.9, P < 0.0001, AUC = 0.90 (0.83–0.98); and width (px), 27 ± 2.4 vs. 23 ± 1.5, P < 0.0001, AUC = 0.92 (0.84–0.99). Cell roundness: 2.1 ± 1.0 vs. 2.2 ± 1.2, P = 0.8, AUC = 0.51. SD of the levitation height (px) was 72 ± 25 vs. 47 ± 16, P < 0.001, AUC = 0.80 (0.67–0.93).

Conclusions: Septic patients had circulating leukocytes with especially increased size parameters, which distinguished sepsis from noninfected patients with promising high accuracy. This portal-device compatible technology shows promise as a potential bedside diagnostic.

*Department of Emergency Medicine, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Massachusetts

Department of Medicine, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Massachusetts

Department of Emergency Medicine, Odense University Hospital, University of Southern Denmark, Odense, Denmark

Address reprint requests to Mikkel S. Andersen, BSc, Odense University Hospital, Odense C, 5000, Denmark. E-mail:,,

Received 18 October, 2017

Revised 2 November, 2017

Accepted 14 March, 2018

NIS and ICG equally contributed to this work.

The work was supported from grants from the NIH (R01-HL096795 and R21-TW009915) and the Bill and Melinda Gates Foundation (OPP1032683) to ICG; NIS has current or prior support from Rapid Pathogen Screening, Thermo-Fisher, and Siemens and the National Institutes of Health. ICG and NIS are coinventors of the MELISSA technique, which is covered by a pending patent.

The authors report no conflicts of interest.

Back to Top | Article Outline


Sepsis is a severe illness that affects more than 750,000 patients annually in the United States alone (1). Given the more than 20% mortality rate for septic shock (2), it is imperative to discover early diagnostic detection tests. Circulating leukocytes, especially neutrophil granulocytes (PMNs), become activated in sepsis (3) leading to leukocytosis and a characteristic “left shift” of immature white blood cells (4, 5). These are clinical parameters typically used in the clinical assessment of the septic patient.

During sepsis, leukocytes especially circulating PMNs, undergo significant changes in number, size, roundness, granularity, and chemotaxis, due to mobilization of immature, larger neutrophils from bone marrow, cell degranulation, activation by circulating bacterial fragments, activated complement proteins, and intracellular generation of reactive oxygen species (6, 7). These changes in circulating leukocytes positively correlate with sepsis disease severity and prognosis (reviewed recently in Zonneveld et al. (8)). Currently, precise quantification of leukocytes changes during sepsis requires the use of specialized equipment such as flow cytometry, bright field and fluorescence microscopy, and dedicated methods such as microfluidic-based cell migration, “volume, conductivity and scatter” (VCS) measurements, and trans-well chamber chemotactic and chemokineses assays.

Magnetic levitation is a relatively new technique that uses a uniform magnetic field to measure and separate cells based on their density and magnetic properties (9–12). Although the magnetic levitation technology offers a variety of opportunities for functional interrogation of cells, for this specific study, we use it as a tool to isolate leukocytes from circulating red blood cells (RBCs), visualize leukocytes, and measure size, shape, density, and paramagnetic property parameters of leukocytes. Herein, we exploit the ability of magnetic levitation to separate cells in real time using a cell type-specific concentration of Gadolinium solution, and then record and measure the cell size and densities of circulating leukocytes, using only a small amount of unprocessed blood. We hypothesize that by assessing changes in circulating leukocytes: size (increase of cell contents when activated), shape (due to actin ruffles formation), and distribution in levitation height (due to intracellular paramagnetic ROS generation and degranulation), we will distinguish sepsis patients from noninfected controls. The technology may be implemented as an inexpensive, portable bedside test.

Back to Top | Article Outline


Study design and population

This project is a prospective, observational cohort study of emergency department (ED) patients with suspected infection compared with noninfected control patients either from the ED or outpatient phlebotomy laboratory. For the patients with sepsis, the inclusion criteria were age 18 years or older with suspected infection or confirmed infection and admission to the hospital. The exclusion criteria were inability to obtain informed consent. Noninfected control patients were collected either in the ED or in affiliated outpatient phlebotomy. The inclusion criteria were adult patients 18 years or older, no suspicion of sepsis or infection, and the absence of SIRS criteria. The study was approved by the Beth Israel Deaconess Medical Center Institutional Review Board.

Back to Top | Article Outline

Demographics and clinical data description

For demographic info, we collect age and sex and results of white blood cell count and differential analysis from the clinical laboratory. We further classified sepsis patients according to their severity upon presentation according to modified ACCP/SCCM sepsis criteria (13). The patients’ sepsis syndromes are recorded at the time of enrollment.

Back to Top | Article Outline

Magnetic levitation technology, preparation, and setup

We use a bright-field microscope tipped on the side (90 degrees) and fitted with a magnetic levitation device, consisting of two magnets positioned with the same poles facing each other, separated by a 1 mm space sufficient for insertion of a 5-mm-long squared glass capillary tube (Fig. 1, A and B). An example of images obtained by levitating blood collected from a control and a septic shock patient are depicted in Figure 2A.

Fig. 1

Fig. 1

Fig. 2

Fig. 2

Back to Top | Article Outline

Blood procedure

Blood was collected and analyzed within 2 h to minimize additional activation of leukocytes. Half a microliter of EDTA whole blood was mixed with 100 μL of 21 mM gadolinium solution Gd3+ and incubated for 20 min. Ten micrometer density reference beads (1.06 g/mL) were added to a final concentration of 5 × 10−4% (Flow-Count Fluorospheres, Beckman Coulter, Inc., Fullerton, Calif). Samples were loaded into squared 1 × 1 × 25 mm capillary tubes and placed between the two magnets (Fig. 1B) as previously described (11) and magnetically focused for 20 min (incubation period in the magnetic levitation device). Cells were imaged using an Olympus PlanFl 20 × 0.5 NA lens, and acquired using a Retiga EXi Blue cooled CCD camera (Qimaging, Surrey, BC, Canada) controlled by Image Pro Plus 7.2 (Media Cybernetics, Rockville, Md).

Back to Top | Article Outline

Software analysis

Acquired images were analyzed in a blinded fashion. Images were saved in .tiff format. Before the analysis could commence, a threshold was set depending on the brightness of the background of the images to maintain uniform results. The cells were analyzed using the “automatic” settings based on the calculated segmentation values. In rare cases, when certain cells failed to be correctly recognized or separated from the surrounding cells by the software, the manual function with a 1,600% zoom was used. Images with unfocused cells, too few reference beads, and/or excessive brightness/darkness were rejected from analysis. The data were exported to Excel, and mean, median, and standard deviation (SD) were calculated for 30 cells per sample.

Back to Top | Article Outline

Covariates: measurements and descriptions

Leukocyte covariates of interest were area, length, width, roundness, and SD of levitation height. Figure 2B illustrates the methodology for covariate measurements. Area was defined as the total number of pixels within the encircled cell. Length was defined as the vertical row of pixels through the center of the object and width was defined as the horizontal row of pixels also through the center. Roundness was calculated by the software using the following formula: (Perimeter2)/(4*Π*area), where Perimeter2 reports the length of the outline of each object. Complete circular objects will have a roundness = 1; other shapes will have a roundness >1. For levitation height, we measured the distance from the cell centroid pixel from the reference density beads level. We analyzed up to five reference beads Y coordinate and performed an average height base on that. Formula: Levitation Height = [Leukocyte centroid − Y coordinate value] − [Average reference beads centroid − Y coordinate value]; the image coordinate system is (0,0) in the top left, which is the reason for the subtraction method (the cells are below the reference beads, which gives them a higher Y coordinate value). For each sample, we calculate the SD to assess the range of the levitation heights of circulating leukocytes. Although activated PMNs have decreased density due to intravascular cell degranulation resulting in an increase in the levitation height, they also generate significant intracellular ROS which increases the magnetic signature of the cell, thus lowering its levitation height. When both events happen in the same cell, which is likely the case on most circulating PMNs, the overall impact on the levitation height would depend on the ratio of these two opposing forces.

Back to Top | Article Outline

Statistical considerations

We used the median values for each sample and performed an unpaired t test to compare parameters for sepsis and noninfected controls. To evaluate the ability to distinguish noninfected controls and septic patients, we calculated the area under the curve (AUC) for the receiver-operating characteristic curves. As a secondary analysis, we assessed the association with sepsis syndrome and display the results; however, we did not anticipate finding statistically significance between the groups based on limited sample size.

Back to Top | Article Outline


Study population

We recruited and analyzed 39 noninfected control patients and 22 patients with infection. For the patients with infection, the sepsis syndrome at enrollment was sepsis (7), severe sepsis (sepsis plus organ dysfunction) (9), and septic shock (6). Subjects were enrolled between March 2016 and November 2016. Table 1 shows the demographics, baseline clinical characteristics, and laboratory testing for our patient population.

Table 1

Table 1

Back to Top | Article Outline

Parameters of leukocyte size

Our results show that infected patients had circulating leukocytes with significantly higher mean cell area in pixels (px), as compared with noninfected controls (561.6 ± 111.3 vs. 409.7 ± 44.8, P < 0.0001) (Fig. 3A) with an AUC for discriminating infection from controls of 0.89 (0.80–0.98) (Table 2). Cell length (px) was increased in infected patients (29.4 ± 2.5 vs. 25.2 ± 1.9, P < 0.0001) with an AUC = 0.90 (0.83–0.98) (Fig. 3B). Similarly, the leukocytes from infected patients were significantly wider than those in controls 26.8 ± 2.4 vs. 22.5 ± 1.5, P < 0.0001 with an AUC = 0.92 (0.84–0.99) (Fig. 3C).

Fig. 3

Fig. 3

Table 2

Table 2

Back to Top | Article Outline

Leukocyte roundness

The software autocalculated leukocyte roundness value (rv), which showed no difference between sepsis and control patients, with an overall roundness value of 2.2 ± 1.0 rv for septic patients, and, a roundness value of 2.2 ± 1.2 rv, P = 0.8 with an AUC = 0.51 (0.35–0.66) for noninfected controls (Table 2).

Back to Top | Article Outline

Magnetic properties represented by SD

We measured the median levitation height, which for sepsis patients was 293.1 ± 61.4 px and 288.8 ± 23.7 px for controls, P = 0.8, with an AUC = 0.52 (0.35–0.69), which proved insignificant (Fig. 3E). The cells’ levitation height varied more within a sample in sepsis patients compared with a noninfected control, likely due to alterations of density after granular content release, and also change in magnetic properties due to intracellular ROS generation. We next compared the SDs of the levitation heights of the samples, which was 71.5 ± 24.8 px for sepsis patients and 47.4 ± 16.0 px for controls, P = 0.0003, with an AUC = 0.80 (0.67; 0.93) (Table 2).

Back to Top | Article Outline

Correlation with complete blood cell and differential laboratory testing

To understand the relationship between leukocyte parameters quantified by magnetic levitation and parameters from complete blood cell testing, we assessed the correlation between these findings in the patients with infection (noninfected controls did not have these parameters available). There was incomplete correlation between mean cell area and neutrophil percentage (r = 0.32), lymphocyte percentage (r = −0.48), band percentage (r = 0.13), and white blood cell (WBC) count (r = 0.19) on the complete blood count (cbc) with differential. This was similar for the correlation between length and neutrophil percentage (r = 0.43), lymphocyte percentage (r = −0.54), band percentage (r = 0.07), and WBC count (r = 0.18), as well as width and neutrophil percentage (r = 0.38), lymphocyte percentage (r = −0.54), and band percentage (r = 0.07) and WBC count (r = 0.25). Further studies are needed to determine if magnetic levitation is either superior or provides new information compared with the cbc with differential.

Back to Top | Article Outline


Previous reports suggest that during sepsis, circulating pathogen and host chemotactic factors mobilize and activate leukocytes, promote increase in neutrophil size, and stimulate cell degranulation (14); however, these are not routinely assessed parameters. Other methods, such as Automated Hematology Analysis (AHA), have been introduced to measure morphology, such as size and granularity of cells, using VCS (volume, conductivity, and scatter) (15). Our findings are consistent with other studies using VCS to measure neutrophils from controls and septic patients, which all show septic patients had larger white cells than those of patients with localized infections and noninfected controls (15–17). Flow cytometry has also been used to determine the size of the cells and shows similar results for septic patients, patients with localized infection, and noninfected controls (18, 19). Furthermore, recent reports have shown a positive correlation between the size of activated neutrophil granulocytes and size increase in infected samples (20, 21).

During activation, previous observations suggest leukocyte plasma membranes have a shape that includes “pointed spikes” due to increased exocytosis and release of certain granular content (3). At the resolution used in our approach, we were unable to detect these differences, which is not unexpected given the resolution of the 20 × 0.50 objective of about 0.6 μm. The neutrophil granulocyte has several stages of activation, each of which has specific morphological parameters, so even though they increase in size, the “spikiness” might not parallel that pattern (21). Thus, although using our system, the roundness parameter was not found to be useful in distinguishing infected patients from noninfected.

When activated by bacterial by products or host anaphylatoxins, circulating neutrophils generate reactive oxygen species (ROS), and exocytose granular content (22). The highly paramagnetic ROS increases the magnetic signature of activated neutrophils, positioning the cells closer to the bottom magnet. Conversely, degranulation promotes a significant decrease in cell density process, which would position the neutrophils closer to the middistance between the two magnets. These concurrent yet opposing mechanisms explain the observed wide distribution range in the levitation height of neutrophils from septic patients compared with noninfected controls. Further investigation is required to better understand the dynamics of these conflicting mechanisms and their implications for sepsis screening and detection using magnetic levitation height.

Furthermore, we show promising diagnostic accuracy in a convenience sample, especially for the size parameters (area, length, and width) ranging from 0.89 to 0.92 AUC. For example, calculating the diagnostic operating characteristics with a cutoff of 23.3 px for width yielded a sensitivity of 91% and a specificity of 74% with a positive likelihood ratio of 3.5. If we alter the cutoff to 24.6 px the sensitivity decreases to 82%, but the specificity increases to 92%, with a positive likelihood ratio of 10.6. In comparison to AHA technology using VCS parameter, the sensitivity was 83% and specificity 78% (16). The challenge of AHA is that it requires an expensive piece of equipment, the UniCel DxH 800 Coulter system and highly trained technicians. As previously described, other technologies such as flow cytometry have proven useful (18, 19). However, these are more complex and time-consuming procedures. Furthermore, published diagnostic accuracies show variable results, most of which are less discriminatory compared with our preliminary findings (23, 24).

For this study, we measure all the leukocytes in a field from whole blood as we intend for the technology to ultimately be implemented at the bedside as a quick, inexpensive, and easy test. Even though the leukocytes mainly consist of neutrophils and lymphocytes, 60% and 30%, respectively, for a healthy individual (25), an alternative approach would be to include only subpopulations such as neutrophil granulocytes in our analyses, which will be applied in future studies.

Back to Top | Article Outline


There are a number of limitations to this investigation. First, as this was an initial investigation into the methodology, our sample size was small. Second, we enrolled a convenience sample population, which is prone to selection bias. For future studies, we will need to test the technology in an undifferentiated population to assess validity of the test. Third, we compared patients to noninfected control presumed to not be acutely ill; this may be different in clinical application to acutely ill patients when we are trying to decide on whether infection is the source of the illness. Fourth, we did not compare the leukocyte measurements to commonly used laboratory parameters such as a complete blood count with differential to determine if the new method is superior to existing technologies. Fifth, we performed an offline analysis. Although the bedside application is theoretically identical and will ultimately be automatic, this will need to be verified. Sixth, we do not know how comorbidities will impact our test. Finally, this was an observational trial, and future investigations on how the results of testing inform practice are needed to improve patient-oriented outcomes.

Back to Top | Article Outline


In this pilot investigation, we describe a novel methodology using magnetic levitation to separate leukocytes from circulating RBCs and assess leukocyte parameters. The technique demonstrated a promising diagnostic accuracy for identifying patients with an infection regarding size parameters (area, length, and width) especially. Infected patients had leukocytes that were larger in size and had a pattern consistent with more variable density and magnetic properties. Secondary analysis suggests that these parameters may also be associated with sepsis severity. This technique, which is potentially inexpensive, portable, and implementable at the bedside, shows promise as a novel screening and diagnostic test for infection.

Back to Top | Article Outline


1. Angus DC, Linde-Zwirble WT, Lidicker J, Clermont G, Carcillo J, Pinsky MR. Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care. Crit Care Med 29 7:1303–1310, 2001.
2. Wang HE, Shapiro NI, Angus DC, Yealy DM. National estimates of severe sepsis in United States emergency departments. Crit Care Med 35 8:1928–1936, 2007.
3. Nauseef WM, Borregaard N. Neutrophils at work. Nat Immunol 15 7:602–611, 2014.
4. Abramson N, Melton B. Leukocytosis: basics of clinical assessment. Am Fam Physician 62 9:2053–2060, 2000.
5. Ploppa A, Schmidt V, Hientz A, Reutershan J, Haeberle HA, Nohe B. Mechanisms of leukocyte distribution during sepsis: an experimental study on the interdependence of cell activation, shear stress and endothelial injury. Crit Care 14 6:R201, 2010.
6. Kolaczkowska E, Kubes P. Neutrophil recruitment and function in health and inflammation. Nat Rev Immunol 13 3:159–175, 2013.
7. King EG, Bauza GJ, Mella JR, Remick DG. Pathophysiologic mechanisms in septic shock. Lab Invest 94 1:4–12, 2014.
8. Zonneveld R, Molema G, Plotz FB. Analyzing neutrophil morphology, mechanics, and motility in sepsis: options and challenges for novel bedside technologies. Crit Care Med 44 1:218–228, 2016.
9. Tasoglu S, Khoory JA, Tekin HC, Thomas C, Karnoub AE, Ghiran IC, Demirci U. Levitational image cytometry with temporal resolution. Adv Mater 27 26:3901–3908, 2015.
10. Knowlton S, Yu CH, Jain N, Ghiran IC, Tasoglu S. Smart-phone based magnetic levitation for measuring densities. PLoS One 10 8:e0134400, 2015.
11. Felton EJ, Velasquez A, Lu S, Murphy RO, ElKhal A, Mazor O, Gorelik P, Sharda A, Ghiran IC. Detection and quantification of subtle changes in red blood cell density using a cell phone. Lab Chip 16 17:3286–3295, 2016.
12. Andersen MS, Ericson E, Lu S, Richard M, Gregory M, Ogembo JG, Mazor O, Gorelik P, Shapiro N, Sharda A, et al. Detection of membrane-bound and soluble antigens by magnetic levitation. Lab Chip 17 20:3462–3473, 2017.
13. Levy MM, Fink MP, Marshall JC, Abraham E, Angus D, Cook D, Cohen J, Opal SM, Vincent JL, Ramsay G. 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Conference. Intensive Care Med 29 4:530–538, 2003.
14. O’Flaherty JT, Kreutzer DL, Ward PA. Chemotactic factor influences on the aggregation, swelling, and foreign surface adhesiveness of human leukocytes. Am J Pathol 90 3:537–550, 1978.
15. Celik IH, Demirel G, Aksoy HT, Erdeve O, Tuncer E, Biyikli Z, Dilmen U. Automated determination of neutrophil VCS parameters in diagnosis and treatment efficacy of neonatal sepsis. Pediatr Res 71 1:121–125, 2012.
16. Lee AJ, Kim SG. Mean cell volumes of neutrophils and monocytes are promising markers of sepsis in elderly patients. Blood Res 48 3:193–197, 2013.
17. Mardi D, Fwity B, Lobmann R, Ambrosch A. Mean cell volume of neutrophils and monocytes compared with C-reactive protein, interleukin-6 and white blood cell count for prediction of sepsis and nonsystemic bacterial infections. Int J Lab Hematol 32 4:410–418, 2010.
18. Roussel M, Benard C, Ly-Sunnaram B, Fest T. Refining the white blood cell differential: the first flow cytometry routine application. Cytometry A 77 6:552–563, 2010.
19. Roussel M, Gros A, Gacouin A, Le Meur N, Le Tulzo Y, Fest T. Toward new insights on the white blood cell differential by flow cytometry: a proof of concept study on the sepsis model. Cytometry B Clin Cytom 82 6:345–352, 2012.
20. Brinkmann V, Reichard U, Goosmann C, Fauler B, Uhlemann Y, Weiss DS, Weinrauch Y, Zychlinsky A. Neutrophil extracellular traps kill bacteria. Science 303 5663:1532–1535, 2004.
21. Iba T, Hashiguchi N, Nagaoka I, Tabe Y, Murai M. Neutrophil cell death in response to infection and its relation to coagulation. J Intensive Care 1 1:13, 2013.
22. Mayer-Scholl A, Averhoff P, Zychlinsky A. How do neutrophils and pathogens interact? Curr Opin Microbiol 7 1:62–66, 2004.
23. Venet F, Lepape A, Monneret G. Clinical review: flow cytometry perspectives in the ICU—from diagnosis of infection to monitoring of injury-induced immune dysfunctions. Crit Care 15 5:231, 2011.
24. Umlauf VN, Dreschers S, Orlikowsky TW. Flow cytometry in the detection of neonatal sepsis. Int J Pediatr 2013:763191, 2013.
25. Ehrlich P LA, Myers W, Woodhead G. Histology of the Blood: Normal and Pathological. Cambridge: Cambridge University Press; 1900.

Leukocytes; magnetic fields; magnets; point-of-care systems; sepsis; systemic inflammatory response syndrome

© 2019 by the Shock Society