Cytometry Part B (Clinical Cytometry) 78B:31–36 (2010)

Clinical Evaluation of a Simple Image Cytometerfor CD4 Enumeration on HIV-Infected Patients

Xiao Li,1 Christian Breukers,1 Aurel Ymeti,1 Kovit Pattanapanyasat,2{

Kasama Sukapirom,2 Leon W. M. M. Terstappen,1* and Jan Greve11Department of Medical Cell Biophysics, Faculty of Science and Technology, University of Twente,

Enschede,The Netherlands2Center of Excellence for Flow Cytometry, Office for Research and Development, Faculty of Medicine,

Siriraj Hospital, Mahidol University, Bangkok, Thailand

Background: Affordable, easy-to-use, and reliable CD41 T lymphocyte enumeration systems areneeded in resource-constrained settings to monitor HIV.

Methods: A simple image cytometer was used to count fluorescently labeled CD41 T and CD81 T lym-phocytes from CD3 immunomagnetically selected cells on blood specimens of 460 HIV-1-infectedpatients in Siriraj Hospital, Bangkok, Thailand. Results were compared with flow cytometry (FCM).

Results: CD41 T lymphocyte counts by image cytometer were comparable (R � 0.97) with those bythe FACSCount and the FACScan with a bias of 7.3 and 9.1%, respectively. At very low CD41 T lympho-cyte counts (�50/ll) some over-count outliers were observed by the FACScan and image cytometerwhen compared with the FACSCount. For CD8 enumeration, the image cytometer showed a good correla-tion (R 5 0.96) and a consistent undercount (�17%) when compared with the FACSCount.

Conclusions: Evaluation of the image cytometer for CD4 and CD8 enumeration demonstrated comparableresults with FCM on a population of HIV-1-infected patients. The image cytometer is a good alternativemethod for point-of-care settings in resource-constrained countries. VC 2009 Clinical Cytometry Society

Key terms: image cytometry; AIDS; CD4; CD8

How to cite this article: Li X, Breukers C, Ymeti A, Pattanapanyasat K, Sukapirom K, Terstappen LWMM,Greve J. Clinical evaluation of a simple image cytometer for CD4 enumeration on HIV-infected patients.Cytometry Part B 2010; 78B: 31–36.

It is estimated that in 2007, 33 million people wereliving with HIV, of whom �95% are in the developingcountries (1). Recently, inexpensive and generic antire-troviral treatment (ART) has become available to moreand more patients in middle- and low-income countries,such as Thailand. In Thailand, more than 1 million peo-ple are infected with HIV, and about 30,000 new infec-tions occur annually (2). The availability of affordableand reliable CD4 enumeration for the initiation andmonitoring of ART has become a critical issue.

Flow cytomety (FCM)-based methods for CD4þ T lym-phocyte enumeration are currently the most accepted.Single-platform (SP) FCM uses calibration beads (3) ormeasure the volume to obtain absolute counts (4). Dual-platform (DP) FCM calculates the absolute CD4þ T lym-phocyte count by multiplying the lymphocyte countobtained from an automatic hematology analyzer with

the CD4 percentage obtained by FCM. The latter meth-ods are usually less accurate especially when testingblood samples after 6–12 h of blood collection (5,6).Both methods are expensive in equipment, mainte-nance, technician training, and assay cost (7).

{K. Pattanapanyasat is a Senior Research Scholar of the ThailandResearch Fund.

Grant sponsor: STW, The Dutch Technology Foundation; Grantnumber: TGT. 6146.

*Correspondence to: Leon W. M. M. Terstappen, Faculty of Sciencesand Technology, Medical Cell Biophysics, University of Twente, P.O.Box 217, Drienerlolaan 5, 7500 AE Enschede, The Netherlands.

E-mail: l.w.m.m.terstappen@utwente.nlReceived 22 April 2009; Revision 5 June 2009; Accepted 10 June

2009Published online 7 July 2009 in Wiley InterScience (www.

interscience.wiley.com).DOI: 10.1002/cyto.b.20488

VC 2009 Clinical Cytometry Society

Affordable, easy-to-use, and reliable CD4þ T lympho-cyte enumeration systems are urgently needed and vari-ous groups are attempting to achieve this goal (8–16).Earlier, we reported the development of an imagecytometer for enumeration of CD4þ T and CD8þ T lym-phocytes (10–12). In this system, immunomagneticselection of CD3þ T lymphocytes is combined withimmunofluorescent labeling of CD4-phycoerythrin (PE)and CD8-peridinin-chlorophyll-protein (PerCP). AfterCD3þ T lymphocytes in whole blood are immunomag-netically attracted to an analysis surface, the fluorescentimages of the CD4þ T lymphocytes and CD8þ T lympho-cytes are captured and the individual cells are identifiedand counted. In this study, we evaluate the performanceof this system in 460 HIV-1-infected patients in Thailand.CD4þ T and CD8þ T lymphocyte counts and CD4/CD8ratios in whole blood samples obtained by the imagecytometer were compared with the results fromFACSCount and FACScan. The FACSCount is a dedicatedSP FCM for CD4þ T and CD8þ T lymphocyte countingand has been extensively validated and is used inresource-poor settings (4,17,18). The FACScan, a DPFCM, is the accepted standard method for CD4enumeration in Thailand (5).

MATERIALS AND METHODS

Patients and Blood Samples

Blood specimens from 460 HIV-1-infected patientsfrom the Siriraj Hospital, Bangkok, Thailand were usedfor this study. All blood specimens were collected insterile K3EDTA vacutainer blood collection tubes byvenipuncture and processed within 10 h after blooddraw. The specimens were leftover clinical specimensfrom the Department of Immunology, Faculty ofMedicine, Siriraj Hospital. The patients consisted of 215male and 245 female, ranging in age from 18 to 75 years(mean 40 � 10). This study was approved by the EthicsCommittee of the Faculty of Medicine, Siriraj Hospital,Mahidol University.

Methods

Image cytometry. Sample preparation was per-formed as follows: to 100 ll of whole blood, 10 ll ofreagent cocktail, which contains 3 ll of 0.655 mg/mlCD3-FerroFluid (CD3FF, clone: CRIS-7, isotype: mouseIgG2a-j, Veridex LLC, Raritan, NJ), 1 ll of 12.5 lg/mlCD4PE (clone: RPA-T4, isotype: Mouse IgG1-j, BDPharmingen), 3 ll of 6.25 lg/ml CD8PerCP (clone: SK1,isotype: Mouse IgG1-j, BD ), and 3 ll of system buffer(Veridex LLC, Raritan, NJ), was added and mixed. After15 min incubation, the sample was diluted with 290 llof system buffer to a final volume of 400 ll. Approxi-mately 340 ll of the sample solution was transferredinto the analysis chamber. The chamber was pluggedand placed into a magnet assembly (MagNest

VR

; VeridexLLC, Raritan, NJ). After 20 min magnetic separation, thesample was ready to be analyzed.

Fluorescent mages were captured using the camerasoftware (CCDOPs, SBIG, USA). The blood volume corre-sponding to each captured image is 1.16 ll. Imageswere taken at three different positions by manuallymoving the chamber arriving at a total blood volumeanalyzed of 3.48 ll. This approach was chosen to obtaina theoretical statistical Poisson variation of �5.4% at 100cells/ll and �7.6% at 50 cells/ll (10,12). At each posi-tion, the PE and PerCP filters were manually changedresulting in three CD4 and three CD8 captured andstored images per sample. This process takes �3 min. Adedicated image analysis algorithm was written inImageJ, a public-domain Java image processing softwarepackage (nih, MD) (19). It runs on a single-board com-puter (665 MHz processor with 240 MB RAM) of theimage cytometer. The algorithm needs �120–180 s toanalyze the three CD4 and three CD8 images that con-tain �58–1,393 cells per image, which translates into50–1,200 cells/ll.

FACSCount. The FACSCount system consists of abench-top instrument, paired tubes, controls, and soft-ware. The instrument is equipped with a green laser.The paired tubes contain calibration beads, CD4PE/CD3PE-Cychrome5 (PE-Cy5), and CD8PE/CD3PE-Cy5,respectively, to enumerate CD3þ T lymphocytes, CD4þ

T, and CD8þ T lymphocytes and to calculate the CD4/CD8 ratio. The controls are used for quality control ofthe system. Sample preparation and acquisition wereperformed according to the manufacturer’s recommen-dation (BD Biosciences ) (20).

FACScan. CD4þ T lymphocyte count was obtainedusing FACScan (BD Biosciences) and a hematologyanalyzer (Sysmex XT2000, TOA Medical Electronics,Kobe, Japan). The absolute number of lymphocytes wasdetermined by a hematology analyzer. The percentage ofCD4þ T lymphocytes from total lymphocytes wasobtained by FACScan using CellQUEST software. Thistest was performed by the routine clinical laboratory ofthe Siriraj Hospital, Bangkok, Thailand.

Statistical Analysis

Based on the CD4þ T, CD8þ T lymphocyte countsand CD4/CD8 ratio obtained from different methods,linear regression lines were drawn and correlation coeffi-cients (R) were calculated. Bland–Altman plots (21,22)were used to evaluate the interchangeability betweenmethods. In Bland–Altman plots, the average of theCD4þ T or CD8þ T lymphocyte counts or CD4/CD8ratio obtained from two methods is plotted on thehorizontal axis, and the difference/average (%) is plottedon the vertical axis. The solid line in the plot representsthe bias (the average difference between the two meth-ods), and the dashed lines in the plot illustrate theupper and lower limits of agreement (�1.96 SD). Toexamine possible differences of the CD4þ T lymphocytecounts obtained by these methods in the significant clin-ical range for HIV, statistical analysis of the blood speci-mens with CD4þ T lymphocyte count �200/ll and

32 LI ET AL.

Cytometry Part B: Clinical Cytometry

CD4þ T lymphocyte count >200/ll was performedseparately.

RESULTS

CD4 Enumeration

Blood specimens of HIVþ patients were tested byimage cytometry, FACSCount, and FACScan, respectively,and the obtained CD4þ T lymphocyte counts were com-pared with each other. Figure 1 shows the Bland–Altmanplots of the comparisons of CD4þ T lymphocyte countsof HIVþ patients between the different methods, andTable 1 gives the statistical analysis. In panels A1–A3 allsamples are shown, in Panels B1–B3 only those withCD4þ T lymphocyte counts �200/ll, and in PanelsC1–C3 those with CD4þ T lymphocyte counts �50/ll.Comparison of FACScan with FACSCount showed a cor-relation of R ¼ 0.98 with a slope of 0.92. Bland–Altman

plots showed the interchangeability between the twoFCM methods with a bias of �2.3%. Comparison ofimage cytometry with the FACSCount showed a correla-tion of R ¼ 0.99 with a slope of 1.04 and a bias of 7.3%.Comparison of image cytometry and FACScan showed acorrelation of R ¼ 0.97 with a slope of 1.10 and a biasof 9.1%. It should be noted that in the very low CD4þ Tlymphocyte count region (�50/ll), outliers wereobserved in all three comparisons. Consequently, thebiases were raised because of the over-count outliers(Figs. A1–A3).

To examine the possible differences of the CD4þ Tlymphocyte counts obtained by these methods in theclinically significant range of CD4þ T lymphocyte countsof �200/ll for HIVþ adults, statistical analysis of allblood specimens, blood specimens with CD4þ T lym-phocyte count � 200/ll, and CD4þ T lymphocytecounts of more than 200/ll was performed separately.

FIG. 1. Method comparisons of CD4þ T lymphocyte counts of HIVþ patients. Panels A show the Bland–Altman plots of all samples; Panels B ofsamples with CD4 counts �200/ll and Panels C of samples with CD4 counts �50/ll. FACScan versus the FACSCount A1, B1, and C1; Imagecytometry versus FACSCount A2, B2, and C2; Image cytometry versus the FACScan A3, B3, and C3. See Table 1 for statistical analysis.

IMAGE CYTOMETER FOR CD4 AND CD8 ENUMERATION 33

Cytometry Part B: Clinical Cytometry

Slopes and correlation coefficients (R) from the linearregression plots and biases (%) and limits of agreement(%) from the Bland–Altman plots were summarized inTable 1.

The linear regression plots of these three comparisonsin the CD4þ T lymphocyte count range of more than200/ll followed the same tendency as the plots in thewhole CD4þ T lymphocyte counts range.

The Bland–Altman plots of these three comparisons inthe CD4þ T lymphocyte count range of more than 200/ll showed the ability to interchange the different meth-ods. For the comparison between the FACScan and theFACSCount, the bias of �6.5% indicated that CD4þ Tlymphocyte count by the FACScan was consistentlyabout 6.5% less than that by FACSCount, which agreedwell with the slope of 0.92 for adults in the CD4þ Tlymphocyte count range of more than 200/ll. Theincrease in the bias between the FACScan and the FACS-Count can be attributed to the over-count outliers forthe very low CD4þ T lymphocyte counts (bias: �2.3%for all, �6.5% for >200/ll, and 13.4% for �200/ll).

For the comparison between the image cytometry andthe FACSCount, the bias (3.5%) indicated that the twomethods were interchangeable, which also agreed wellwith the slope of 1.04 for adults in the CD4þ T lympho-cyte count range of more than 200/ll. The bias betweenthe image cytometry and the FACSCount increased from3.5 (>200/ll) to 7.3% (the whole range) because of theover-count outliers observed for some of the very lowCD4þ T lymphocyte counts.

For the comparison between the image cytometry andthe FACScan, the bias indicated that CD4þ T lymphocytecount by the image cytometry was consistently about10% higher than that of the FACScan. The biases almostremained the same (9.8% for >200/ll and 9.1% for thewhole range).

For the CD4þ T lymphocyte count, �200/ll good cor-relations (R ¼ 0.97–0.98) were still achieved. However,for the comparisons between the FACScan and the FACS-Count and between the image cytometry and the FACS-Count, the biases of the Bland–Altman plots were raisedbecause of the over-count outliers at CD4þ T lympho-cyte counts �50/ll. (Table 1).

To investigate the different performance of differentmethods for samples with very low CD4þ T lymphocytecount (�50/ll), the Bland–Altman analyses were per-formed on samples in this region in a different way; thedifference of absolute CD4þ T lymphocyte counts wasplotted on the vertical axis (as shown in Fig. 1C).Results indicated that on an average the FACScancounted 7 cells/ll (limits of agreement: �7 to 21 cells/ll) more than the FACSCount did; the image cytomtrycounted 6 cells/ll (limits of agreement: �12 to 25 cells/ll) more than the FACSCount did. And, the biasbetween the image cytometry and the FACScan is only 1cell/ll (limits of agreement: �21 to 23 cells/ll).

CD8 Enumeration

CD8þ T lymphocyte counts were obtained by imagecytometry and by FACSCount. Thirty-three of the 460HIVþ patient samples were excluded from the compari-son, because the FACSCount reports CD8þ T lympho-cytes �2,000/ll as greater than 2,000. Figure 2 showsthe linear regression plot (A) and Bland–Altman plots(B) of CD8þ T lymphocyte counts obtained by the imagecytometry and FACSCount. Comparison of the resultsfrom 427 HIVþ patients showed a correlation of R ¼0.96 with a slope of 0.83. Bland–Altman plots showedthe interchangeability between the two methods with abias of �16.8% (LL ¼ �42.8%, UL ¼ 8.8%).

CD4/CD8 Ratio Determination

The CD4/CD8 ratios were determined by dividing theCD4þ T lymphocyte count by the CD8þ T lymphocytecount. The comparisons were performed between theimage cytometer and the FACSCount. Regression analysisshowed a correlations of R ¼ 0.99 with a slope of 1.22.Bland–Altman plot shows the interchangeabilitybetween two methods with a bias of: 23.5% (LL ¼�9.4%, UL ¼ 56.4%). The �24% higher CD4/CD8 ratiosobtained with the image cytometer when comparedwith the FACSCount are a results of the slightly lowerCD8þ T lymphocyte counts and the slightly higherCD4þ T lymphocyte counts in the image cytometer.

Table IComparison of CD4þ T Lymphocyte Counts Obtained by Image Cytometry, FACSCount, and FACScan for All HIVþ Patients,

Those with CD4 Count Range �200/ll, >200/ll, and <50/ll

Range All

FACSCan vs. FACSCount

All

Image cytometryvs. FACSCount

All

Image cytometryvs. FACSCan

�200/ll >200/ll <50/ll �200/ll >200/ll <50/ll �200/ll >200/ll <50/ll

# 435 92 343 26 460 106 354 29 437 97 340 27Slope 0.92 0.97 0.92 1.26 1.04 1.07 1.04 1.25 1.10 1.10 1.10 1.04R 0.98 0.98 0.97 0.79 0.99 0.98 0.98 0.77 0.97 0.97 0.96 0.80Bias �2.3 13.4 �6.5 7.2 7.3 20.6 3.5 6.4 9.1 6.3 9.8 0.7L L �42.2 �56.0 �26.8 �6.5 �33.7 �51.9 �17.5 �12.3 �24.9 �49.1 �14.8 �21.2U L 37.6 82.8 13.7 20.9 48.2 93.2 24.6 25.1 43.0 61.7 34.5 22.6

Slopes and correlation coefficients (R) are derived from the linear regression plots, and biases (%) and limits of agreement (%)from the Bland–Altman plots.

LL, lower limit; UL, upper limit; #, number of patients.

34 LI ET AL.

Cytometry Part B: Clinical Cytometry

DISCUSSION

CD4þ T lymphocyte counts by image cytometer werecomparable with those by the FACSCount and FACScan.Across the CD4þ T lymphocyte count range, the imagecytometer counted �7.3% more when compared withthe FACSCount and �9.1% more when compared withthe FACScan.

At very low CD4þ T lymphocyte counts (1–50/ll),some outliers were observed in the FACScan and imagecytometer when compared with the FACSCount. TheFACScan and the image cytometer counted �7 and 6cells/ll more than the FACSCount did, respectively.Nevertheless, the biases in this range may not influencethe clinical decision making, because ART is given whenCD4þ T lymphocyte counts are less than 200/ll (23).The omission of the duplicate CD3 count in the morerecent releases of the FACSCount will likely make theresults of the FACSCount at very low CD4þ T lympho-cyte counts (<50/ll) less robust (20,24). These biasesare unlikely to affect the monitoring either, as long asthe same method was used all the time.

A likely explanation for some of the over-count out-liers is the increasing contribution of cell debris, fluoro-phore clumps, and autofluorescent junk in the imageswith lower cell counts. This can lead to the misclassifi-cation of events by the algorithm. Review of the imagesof the outliers indeed showed that some had a largerbackground. In FCM, some of these issues are avoidedby the use of forward- and side-scatter and quantitativeanalysis of the fluorescence signals. Further develop-ment of the image analysis algorithms used in the imagecytometer may improve the ability to discriminate theintact target cells from debris and reduce the outliers.Another explanation for the outliers observed betweenmethods in the very low CD4þ T lymphocyte count(<50/ll) may be contributed to the higher variation ofthe FACSCount method (CV > 10% at CD4þ T lympho-cyte counts below 50/ll) (20).

The possible explanations for the CD8 undercount(�16.8%) by the image cytometer when compared withthe FACSCount may be attributed to the CD8þdim T lym-phocytes (about 10% of total CD8þ T lymphocytes).They were either not detected by the image cytometerbecause of the limited sensitivity of the instrument ornot counted by the image analysis because of the inter-ference of the cross-talked CD4PE cells in CD8PerCPimage (12). Similarly, the CD8 undercount was alsofound in some dedicated CD4 and CD8 enumerationflow cytometric methods, such as Guava PCA method(18), because of its limited sensitivity and gate settingby which only the CD8þbright T lymphocytes arecounted.

This study demonstrated that image cytometry has thepotential to simplify enumeration of CD4þ and CD8þ Tlymphocytes. Before such a system can be introduced inresource-poor countries, several issues will have to beresolved. Although the optical components are in afixed position, and no alignment or calibration isrequired, the optical filters for PE and PerCP detectionas well as the chamber are moved manually to obtainthe three PE and three PerCP images. The system wasoperated in Siriraj Hospital in Bangkok, Thailand on a 12V battery. However, the operator was integral to the de-velopment of the system and by no means representsthe ultimate user of the system. The assay uses liquidreagents that need to be refrigerated and require accu-rate pipetting of blood and reagents. The componentcost of the image cytometer was US �$3,000, and theassay cost US �$3 for CD4 and CD8 enumeration. Theactual manufacturing cost and resale price can only beestablished in concert with an industrial partner thathas an interest in distribution of the systems inresource-constrained countries.

In conclusion, image cytometry can be a good alterna-tive method for FCM-based methods in point-of-care set-tings of resource-constrained countries.

FIG. 2. Linear regression plots (A) and Bland–Altman plots (B) of comparisons of CD8 counts obtained by the image cytometry and the FACSCountfor 427 HIVþ patients.

IMAGE CYTOMETER FOR CD4 AND CD8 ENUMERATION 35

Cytometry Part B: Clinical Cytometry

ACKNOWLEDGMENTS

The authors thank Mr. Charin Thepthai in the routineLab of the Siriraj Hospital, Bangkok, Thailand for theadministrative assistance. They acknowledge all patientsand healthy donors whose blood was tested in this study.

LITERATURE CITED

1. Report on the global HIV/AIDS epidemic 2008. ‘‘UNAIDS/08.25E /JC1510E’’. ISBN 978 92 9 173711 6. Available at: www.unaids.org.

2. Thai Working Group on HIV/AIDS Projection. Projections for HIV/AIDS in Thailand: 2000–2020. Bangkok: Ministry of Public Health;2001.

3. Schnizlein-Bick CT, Spritzler J, Wilkening CL, Nicholson JKA, O’Gor-man MRG. Evaluation of TruCountTM absolute-count tubes for deter-mining CD4 and CD8 cell numbers in human immunodeficiencyvirus-positive adults. Clin Diagn Lab Immunol 2000;7:336–343.

4. Pattanapanyasat K, Lerdwana S, Noulsri E, Chaowanachan T,Wasinrapee P, Sakulploy N, Pobkeeree V, Suksripanich O, Thanpra-sertsuk S, Spira TJ, Tappero JW, Levine WC. Evaluation of a newsingle-parameter volumetric flow cytometer (CyFlowgreen) for enu-meration of absolute CD4þ T lymphpocytes in human immunodefi-ciency virus type 1-infected Thai Patients. Clin Diagn Lab Immunol2005;12:1416–1424.

5. Glencross D, Scott LE, Jani IV, Barnett D, Janossy G. CD45-assistedPanLeucogating for accurate, cost-effective dual-platform CD4þT-cell enumeration. Cytometry B 2002;50B:69–77.

6. Mandy F, Janossy G, Bergeron M, Pilon R, Faucher S. AffordableCD4 T-cell enumeration for resource-limited regions: A status reportfor 2008. Cytometry Part B 2008;74B:S27–S39.

7. Medical Mission Institute. Access to antiretroviral therapy in devel-oping Countries: A continuum of care approach; Guidelines andpolicy issues: Appropriate laborotary methods in the management ofHIV infection. In: CI/CIDSE Policy Workshop. Wuerzburg, Germany:Medical Mission Institute; 2001.

8. Janossy G, Jani IV, Brando B. New trends in affordable CD4þ T-cellenumeration by flow cytometry in HIV/AIDS. Clin Appl ImmunolRev 2003;4:91–107.

9. Rodriguez WR, Christodoulides N, Floriano PN, Graham S, MohantyS, Dixon M, Hsiang M, Peter T, Zavahir S, Thior I, Romanovicz D,Bernard B, Goodey AP, Walker BD, McDevitt JT. A microchip CD4counting method for HIV monitoring in resource-poor settings.PLoS Med 2005;2:e182.

10. Li X, Ymeti A, Lunter B, Tibbe AG, Terstappen LW, Greve J. CD4þT lymphocytes enumeration by an easy-to-use single platform imagecytometer for HIV monitoring in resource-constrained settings.Cytometry Part B 2007; 72B:397–407.

11. Ymeti A, Li X, Lunter B, Breukers C, Tibbe AG, Terstappen LW,Greve J. A single platform image cytometer for resource-poor set-tings to monitor disease progression in HIV infection. CytometryPart A 2007;71A:132–142.

12. Li X, Breukers C, Ymeti A, Lunter B, Terstappen LWMM, Greve J.CD4 and CD8 enumeration for HIV monitoring in resource-con-strained settings. Cytometry Part B 2009;76B:231–236.

13. Srithanaviboonchai K, Rungruengthanakit K, Nouanthong P, Pata S,Sirisanthana T, Kasinrerk W. Novel low-cost assay for the monitoringof CD4 counts in HIV-infected individuals. J Acquir Immune DeficSyndr 2008;47:135–139.

14. Moon SJ, Keles HO, Ozcan A, Khademhosseini A, Hæggstrom E,Kuritzkes D, Demirci U. Integrating microfluidics and lenslessimaging for point-of-care testing. Biosens Bioelectron 2009;24:3208–3214.

15. Nouanthong P, Pata S, Sirisanthana T, Kasinrerk W. A simple manualrosetting method for absolute CD4 lymphocyte counting inresource-limited countries. Clin Vaccine Immunol 2006;13:598–601.

16. Thorslund S, Larsson R, Bergquist J, Nikolajeff F, Sanchez J. A PDMS-based disposable microfluidic sensor for CD4þ lymphocyte count-ing. Biomed Microdevices 2008;10:851–857.

17. Strauss K, Hannet I, Engels S, Shiba A, Ward DM, Ullery S, JingujiMG, Valinsky J, Barnett D, Orfao A, Kestens L. Performance evalua-tion of the FACSCount system: A dedicated system for clinical cellu-lar analysis. Cytometry 1996;26:52–59.

18. Pattanapanyasat K, Phuang-Ngern Y, Lerdwana S, Wasinrapee P,Sakulploy N, Noulsri E, Thepthai C, McNicholl JM. Evaluation of asingle-platform microcapillary flow cytometer for enumeration ofabsolute CD4þ T-lymphocyte counts in HIV-1 infected Thaipatients. Cytometry Part B 2007;72B:387–396.

19. Rasband WS. ImageJ. U.S. National Institutes of Health:Bethesda,MD; 1997–2005. Avialable at: http://rsb.info.nih.gov/ij/.

20. BD Bioscience Immunocytometry Systems. Technical SpecificationsBD FACSCount System. Available at: http://www.bdbiosciences.com/cgi-bin/literature/view?part_num¼23-6575-01.

21. Bland JM, Altman DG. Statistical methods for assessing agreementbetween two methods of clinical measurement. Lancet 1986;1:307–310.

22. Pollock MA, Jefferson SG, Kane JW, Lomax K, MacKinnon G,Winnard CB. Method Comparison—A different approach. Ann ClinBiochem 1992;29:556–560.

23. Hoover DR, Graham NM, Chen B, Taylor JM, Phair J, Zhou SY,Munoz A. Effect of CD4þ cell count measurement variability onstaging HIV-1 infection. J Acquir Immune Defic Syndr 1992;5:794–802.

24. Pattanapanyasat K, Sukapirom K, Kowawisatsut L, Thepthai C. NewBD FACSCount CD4 reagent system for simultaneous enumerationof percent and absolute CD4 T-lymphocytes in HIV-1-infectedpediatric patients. Cytometry Part B 2008;74B:S98–106.

36 LI ET AL.

Cytometry Part B: Clinical Cytometry