Lab on a Chip

Publ

ishe

d on

07

Febr

uary

201

4. D

ownl

oade

d by

Cor

nell

Uni

vers

ity o

n 29

/04/

2014

20:

48:4

0.

PAPER View Article OnlineView Journal | View Issue

Lab ChipThis journal is © The Royal Society of Chemistry 2014

a Sibley School of Mechanical and Aerospace Engineering, Cornell University, Ithaca,

NY 14853, USA. E-mail: de54@cornell.edubDepartment of Biomedical Engineering, Cornell University, Ithaca, NY 14853, USAc Division of Nutritional Sciences, Cornell University, Ithaca, NY 14853, USA

Cite this: Lab Chip, 2014, 14, 1437

Received 11th December 2013,Accepted 7th February 2014

DOI: 10.1039/c3lc51375k

www.rsc.org/loc

A smartphone platform for the quantificationof vitamin D levels

Seoho Lee,a Vlad Oncescu,a Matt Mancuso,b Saurabh Mehtac and David Erickson*a

Vitamin D deficiency has been linked to a number of diseases and adverse outcomes including:

osteoporosis, infections, diabetes, cardiovascular diseases, and even cancer. At present the vast majority

of vitamin D testing is performed in large-scale laboratories at the request of a physician as part of an

annual panel of blood tests. Here we present a system for rapid quantification of vitamin D levels on a

smartphone. The system consists of a smartphone accessory, an app, and a test strip that allows the col-

orimetric detection of 25-hydroxyvitamin D using a novel gold nanoparticle-based immunoassay. We

show that the system can be used to accurately measure physiological levels of 25-hydroxyvitamin D

with accuracy better than 15 nM and a precision of 10 nM. We compare our system with well-established

ELISA test kits for serum samples of unknown concentration and demonstrate equivalency of the results.

We envision this as the first step towards the development of the NutriPhone, a comprehensive system

for the analysis of multiple vitamins and micronutrients on a smartphone.

Introduction

Vitamin D deficiency is prevalent worldwide1 and has beenlinked to diabetes,2 cardiovascular diseases,3,4 infections,5,6 andcancer7,8 in addition to bone diseases such as osteoporosis9

and rickets.10 Deficiency is commonly defined in humans bymeasuring serum concentrations of 25-hydroxyvitamin D, or25(OH)D, the circulating form of vitamin D. The currentlyrecommended cutoffs for vitamin D deficiency and insuffi-ciency are defined by 25(OH)D levels below 50 nM and 80 nMrespectively. Low levels of vitamin D can usually be improvedby consuming supplements and increasing skin exposureto sunlight.11 However, many people are unaware of theirvitamin D status due to the lack of simple methods fordiagnosing and tracking 25(OH)D concentrations in the blood.The ability to monitor vitamin D levels may help those at riskto detect low levels at an earlier stage and enable correctiveinterventions before the occurrence of potentially irreversibledamage. Reliance on supplements alone without knowledge oftrue vitamin D status may fall short in addressing severevitamin D deficiencies (<20 nM) or in some cases lead topossible vitamin D intoxication.11,12 At present, most 25(OH)Dtests are conducted using high-performance liquid chromatog-raphy (HPLC), liquid chromatography-tandem mass spec-trometry (LC-MS) techniques, or certain immunoassays such as

radioimmunoassay (RIA) and enzyme-linked immunosorbentassay (ELISA).13–15 These methods are time consuming, costlyand difficult to perform in most settings outside sophisticatedlaboratories. Recently, a lateral flow assay for vitamin D(Test4D; Nanospeed Diagnostics) has been introduced,16 demon-strating the commercial need for such rapid, point-of-carevitamin D analysis. Smartphone based diagnostics offer addi-tional advantages over lateral flow assays including the abilitiesto quantify results, track changes with time, and provide user-free error tracking.

We present here the vitaAID – vitamin AuNP-based Immuno-assay Device – a system for rapid quantification of vitamin Dlevels on a smartphone. This is achieved through the develop-ment of: a smartphone accessory, app, and novel gold nanoparti-cle (AuNP) based colorimetric competitive direct-antigenimmunoassay. This assay enables us to quantify 25(OH)Dmolecules whose small size (~400 g mol−1) restricts theirbinding to more than one antibody at a time.17,18 Gold nano-particle based colorimetric assays19,20 coupled with smartphonetechnology have previously been used for the quantificationof mercury contamination in water samples as demonstratedby Wei et al.21 In addition, smartphone platforms wereused for the detection of biomarkers in complex bodilyfluids,22–25 thereby demonstrating the ability to do rapid,personalized diagnostics.

In this paper we describe the vitaAID and show how25(OH)D levels can be quantified by evaluating brightnessdifferences between the detection area and a reference areaon the vitaAID test strip. We then demonstrate that thevitaAID system can be used to quantify vitamin D levels by

, 2014, 14, 1437–1442 | 1437

Lab on a ChipPaper

Publ

ishe

d on

07

Febr

uary

201

4. D

ownl

oade

d by

Cor

nell

Uni

vers

ity o

n 29

/04/

2014

20:

48:4

0.

View Article Online

evaluating serum samples with unknown 25(OH)D concentra-tions and comparing the results to those obtained using acommercial ELISA method.

ResultsvitaAID system and test strip surface chemistry

The vitaAID system consists of a smartphone accessory andcustom test strip as shown in Fig. 1a. The accessory has beendesigned to minimize the effect of variability in external light-ing conditions with a 5 mm amber LED (λ = 592 nm) embed-ded in a polydimethylsiloxane (PDMS) diffuser used toilluminate the back of the test strip. The test strip is com-posed of a detection area and reference area. The referencearea allows the algorithm to further adjust to differences insmartphone cameras. The detection area that enables the col-orimetric reaction to occur consists of a surface-immobilizedlayer of 25(OH)D. There is currently no standard method forobtaining a chemically stable coating layer of 25(OH)D due tothe small size and absence of linkable functional groups onthe 25(OH)D. We addressed this by developing a novel25(OH)D coating process onto a fused Si-based substrate thatserves as the detection area on the test strip. As shown inFig. 1b, we coated the 3-aminopropyltriethoxysilane (APTES)and polystyrene-co-maleic anhydride (PSMA) layers sequen-tially on a fused Si substrate and covalently linked

1438 | Lab Chip, 2014, 14, 1437–1442

Fig. 1 a) vitaAID accessory on a iPhone with the inset showing the compontion of the APTES, maleic anhydride and aminopropylated 25(OH)D3 layers t

Fig. 2 a) Test strip and schematic of the gold nanoparticle-based immunoation at different AuNP-anti-25(OH)D3 incubation times on the detection are

aminopropylated 25(OH)D to achieve a stable 25(OH)D coat-ing. In order to validate this method, the surface treatmentswere characterized by Fourier-transform infrared spectroscopy(FT-IR) as shown in Fig. 1b. The initial introduction of APTESlayer on the Si substrate is evident from the transmittancepeak at 1654 cm−1 that is associated with primary aminegroups (–NH2).

26 The PSMA coating is confirmed by theappearance of peaks at 1850 and 1780 cm−1 which have beenlinked with maleic anhydrides in other studies.26–28 Lastly, thepeak at 3300 cm−1 corresponds to the formation of the second-ary amine (R2NH) bond between the PSMA and aminopropyl-ated 25(OH)D and verifies the 25(OH)D immobilization.29

Surface-based gold nanoparticle immunoassay

The colorimetric reaction on the detection area of the teststrip is based on a novel surface-based gold nanoparticle-based immunoassay as shown in Fig. 2a. When a sample isapplied onto the detection area of the test strip, only the anti-body conjugates that are not bound to the 25(OH)D presentin the initial sample are captured by the coated 25(OH)D onthe surface. The colorimetric signals from the immobilizedAuNP-antibody conjugates are then amplified using a silverenhancement scheme as the silver ions undergo reduction onthe surface of the AuNP to increase their size and therebyincrease the sensitivity of the system. For samples with high

This journal is © The Royal Society of Chemistry 2014

ents of the accessory b) FT-IR spectra showing the chemical composi-hat constitute the detection area.

ssay reaction on the detection area b) variation in predicted concentra-a for 0 nM sample 25(OH)D.

Lab on a Chip Paper

Publ

ishe

d on

07

Febr

uary

201

4. D

ownl

oade

d by

Cor

nell

Uni

vers

ity o

n 29

/04/

2014

20:

48:4

0.

View Article Online

vitamin D levels, most of the antibody conjugates are occu-pied with 25(OH)D from the initial sample, resulting in only asubtle change in the colorimetric signal on the test strip. Forsamples with low vitamin D levels, the test strip develops anintense color that reflects the high number of antibody conju-gates bound on the surface. The steps involved in the prepa-ration of the AuNP-anti-25(OH)D conjugate solution and thedetection area are discussed in the methods section.

A critical step during testing is the incubation of the AuNP-anti-25(OH)D sample solution on the test strip's detectionarea. It is important to characterize the time it takes for theAuNP-anti-25(OH)D to immobilize in order to minimize thetotal assay time and to improve accuracy. In Fig. 2b, we showthe effect of different incubation times on the brightness dif-ference between the detection and reference areas of the teststrip (ΔH) for a sample without 25(OH)D. After 6 h the bright-ness difference is within 10% of that obtained after a typical12 h overnight incubation. This indicates that a 6 h incuba-tion time is sufficient in order to obtain accurate results dur-ing such vitamin D measurements. The incubation time canhowever be significantly reduced by using the obtained nega-tive exponential relationship to determine the minimum incu-bation time for ensuring that sufficient conjugate bindingevents have occurred on the detection area. In a recent study,Kai et al.30 have demonstrated that the incorporation of micro-fluidic channels could further reduce such incubation timesto the order of minutes through improved binding kinetics.

Quantification of colorimetric reaction

Once the competitive binding of AuNP-anti-25(OH)D wasperformed on the test strip, the quantification of the 25(OH)D

This journal is © The Royal Society of Chemistry 2014

Fig. 3 a) Colorimetric variation on the test strip at different known 25(OH)Dthe detection area and reference area ΔH at different known 25(OH)D concentra

levels in the initial sample can be achieved using the vitaAIDsmartphone platform. First, the colorimetric change is cap-tured using the smartphone's camera after inserting the teststrip in the vitaAID accessory. In Fig. 3a we show the colori-metric change in the transparent fused silica detectionregion at different known concentrations of 25(OH)D andthe colorimetric change on the white reference region due tochanges in exposure time of the camera. The iPhone auto-matically adjusts its camera exposure times in response tothe amount of light passing through the detection region.Therefore, by comparing the differences in brightnessbetween the detection area (Bdet) and reference area (Bref) wecan estimate the concentration of 25(OH)D independent ofthe exposure setting of the iPhone camera. In Fig. 3b weshow that ΔH = Bdet − Bref can be correlated to the 25(OH)Dconcentration in the initial sample. A second order polyno-mial was then fitted onto this calibration curve in order toobtain a function such that [25(OH)D] = f (ΔH).

Fig. 3c shows the algorithm that allows the quantificationof 25(OH)D across the entire range of physiological values.First, the detection area is scanned for silver enhancedregions where AuNP-anti-25(OH)D is bound. This is impor-tant because at higher 25(OH)D concentrations in the initialsample, the detection area rarely exhibits a uniform colori-metric change. A 100px by 100px area around the high inten-sity silver enhanced region is taken and the brightness isaveraged across all the pixels in that area. The brightness is acoordinate in the Hue Saturation Brightness (HSB) colormodel and is computed from the Red Green Blue (RGB) aver-age values as described elsewhere.31 The same steps are thenperformed on the reference region and an average brightnessis calculated. Once the brightness difference between the

Lab Chip, 2014, 14, 1437–1442 | 1439

concentrations b) calibration curve showing brightness difference betweentions c) algorithm used in quantifying 25(OH)D levels on test strip.

Lab on a ChipPaper

Publ

ishe

d on

07

Febr

uary

201

4. D

ownl

oade

d by

Cor

nell

Uni

vers

ity o

n 29

/04/

2014

20:

48:4

0.

View Article Online

detection area and the reference area is computed, the algo-rithm uses a second order polynomial [25(OH)D] = f (ΔH)derived from Fig. 3b to calculate the 25(OH)D concentrationin the initial sample.

Accuracy and precision compared to ELISA

The accuracy of the vitaAID system was characterized byquantifying 25(OH)D levels in solutions of known concentra-tions and comparing them to results obtained using a com-mercial ELISA kit. In Fig. 4a we show variations in predicted25(OH)D levels for 3 standard solutions of 15 nM, 40 nM and110 nM. At each concentration, 3 test strips were used andthe maximum error across all the samples and concentra-tions was measured as 14.1 nM at a concentration of 40 nM.The maximum average inter-sample difference was 9.8 nMalso at 40 nM. These errors are of the same order as those ofELISA methods for the quantification of 25(OH)D levels.

In Fig. 4b, 3 serum samples of unknown 25(OH)D concen-tration are evaluated using both the vitaAID system and anELISA method. As indicated previously, 25(OH)D levels below50 nM indicate vitamin D deficiency while level below 20 nMindicate a severe deficiency. Both methods are in agreementthat 2 of the samples are 25(OH)D deficient, while the otheris borderline deficient. Moreover, 2 of the ELISA measure-ments fall within the range of values measured using thevitaAID system and one is within 2 nM of that range.

Discussion

In this paper we presented the vitaAID system which allowsfor the fast and accurate quantification of 25(OH)D levels.This system uses a novel gold nanoparticle-based immuno-assay in order to produce a surface colorimetric reaction thatcan be quantified directly on a smartphone. We have demon-strated that we can measure physiological levels of 25(OH)Din solution with accuracy better than 15 nM and a precisionof 10 nM. Moreover we show that the results obtained usingthe vitaAID system are comparable with those of commercialELISA kits. By analyzing 3 serum samples with unknown25(OH)D concentration we were able to determine accuratelythe extent of vitamin D deficiency in each case. These results

1440 | Lab Chip, 2014, 14, 1437–1442

Fig. 4 a) Predicted concentration vs. actual concentration (grey line) for c(black) and ELISA kit (red) b) predicted concentrations of 25(OH)D in serum

represent a promising step in the development of an accuratesystem for determining vitamin D levels which has manypotential applications as a point-of-care tool both in thedeveloped and the developing world.

In the future, this method for vitamin D quantificationcan be further improved in terms of precision and accuracy.Here, we used a specific form of 25(OH)D for coating anddetection, namely 25(OH)D3 and anti-25(OH)D3. The mono-clonal anti-25(OH)D3 has 68% cross reactivity with 25(OH)D2

and 100% with 25(OH)D3. The use of 25(OH)D3 for the detec-tion zone coating allows for the capturing of all the unboundAuNP-anti-25(OH)D3 conjugates after the initial interactionwith the sample. Nonetheless, this difference in the antibodyaffinity inevitably causes our method to underestimate thetotal 25(OH)D level which is the sum of the 25(OH)D2 and25(OH)D3 levels. This represents the shortcoming of mostimmunoassays for vitamin D (e.g. ELISA kit; Alpco, radioim-munoassay; DiaSorin) as demonstrated through clinical com-parison studies with other detection standards such as HPLCand LC-MS.13,14,32

MethodsSample preparation

For a vitamin D deficiency test, once the sample has beenacquired, several steps are performed in solution prior to itsapplication onto the vitaAID test strip. First, the filteredserum sample is mixed 1 : 10 (v/v) with 0.78 g ml−1 acetoni-trile (Thermo Fisher Scientific Inc.) in order to liberate the25(OH)D molecules that are in proportion of 95–99% boundto vitamin D binding proteins (DBP).33,34 The sample is thenmixed with AuNP-anti-25(OH)D conjugate solution for 30 min.This ensures that all the 25(OH)D initially present in the bloodsample is bound to AuNP-anti-25(OH)D before being appliedonto the test strip.

AuNP-anti-25(OH)D3 conjugate preparation

The spherical AuNP (Nanopartz Inc., 30 nm) came pre-treatedwith N-hydroxysuccinimide ester terminal (NHS) groupswhich specifically reacted with the primary amines of mono-clonal anti-25(OH)D3 IgG (Raybiotech Inc.) to form the AuNP-

This journal is © The Royal Society of Chemistry 2014

alibration solutions at 15 nM, 40 nM and 110 nM using vitaAID systemsamples and comparison to the average ELISA result.

Lab on a Chip Paper

Publ

ishe

d on

07

Febr

uary

201

4. D

ownl

oade

d by

Cor

nell

Uni

vers

ity o

n 29

/04/

2014

20:

48:4

0.

View Article Online

antibody conjugates. The antibody was first purified usingthe Pierce Antibody Clean-up Kit (Thermo Fisher ScientificInc.) because 2% bovine serum albumin (BSA) stabilizers inanti-25(OH)D3 are known to interfere with the amine-reactiveconjugation.35 The antibody solution was placed into theMelon Gel-based purification support which binds non-antibody proteins while allowing the IgG antibody to flowthrough in a purified form during the one-minute centrifuga-tion at 6000 g. The successful removal of BSA was checked byperforming sodium dodecyl sulfate polyacrylamide gel electro-phoresis (SDS-PAGE). For conjugation, the AuNP were mixedwith the purified anti-25(OH)D3 at 0.1 mg ml−1 in 0.01 Mamine-free phosphate buffer saline (PBS) buffer at pH 7.4. Themixture was sonicated for 30 s to re-suspend AuNP into solu-tion, followed by vortexing for 30 min at room temperature.36

The centrifugation was performed at 15000 g for 10 min toremove the excess antibody in supernatant form and the finalconjugates were reconstituted in 0.01 M PBS with 0.1%Tween-20 at pH 7.4. The successful conjugation was confirmedthrough surface plasmon resonance changes using ultraviolet-visible spectroscopy. The conjugates were diluted to 10 μg ml−1

and stored at 4 °C until use.

Detection area preparation

The covalent immobilization of 25(OH)D was achieved byobtaining 25(OH)D3, 3′-aminopropyl ether (Toronto ResearchChemicals Inc.) and using its primary amines as linkers tothe test strip surface. Immobilization of the peptides to sur-face using maleic anhydride chemistry has been demon-strated previously by others.37,38 Here, the aminopropylated25(OH)D3 was immobilized on a flat Si substrate other thanon a typical well-plate which represents a compatibilityimprovement for use in our smartphone-based detection.Briefly, 4′′ fused Si wafers were cleaned in piranha solution,immersed in 20 mM APTES (Sigma-Aldrich Co. LLC) in iso-propanol for 2 h and annealed at 120 °C for 1 h. The APTEScoating acted as an activation layer for the binding of 1%PSMA (Sigma-Aldrich Co. LLC) dissolved in tetrahydrofuran,which was spin-coated at 3500 rpm for 30 s followed by cur-ing at 120 °C for 2 h. The treated Si wafer was cooled andimmersed in acetone for 10 min and subsequently diced into4 by 7 mm strips. Finally, the 25(OH)D immobilization wasachieved by incubating the PSMA-treated strips with 20 μg ml−1

aminopropylated 25(OH)D3 in the coating buffer (0.1 Mcarbonate/bicarbonate buffer at pH 9.4) for 1 h at 37 °C.The unreacted PSMA sites were treated by incubating theblocking buffer (0.01 M PBS with 1 mg ml−1 Casein and0.05% Kathon preservative at pH 7.4) for 30 min at roomtemperature, and cleaned with washing buffer (0.01 MPBS with 0.05% Tween-20 at pH 7.4). The incubation procedureswere performed in custom incubation chambers that housedthe test strips and prevented pre-mature drying of the treat-ment solutions. The modified Si surfaces after each surfacetreatment were characterized by FT-IR using a Vertex 80-vspectrometer (Bruker Optics) equipped with a 60° germanium

This journal is © The Royal Society of Chemistry 2014

attenuated total reflection (VeeMax Ge ATR) crystal. For eachspectrum, 256 scans at a spectral resolution of 4 cm−1 wereperformed using a liquid nitrogen detector. After the 6 h incu-bation of AuNP-antibody conjugates with the sample on thedetection area, the strip was rinsed three times with thewashing buffer to remove unbound conjugates and incubatedwith silver enhancement solution from the Silver Enhancer Kit(Sigma-Aldrich Co. LLC). After 20 min, the detection area wasrinsed with the washing buffer and air dried at room temperature.

ELISA kit procedures

The standard and serum samples were analyzed with 25-OHVitamin D ELISA kit (ALPCO Diagnostics) according to the pro-vided protocol. Briefly, the samples were incubated with thereleasing agent to separate 25(OH)D molecules from DBP at37 °C for 1 h. The samples and anti-25(OH)D were then addedonto the 25(OH)D coated microtiter plates. During an overnight(18–22 h) incubation, the 25(OH)D molecules in the sampleand those immobilized on the plates competitively bound tothe available anti-25(OH)D. The plates were washed to removeany unbound anti-25(OH)D and peroxidase-conjugated antibodywere added to form 25(OH)D–anti-25(OH)D–peroxidase antibodycomplex on the plate surface. After 1 h incubation, the plateswere washed and tetramethylbenzidine (TMB) substrates wereadded which resulted in a reaction causing the solution colorto change. After 20 min, the reaction was stopped with anacidic solution and the absorption of the solution was mea-sured using Spectramax 384 at 450 nm.

Acknowledgements

D.E. and S.M. acknowledge funding support through theAcademic Venture Fund of the Atkinson Center for a Sus-tainable Future at Cornell University. In addition, V.O.acknowledges the support of the National Science and Engi-neering Research Council of Canada (NSERC) through apostgraduate scholarship. M.M. would like to acknowledge aNational Science Foundation Graduate Research Fellowshipunder grant no. DGE-0707428. The fabrication stepsdescribed in this paper were carried out at the CornellNanoscale Facility (CNF) and the Nanobiotechnology Centerat Cornell University (NBTC).

References

1 M. F. Holick, N. Engl. J. Med., 2007, 357, 266–281.

2 M. F. Holick, Am. J. Clin. Nutr., 2004, 79, 362–371. 3 T. J. Wang, M. J. Pencina, S. L. Booth, P. F. Jacques,

E. Ingelsson, K. Lanier, E. J. Benjamin, R. B. D'Agostino,M. Wolf and R. S. Vasan, Circulation, 2008, 117, 503–511.

4 J. L. Anderson, H. T. May, B. D. Horne, T. L. Bair, N. L. Hall,

J. F. Carlquist, D. L. Lappe and J. B. Muhlestein, Am. J.Cardiol., 2010, 106, 963–968.

Lab Chip, 2014, 14, 1437–1442 | 1441

Lab on a ChipPaper

Publ

ishe

d on

07

Febr

uary

201

4. D

ownl

oade

d by

Cor

nell

Uni

vers

ity o

n 29

/04/

2014

20:

48:4

0.

View Article Online

5 S. Mehta, E. Giovannucci, F. M. Mugusi, D. Spiegelman,

S. Aboud, E. Hertzmark, G. I. Msamanga, D. Hunter andW. W. Fawzi, PLoS One, 2010, 5, e8770.

6 S. Mehta, D. J. Hunter, F. M. Mugusi, D. Spiegelman,

K. P. Manji, E. L. Giovannucci, E. Hertzmark, G. I. Msamangaand W. W. Fawzi, J. Infect. Dis., 2009, 200, 1022–1030.

7 C. F. Garland, F. C. Garland, E. D. Gorham, M. Lipkin,

H. Newmark, S. B. Mohr and M. F. Holick, Am. J. PublicHealth, 2006, 96, 252–261.

8 J. M. Lappe, D. Travers-Gustafson, K. M. Davies, R. R. Recker

and R. P. Heaney, Am. J. Clin. Nutr., 2007, 85, 1586–1591.

9 M. XunWu, Chin. J. Pract. Intern. Med., 2009, 29, 965–967.

10 R. Chesney, Rev. Endocr. Metab. Disord., 2001, 2,

145–151.11 P. Lips, J. Steroid Biochem. Mol. Biol., 2010, 121, 297–300.

12 A. Mithal, D. A. Wahl, J. P. Bonjour, P. Burckhardt,

B. Dawson-Hughes, J. A. Eisman, G. El-Hajj Fuleihan,R. G. Josse, P. Lips and J. Morales-Torres, Osteoporosis Int.,2009, 20, 1807–1820.

13 H. J. Roth, H. Schmidt-Gayk, H. Weber and C. Niederau,

Ann. Clin. Biochem., 2008, 45, 153–159.

14 A. M. Wallace, S. Gibson, A. de la Hunty, C. Lamberg-Allardt

and M. Ashwell, Steroids, 2010, 75, 477–488.

15 B. W. Hollis and R. L. Horst, J. Steroid Biochem. Mol. Biol.,

2007, 103, 473–476.

16 R. Gupta and S. Gupta, Lateral flow immunoassay for

detecting vitamins, Report WO/2012/129650, 2012.

17 E. Fu, K. E. Nelson, S. A. Ramsey, J. O. Foley, K. Helton and

P. Yager, Anal. Chem., 2009, 81, 3407–3413.

18 P. Pradelles, J. Grassi, C. Creminon, B. Boutten and

S. Mamas, Anal. Chem., 1994, 66, 16–22.

19 K. Lei and Y. C. Butt, Microfluid. Nanofluid., 2010, 8,

131–137.

20 C.-H. Yeh, C.-Y. Hung, T. Chang, H.-P. Lin and Y.-C. Lin,

Microfluid. Nanofluid., 2009, 6, 85–91.

1442 | Lab Chip, 2014, 14, 1437–1442

21 Q. Wei, R. Nagi, K. Sadeghi, S. Feng, E. Yan, S. J. Ki,

R. Caire, D. Tseng and A. Ozcan, ACS Nano, 2014, DOI:10.1021/nn406571t.

22 A. F. Coskun, R. Nagi, K. Sadeghi, S. Phillips and A. Ozcan,

Lab Chip, 2013, 13, 4231–4238.

23 A. F. Coskun, J. Wong, D. Khodadadi, R. Nagi, A. Tey and

A. Ozcan, Lab Chip, 2013, 13, 636–640.

24 V. Oncescu, M. Mancuso and D. Erickson, Lab Chip,

2014, 14, 759–763.

25 V. Oncescu, D. O'Dell and D. Erickson, Lab Chip, 2013, 13,

3232–3238.

26 J. H. Wang, L. P. Zhu, B. K. Zhu and Y. Y. Xu, J. Colloid

Interface Sci., 2011, 363, 676–681.

27 G. C. Chitanu, I. Popescu and A. Carpov, Rev. Roum. Chim.,

2006, 51, 923–929.

28 J.-H. Jeong, Y.-S. Byoun and Y.-S. Lee, React. Funct. Polym.,

2002, 50, 257–263.

29 R.-H. Lin and J.-H. Hsu, Polym. Int., 2001, 50, 1073–1081.

30 J. Kai, A. Puntambekar, N. Santiago, S. H. Lee, D. W. Sehy,

V. Moore, J. Han and C. H. Ahn, Lab Chip, 2012, 12, 4257–4262.31 J. Bigun, in Vision with Direction: A Systematic Introduction to

Image Processing and Computer Vision, Springer, Germany,2006, ch. 2, pp. 31–32.

32 C. S. Højskov, L. Heickendorff and H. J. Møller, Clin. Chim.

Acta, 2010, 411, 114–116.

33 A. M. Wootton, Clin. Biochem. Rev., 2005, 26, 33.

34 G. D. Carter, Curr. Drug Targets, 2011, 12, 19–28. 35 A. C. Grodzki and E. Berenstein, in Immunocytochemical

Methods and Protocols, Springer, 2010, pp. 15–26.36 J. Ljungblad, MPhil thesis, Department of Physics, Chemistry

and Biology, Linköping University, 2009.37 M. Tasso, A. L. Cordeiro, K. Salchert and C. Werner,

Macromol. Biosci., 2009, 9, 922–929.38 A. L. Cordeiro, T. Pompe, K. Salchert and C. Werner, in

Bioconjugation Protocols, Springer, 2011, pp. 465–476.

This journal is © The Royal Society of Chemistry 2014