Biomedical Microdevices

, Volume 12, Issue 5, pp 887–896 | Cite as

An integrated microfluidic chip for non-immunological determination of urinary albumin

  • Chun-Che Lin
  • Chin-Chung Tseng
  • Chao-June Huang
  • Jung-Hao Wang
  • Gwo-Bin Lee


This study presents an integrated microfluidic chip for non-immunologically determining the concentrations of albumin in clinical urine samples. This microchip integrates membrane-type micromixers and a fan-shaped micropump capable of simultaneously and precisely delivering assay reagents to react with 6 urine samples in one single operation. The experimental results show that the coefficient of variation in the pumping rate is 2.42%. More importantly, using this unique chip design, only 2 electromagnetic valves are required for the actuation of the micromixer and the micropump. The working range of the proposed microchip is 2–200 mg/L of albumin, which covers the range of interest for the determination of microalbuminuria. Moreover, statistical analysis show that the results obtained by the proposed microchip are in good agreement with the conventional detection method, based on immunological assays. This simple, inexpensive and microchip-based platform presents a promising alternative to conventional immunological assays for measurement of urinary albumin, and is well suited for clinical applications.


Microalbuminuria Immunoassay Microfluidics Micropump 



albumin blue


albumin excretion rate


biological field-effect transistor


coefficients of variation


computer-numerically controlled


confidence interval


dextran sulfate


electromagnetic valves


enzyme-linked immunosorbent assay


human serum albumin


limit of detection




metal oxide semiconductor field effect transistor


molecularly imprinted polymer


sodium hydroxide


normally closed valves








quartz crystal microbalance


self-assembled monolayer


standard deviation

1 Introduction

Human serum albumin (HSA) which is composed of one long polypeptide chain (66.5 kDa) is a major constituent of plasma proteins (Dockal et al. 2000). A kidney is capable of retaining beneficial proteins in blood such as albumin and filtering out waste components. However, when the kidneys are severely damaged, albumin may appear in the urine. The quantitative determination of urinary albumin is of great importance because it can provide critical information on different aspects of renal function and many diseases have been linked to an increased albumin excretion rate (AER) (Waller et al. 1989; Chase et al. 1991). A slightly increased AER is termed as microalbuminuria (MAU) and is recognized as an early predictor for nephropathy in patients who suffer from diabetes and hypertension (Mogensen 1987; Vigstrup and Mogensen 1985). In addition, a number of studies suggest that MAU is also an indicator of cardiovascular diseases in nondiabetic individuals (Lydakis and Lip 1998; Haffner et al. 1990). Since diabetic nephropathy is potentially reversible at the early stage (Viberti et al. 1979), early detection and control of MAU can reduce the risk of nephropathy and renal failure. Thus, routine monitoring of MAU is essential for patients at high-risk for diabetes. For the monitoring of MAU, the range of interest of urinary albumin concentration is between 2–200 mg/L, particularly the 15–40 mg/L range that covers the usual cutoff limits between normal and increased albumin excretion (Rowe et al. 1990).

Since a urine sample is very complex and the urinary albumin concentration is relatively low for MAU determination, several immunoassay-based methods such as a solid-phase fluorescent immunoassay (Chavers et al. 1984), an enzyme-linked immunosorbent assay (ELISA) (Aybay and Karakus 2003), piezoelectric immunosensors (Navrátilová et al. 2001), a liposomal immunoassay (Frost et al. 1996), magnetic immunoassay (Lu et al. 2006), and a molecularly imprinted polymer (MIP)-quartz crystal microbalance (QCM) sensor (Hu et al. 2005) have been developed to determine urinary albumin due to these specificity and sensitivity issues. Although these immunoassay-based methods are reliable for urinary albumin determination, the costs, mainly attributed to antibodies required for immunoreactions, are high and may limit their applications in local laboratories for routine monitoring of MAU.

Recently, microfluidic systems have attracted considerable interest and shown their great potential for chemical analysis and biomedical applications due to unique advantages including fast analysis time, high throughput, low sample/reagent consumption, disposability, portability, reliability, and their capability for integration and automation (Manz et al. 1990; Liao et al. 2005). Recently, several microfluidic devices have been developed to determine urinary albumin concentrations, such as a microchip electrophoresis (Chan and Herold 2006) and a biological field-effect transistor (BioFET) (Park et al. 2008). The assay based on microchip electrophoresis used a commercial assay kit for quantifying urinary albumin by adding chicken albumin as an internal calibrator. There, sodium dodecyl sulfate in the buffer binds to the albumin. Fluorescent dye then binds to the sodium dodecyl sulfate micelles as an indicator of albumin concentration. Although this method quantifies urinary albumin with good sensitivity, precision, and accuracy, an electrophoresis separation step is required. The BioFET is a metal-oxide-semiconductor field effect transistor (MOSFET)-type protein sensor with gold deposited on the gate insulator. The gold on the gate surface was first chemically modified by a self-assembled monolayer (SAM), followed by immobilizing anti-albumin antibody to conjugate with albumin in the samples. The drain current was then modulated by the albumin bound to the anti-albumin and the concentration of urinary albumin was estimated according to the current variation ratio. Although the BioFET provides advantages such as small size, fast response, high reliability, and it is potentially portable, it is an antibody-based sensor and hence its application in routine monitoring of urinary albumin is still limited due to the costs attributed to the antibody reagents.

In order to simplify the procedures and reduce the costs for urinary albumin determination, a non-immunological method based on a dye-binding assay has been developed. The method is based on the characteristics that fluorescence probes, namely albumin blue (AB) 633 and AB 670, are highly selective for HSA and the fluorescence intensity is enhanced by orders of magnitude after binding with albumin (Kessler et al. 1992; Kessler and Wolfbeis 1992). These two dyes, however, are not stable and hence their practical applications are limited. Later a derivative, AB 580, was developed to overcome the stability issue (Kessler et al. 1997a). These dyes are capable of highly specific binding with albumin to undergo strong fluorescence enhancement and are not subject to interference by other urinary proteins with minimum sample pretreatment (Kessler et al. 1997b). The AB 580-based assay was also applied to a glass microchip equipped with a fluorescence microscope for detection of albumin. However, the limit of detection (LOD) was not determined and further clinical applications were not performed (Kamholz et al. 1999). Recently, this AB 580-based assay was performed on a disposable microchip with an integrated fluorescence detection system based on thin-film, organic, light emitting diodes. A linear range of 10–100 mg/L was obtained (Hofmann et al. 2005). Although the LOD obtained by this integrated microchip is sufficiently low enough for the determination of MAU, only one sample can be assayed in one single operation in this work. Moreover, clinical samples were not assayed.

In this study, we present a microchip integrated with a fan-shaped micropump and micromixers capable of determining the albumin concentrations of six clinical urine samples in one single operation using an AB 580-based assay. Due to the unique design of the microchip, only two electromagnetic valves (EMVs) are required for actuating the on-chip micropump and micromixers. We also compare the results from clinical urine samples processed by the microchip with those obtained by the clinically used immunoturbidimetric method (Thakkar et al. 1997), to evaluate the performance of the proposed microchip.

2 Materials and methods

2.1 Chemicals and reagents

Dextran sulfate (DS, MWav 5000), hexadimethrine bromide (Polybrene, PB), sodium hydroxide (NaOH) and ethanol were obtained from Sigma-Aldrich (Louis, USA). The poly(dimethylsiloxane) (PDMS) kit was purchased from Dow Corning Corp. (Midland, USA). For the casting process, the PDMS elastomer and the curing agent were mixed in a weight ratio of 10:1 and cured at 80 °C for 4 hours. The albumin fluorescence assay kit was purchased from Fluka (Buchs, Switzerland). The assay reagent was freshly prepared by mixing reagent A (solution of AB 580 in 2-propanol) with reagent B (buffer solution, pH 7.0 ± 0.2) in both a volume ratio of 1:50 or 1:100. The albumin stock solution was prepared by dissolving 10 mg of HSA in 5 mL of Milli-Q water, and the albumin standard solutions for establishing calibration curves were prepared from this stock solution using the diluted solution provided in the kit. For fluorescence detection, the samples or standard solutions (6 μL) were mixed and reacted with the assay reagent (30 μL) through the operation of the micropump and micromixers.

2.2 Chip design

Figure 1 shows a schematic illustration of the chip design, the detailed dimensions, the actuating mechanism of the micropump and micromixers, and a photograph of the prototype microchip. The prototype microchip is composed of four layers, in which the glass layer functions as a support layer and one side of the channel, the resilient membrane layer forms a liquid channel with the glass layer, the air chamber layer forms cavities with the liquid channel layer when compressed air is applied from the air inlet deflects the membranes underneath, and the reservoirs are the wells where the reactions occur.
Fig. 1

(a) Exploded view, (b) photograph, and (c) dimensions of the microchip. (d) Schematic illustration of the pumping (d-1) and mixing (d-2) mechanisms. The dashed line in the upper part of (c) represents the air chamber layer. The dashed arrows in (d-1) and (d-2) indicate that the compressed air is supplied to the air chambers through the connecting air channels

The fan-shaped micropump consists of six liquid channels, six air chambers, resilient PDMS membrane structures and normally closed valves (NCVs). As shown in Fig. 1(c), the NCV of the micropump is a PDMS-based floating block structure located inside the liquid channel, and is activated by hydraulic pressure generated by the motion of the PDMS membrane (Yang et al. 2009a). The NCV is used to prevent the backflow of the liquid during the pumping (Fig. 1(d-1)) and mixing (Fig. 1(d-2)) steps.

The microchip integrates micromixers which utilize pneumatically-driven membranes to generate the swirling flow in mixing chambers. As shown in Figure (d-2), the micromixer consists of two layers of PDMS substrates and a glass substrate layer. The top PDMS substrate contains two air chambers with connecting air channels. The width and height of the connecting air channels are 1,000 μm and 300 μm, respectively. The bottom PDMS substrate is comprised of a mixing chamber with two liquid channels. When compressed air is supplied to the two air chambers through the connecting air channels, the PDMS membranes are deflected sequentially and generate the swirling flow inside the circular mixing chamber. Meanwhile, the PDMS membrane above the NCV is also deflected to prevent the NCV from opening during the mixing step. Note that the connecting air channels are used to connect all the six micromixers so that only one EMV is required to activate the six micromixers in one single operation.

Prior to fluorescence detection, 6 μL of standard solutions or urine samples are first pipetted into each of the reservoirs and 190 μL of freshly prepared assay reagent is pipetted into the reagent well. The assay reagent is then pumped into the reservoirs (30 μL for each reservoir) through the fan-shaped micropump, followed by mixing with the standard solutions or samples in the reservoirs by the micromixers. After mixing, the compressed air is continuously applied to maintain the liquid level until the fluorescence detection is completed.

2.3 Fabrication

The integrated microfluidic chip is comprised of a three-layer PDMS structure and a glass substrate. Figure 2 shows a schematic illustration of the microfabrication process which includes a computer-numerically controlled (CNC) machining process, a PDMS replication process, and a bonding process (Yang et al. 2009b). Briefly, polymethylmethacrylate (PMMA) master molds with microstructures are first formed by using a CNC machine (EGX-400, Roland Inc., Japan) equipped with a 0.5-mm end mill (Fig. 2(a) and (a’)). The rotational speed and feed rate of the spindle are 26,000 rpm and 15 mm/min, respectively. This is then followed by the PDMS casting process (Fig. 2(b) and (b’)) to form the inverse structures of the air chamber mold and the channel mold. Note that the channel mold is designed to form a thin film with a thickness of 100 μm for deflections. After removing the cover plate on the channel structure (Fig. 2(c)) and peeling off the air chamber structure from the PMMA mold (Fig. 2(c’)), an oxygen plasma treatment is used to form a double-layer PDMS structure (Fig. 2(d)). Again, this double-layer PDMS structure is peeled off from the PMMA mold (Fig. 2(e)), followed by a drilling step to form wells for loading reagents and samples (Fig. 2(f)). Then, this double-layer PDMS structure with wells is bonded with a piece of glass substrate by an oxygen plasma treatment (Fig. 2(g)). Prior to the oxygen plasma treatment, the block structure of the NCV is masked using a black marker to prevent the block from bonding with the glass substrate underneath. This process is extremely important to form a floating block. Finally, the six pieces of drilled PDMS are bonded on top of the 6 sample wells to form the reservoirs for reaction and detection (Fig. 2(h)), followed by rinsing the channels with ethanol to remove the black marker used to mask the NCV area during the bonding process.
Fig. 2

An overview of the fabrication process, which includes the CNC machining processes (a and a’), PDMS replication processes (b and b’), peeling-off process (c and e), cover plate removing process (c), bonding processes (d, g, and h), and drilling process (f)

2.4 Experimental setup

The experimental setup, consisting of an EMV controller and a fluorescent detection system, is schematically shown in Fig. 3. A custom-built EMV controller has been developed to regulate the deflections of the PDMS membranes for liquid delivery and mixing. This EMV controller integrates a pressure regulator, a microcontroller (8051 microcontroller, model AT89C51 24 PC, ATMEL, USA), 12 EMVs (SD70M-6BG-32, SMC, Japan), an air compressor (MDR2-1A/11, Jun-Air Inc., Japan), and a graphical user interface developed using Visual Basic software (Visual Basic 2005, Microsoft, USA). The pressure, frequency, and duration for driving the EMVs are controlled via this custom-built controller. Note that only two EMVs are used in this study for activating the micropumps and the micromixers.
Fig. 3

Experimental setup for the study which includes an optical detection module and an EMV controlling module. The EMV controlling module is consisted of a custom-built EMV controller which integrates a pressure regulator, a microcontroller, an air compressor, and a graphical user interface developed using Visual Basic software. The optical detection module is composed of a light source (mercury lamp), optical filters, an objective lens, a pinhole, a photo-multiplier tube, a commercial data acquisition interface (ADC), and a personal computer

Signals are detected on-chip via fluorescence detection. The detection system is constructed by modifying a commercial reflection microscope (model BX41, Olympus, Tokyo Japan) (Kuo et al. 2009). The light source radiating from a mercury lamp is filtered by a band-pass optical filter (540–580 nm) and is focused into a spot with a diameter of 3 mm on the reaction reservoir by a 50× (numerical aperture = 0.5) long-working-distance objective lens for exciting the AB 580 dyes. The emitted fluorescence is then collected by the same objective lens and is passed through a dichroic beam splitter (595 nm), a band-pass optical filter (600–660 nm), a pinhole with a diameter of 1.0 mm, and finally is detected by a photo-multiplier tube operating at 600 V (C3830, R928, Hamamatsu Photonics, Tokyo, Japan). Amplified photo-electronic signals are converted into digital signals and processed by a computer using a 24-bit commercial data acquisition interface (Model 9924–2; Scientific Information Service, Taipei, Taiwan).

3 Results and discussion

3.1 Surface modification

PDMS is composed of repeating -OSi(CH3)2- units, resulting in an inherently hydrophobic surface and nonspecific adsorption of proteins (McDonald et al. 2000). In this study, the hydrophobic property may hinder the transport of the aqueous solutions (e.g. the assay reagents) and many bubbles may form. More importantly, a higher detection limit is adversely obtained due to the adsorption of proteins. In order to improve the hydrophilicity for transport of reagents and to reduce the adsorption of the negatively charged albumin (Bessonova et al. 2007) onto the walls of the PDMS channels, a multi-layer modification method is used to create a negatively charged surface on the walls of the PDMS channels (Liu and Pietrzyk 2000). Briefly, the PDMS channels are pre-conditioned by first sequentially rinsed with a solution of 0.1 M NaOH and deionized water for 5 min each. Then, the channels are rinsed with a cationic polymer solution, 5% PB, for 2 min and the solution is left for 30 min before rinsing with deionized water. The same process is also applied for a second layer coating of an anionic polymer solution, 3% DS. Figure 4 shows the results of the surface modification. A mixed solution of HSA (200 mg/L) and the assay reagent is first loaded into the channels (Fig. 4(a)). After removing the mixed solution from the channels, rinsing the channels with deionzed water 3 times, and finally removing the deionized water, the photographs of the channels before and after the multi-layer modification are shown in Fig. 4(b) and (c), respectively. It can be clearly observed that protein adsorption has been greatly reduced. Figure 4(d) shows the profiles of the gray-scale intensity that represents the fluorescence intensity across the channels. The measured gray-scale intensities for the channels before and after modification are 85 and 31, respectively. The results indicate that after creating a negatively charged surface on the PDMS channels, the adsorption of albumin is reduced about 63%. The bad effect of the albumin adsorption is the poor detection sensitivity since albumin may adsorb onto the surface of the mixers, resulting in less albumin for detection. Although the adsorption effect is not completely eliminated, later results show that a LOD as low as 2 mg/L can be obtained, which is low enough for clinical applications. Furthermore, the modified channels are more wettable than the original ones due to the negatively charged surfaces, and fewer bubbles are formed during liquid transport.
Fig. 4

Results of the surface modification for (a) the channel filled with HSA solution, (b) the channel without surface modification, and (c) the channel modified with a negatively charged polymer solution. (d) The intensity profile obtained by using an image analysis software in gray scale across the channels shown in (a–c)

3.2 Characterization of the micropump

One of the important performance parameters for the microfluidic chip is the uniformity of the pumping rates for the 6 channels generated by the fan-shaped micropump. Our group previously developed “spider-web” micropumps to successfully perform the simultaneous and precise transport of reagents in an automatic manner (Wang and Lee 2005). However, 3 EMVs were used for the actuation of the “spider-web” micropumps. In this study, the fan-shaped micropump is actuated by only using a single EMV due to its unique design. The measurements of the pumping rates are performed by measuring the volume of the pumped deionized water in the reservoirs after actuating the micropump for 1 min, and the measurements are repeated for 3 times to obtain the average pumping rates. Figure 5 shows the relationship between the measured average pumping rates for the six channels and the driving frequencies at different pressure levels. It is clearly observed that the pumping rates increase with the increased deflections of the PDMS membranes attributed to increased pneumatic pressures. It is also seen that the pumping rates increase with the increased driving frequency. However, the maximum pumping rate at a constant pressure is limited by the restoration of the PDMS membranes. If the driving frequency is too high, the PDMS membranes cannot be restored to their original position and the pumping rate cannot increase but starts to decrease. The average pumping rate for the fan-shaped micropump can be as high as 59.7 μL/min when actuated at a pressure of 15 psi, with a driving frequency of 20 Hz. The error bars shown in Fig. 5 are obtained from measuring the pumping rates for each of the 6 channels. It is observed that the coefficients of variation (CVs) of the pumping rate increase with the increased pumping rate. The CVs of the pumping rates range from 1.41% to 8.19%, respectively. In order to obtain a uniform pumping rate for the 6 channels, in the subsequent experiments, the fan-shaped micropump is activated with a driving frequency of 10 Hz at a pressure of 10 psi, corresponding to a pumping rate of 21.7 μL/min and the CV of the pumping rate for the 6 channels is 2.42%.
Fig. 5

The relationship between the pumping rate and the driving frequency of the EMV at different pressure levels. The error bars are obtained from measuring pumping rates for the six channels

3.3 Characterization of the micromixer

The mixing performance of the proposed micromixer is estimated according to our previous work (Lin et al. 2005) and is shown in Fig. 6. Figure 6 shows the average result that is measured across the centers of the six mixing chambers (x’–x line), in which X+ and C+ are the normalized location and concentration, respectively. The mixing time, the applied air pressure and the driving frequency of the EMV are 5 sec, 10 psi and 5 Hz, respectively. The result shows that the intensity profile for the normalized concentration approaches 0.5 (completely mixed) throughout the mixing chamber, and the mixing indices before and after mixing are calculated to be 20.1% and 96.6%, respectively. Moreover, the error bars are obtained from the six mixers, and the CVs of the normalized concentrations after mixing range from 2.42% to 3.94%, indicating that the mixing performance is uniform for the six mixers under the operating conditions.
Fig. 6

The profile of the concentration distribution before and after the mixing process. The error bars are obtained from the six mixers

3.4 Calibration curves

Figure 7 shows the calibration curves obtained from the prototype microchip under different assay reagent compositions. Figure 7(a) is obtained when the assay reagent is prepared by mixing reagent A with reagent B in a volume ratio of 1:100 where Fig. 7(b) is obtained when the ratio is 1:50. As shown in Fig. 7(a) (reagent A/reagent B = 1/100), the working range is 2–80 mg/L with a linearity (R2) of 0.9898. It is noted that the calibration curve is nonlinear when albumin concentrations over 80 mg/L are included and approaches saturation at concentrations over 200 mg/L. This is probably because the binding of the AB dyes to albumin obeys the Law of Mass Action and results in saturation in the fluorescence intensity (Kessler and Wolfbeis 1992). In order to prevent this saturation and to extend the working range to more than 200 mg/L, an assay reagent in which the ratio of reagent A to reagent B is 1:50 is tested. As shown in Fig. 7(b), a working range of 5–200 mg/L with a linearity of 0.9914 is obtained. After establishing the calibration curves, the concentrations of urinary albumin of the twenty-six clinical urine samples submitted for routine analysis are assayed based on the calibration curves shown in Fig. 7.
Fig. 7

Calibration curves determined when (a) reagent A: reagent B ratio is 1:100 and (b) reagent A: reagent B ratio is 1:50. The error bars of each concentration are obtained from 3 repeated measurements

3.5 Clinical urine sample test

In order to evaluate the performance of the proposed microchip in measuring urinary albumin concentration, 26 centrifuged urine samples (2,500 rpm, 10 min) collected from National Cheng Kung University Hospital are assayed by both the prototype microchip and the conventional immunoturbidity method. The albumin concentrations of the 26 urine samples are determined by averaging results of three repeated measurements with CVs ranging from 0.3% to 6.2%. Although the CVs obtained by the proposed microchip are not superior to those claimed by Sigma (0.6%–3.6%), they are, however, slightly better than those provided by Beckman (5.0%–8.0%). The results indicate that the reproducibility performance of the proposed microchip is acceptable. Moreover, comparisons of the results (average concentrations) determined by the prototype microchip with the conventional method are made based on the Bland-Altman plot (Bland and Altman 1999) and Passing-Bablok regression analysis (Passing and Bablok 1983). Figure 8(a) shows the result of the Bland-Altman plot, indicating that there is no apparent bias for the concentrations of urinary albumin determined by the prototype microchip since the differences between the two methods are all within the mean ±1.96 standard deviation (SD) interval. Figure 8(b) shows the result of Passing-Bablok regression analysis in which the equation for the regression line is \( y = 0.9908{\hbox{x}} + 0.3128 \). The 95% confidence interval (CI) shows that the slope and the intercept are not significantly different from 1 and 0, respectively, indicating that the concentrations of urinary albumin determined by the proposed microchip is in good agreement with that determined by the conventional method. In addition, a comparison of throughput, detection limit, working range, and sample consumption between these two methods is listed in Table 1. The throughput, working range, and sample consumption of the proposed microchip are superior to the conventional method. It is then concluded that the prototype microchip can automatically perform the detection of urinary albumin with a high confidence in the accuracy of the results.
Fig. 8

Comparison of results obtained by the prototype microchip and the conventional method. (a) Bland-Altman bias plot and (b) Passing-Bablok regression analysis

Table 1

A comparison of throughput, detection limit, working range, and sample consumption between the prototype microchip and a conventional assay



Conventional assay a


190 samples/hour

180 samples/hour

Detection limit

2 mg/L

2 mg/L

Working range

2–200 mg/L

2–40 mg/L b

Sample consumption

5 μL

21 μL

aThe conventional assay was performed on IMMAGE Beckman Rate Nephelemetry (Beckman Coulter Inc. CA, USA)

bAlthough the working range of the conventional assay performed on Beckman’s instrument can be expanded to 2–8,640 mg/L according to the data sheet, some additional efforts are required. Further, the concentrations higher than 200 mg/L are out of the range interested for monitoring MAU. Using the proposed microchip, however, no further effort is required for monitoring MAU since the obtained working range fulfils the demand

4 Conclusions

This work presents a unique microfluidic-based platform capable of simultaneously determining low levels of albumin in six urine samples, in one single operation, based on a dye-binding reagent. This platform is less labor intensive, consumes less samples and reagents, and provides a fast analysis (114 sec for each operation). The experimental results are consistent with conventional immunoassays. Moreover, due to the low reagent price (no immunological reagents are required) and the excellent performance of the microchip, this platform may be used as a substitute for clinical immunoassays for the determination of MAU.



The authors gratefully acknowledge the financial support provided to this study by the National Science Council in Taiwan (NSC 97–2120-M-006–007).


  1. C. Aybay, R. Karakus, Turk. J. Med. Sci. 33(1), 1 (2003)Google Scholar
  2. E.A. Bessonova, L.A. Kartsova, A.U. Shmukov, J. Chromatogr. A 1150(1–2), 332 (2007)CrossRefGoogle Scholar
  3. J.M. Bland, D.G. Altman, Stat. Meth. Med. Res. 8(2), 135 (1999)CrossRefGoogle Scholar
  4. O.T.M. Chan, D.A. Herold, Clin. Chem. 52(11), 2141 (2006)CrossRefGoogle Scholar
  5. H. Chase, G. Marshall, S. Garg, S. Harris, I. Osberg, Clin. Chem. 37(12), 2048 (1991)Google Scholar
  6. B.M. Chavers, J. Simonson, A.F. Michael, Kidney Int. 25(3), 576 (1984)CrossRefGoogle Scholar
  7. M. Dockal, D.C. Carter, F. Rüker, J. Biol. Chem. 275(5), 3042 (2000)CrossRefGoogle Scholar
  8. S.J. Frost, J. Chakraborty, G.B. Firth, J. Immunol. Methods 194(2), 105 (1996)CrossRefGoogle Scholar
  9. S. Haffner, M. Stern, M. Gruber, H. Hazuda, B. Mitchell, J. Patterson, Arterioscl. Throm. Vas. 10(5), 727 (1990)Google Scholar
  10. O. Hofmann, X. Wang, J.C. deMello, D.D.C. Bradley, A.J. deMello, Lab Chip 5(8), 863 (2005)CrossRefGoogle Scholar
  11. C.H. Hu, T.C. Chou, T.Y. Lin, Int. Conf. on Bio-Nano-Information (BNI) Fusion (CA, USA, July 20-22, 2005), pp. 20-22.Google Scholar
  12. A.E. Kamholz, B.H. Weigl, B.A. Finlayson, P. Yager, Anal. Chem. 71(23), 5340 (1999)CrossRefGoogle Scholar
  13. M.A. Kessler, O.S. Wolfbeis, Anal. Biochem. 200(2), 254 (1992)CrossRefGoogle Scholar
  14. M.A. Kessler, M. Hubmann, B. Dremel, O. Wolfbeis, Clin. Chem. 38(10), 2089 (1992)Google Scholar
  15. M.A. Kessler, A. Meinitzer, W. Petek, O.S. Wolfbeis, Clin. Chem. 43(6), 996 (1997a)Google Scholar
  16. M.A. Kessler, A. Meinitzer, O.S. Wolfbeis, Anal. Biochem. 248(1), 180 (1997b)CrossRefGoogle Scholar
  17. C.H. Kuo, J.H. Wang, G.B. Lee, Electrophoresis 30(18), 3228 (2009)CrossRefGoogle Scholar
  18. C.S. Liao, G.B. Lee, J.J. Wu, C.C. Chang, T.M. Hsieh, F.C. Huang, C.H. Luo, Biosens. Bioelectron. 20(7), 1341 (2005)CrossRefGoogle Scholar
  19. C.H. Lin, C.H. Tsai, L.M. Fu, J. Micromechanics Microengineering 15(5), 935 (2005)CrossRefGoogle Scholar
  20. Y. Liu, D.J. Pietrzyk, Anal. Chem. 72(24), 5930 (2000)CrossRefGoogle Scholar
  21. M. Lu, F. Ibraimi, D. Kriz, K. Kriz, Biosens. Bioelectron. 21(12), 2248 (2006)CrossRefGoogle Scholar
  22. C. Lydakis, G. Lip, Q. J. Med. 91(6), 381 (1998)Google Scholar
  23. A. Manz, N. Graber, H.M. Widmer, Sens. Actuators, B 1(1–6), 244 (1990)CrossRefGoogle Scholar
  24. J.C. McDonald, D.C. Duffy, J.R. Anderson, D.T. Chiu, H. Wu, O.J.A. Schueller, G.M. Whitesides, Electrophoresis 21(1), 27 (2000)CrossRefGoogle Scholar
  25. C.E. Mogensen, Kidney Int. 31(2), 673 (1987)CrossRefGoogle Scholar
  26. I. Navrátilová, P. Skládal, V. Viklický, Talanta 55(4), 831 (2001)CrossRefGoogle Scholar
  27. K.M. Park, S.K. Lee, Y.S. Sohn, S.Y. Choi, Electron. Lett. 44(3), 185 (2008)CrossRefGoogle Scholar
  28. H. Passing, W. Bablok, Clin. Chem. Lab. Med. 21(11), 709 (1983)CrossRefGoogle Scholar
  29. D.J.F. Rowe, A. Dawnay, G.F. Watts, Ann. Clin. Biochem. 27(4), 297 (1990)Google Scholar
  30. H. Thakkar, D.J. Newman, P. Holownia, C.L. Davey, C.C. Wang, J. Lloyd, A.R. Craig, C.P. Price, Clin. Chem. 43(1), 109 (1997)Google Scholar
  31. G. Viberti, J. Pickup, R. Jarrett, H. Keen, N. Engl. J. Med. 300(12), 638 (1979)CrossRefGoogle Scholar
  32. J. Vigstrup, C.E. Mogensen, Acta Ophthalmol. 63(5), 530 (1985)CrossRefGoogle Scholar
  33. K. Waller, K. Ward, J. Mahan, D. Wismatt, Clin. Chem. 35(5), 755 (1989)Google Scholar
  34. C.H. Wang, G.B. Lee, Biosens. Bioelectron. 21(3), 419 (2005)MATHCrossRefMathSciNetGoogle Scholar
  35. Y.N. Yang, S.K. Hsiung, G.B. Lee, Microfluid. Nanofluid. 6(6), 823 (2009a)CrossRefGoogle Scholar
  36. S.Y. Yang, J.L. Lin, G.B. Lee, J. Micromechanics Microengineering 19(3), 035020 (2009b)CrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media, LLC 2010

Authors and Affiliations

  • Chun-Che Lin
    • 1
  • Chin-Chung Tseng
    • 2
  • Chao-June Huang
    • 1
  • Jung-Hao Wang
    • 1
  • Gwo-Bin Lee
    • 1
  1. 1.Department of Engineering ScienceNational Cheng Kung UniversityTainanTaiwan
  2. 2.Division of Nephrology, Department of Internal Medicine, College of MedicineNational Cheng Kung UniversityTainanTaiwan

Personalised recommendations