1 Introduction

Influenza is a serious respiratory infectious disease caused by influenza virus (Hoffmann et al. 2011). To prevent its wide infection, influenza is always closely monitored by the WHO. Influenza viruses are divided into type A, B, and C, of which type A ever caused the most serious epidemics of influenza in the world (Claas et al. 1992; Tharakaraman and Sasisekharan 2015). Influenza A virus contains the glycoproteins haemagglutinin (HA) and neuraminidase (NA). It includes 15 HA subtypes and nine NA subtypes. Influenza A detection can be performed with virus culture, immunofluorescence, nucleic acid diagnosis, serological detection, and antigen detection (Harmon and Pawlik 1982; Quinlivan et al. 2004; Weinberg and Walker 2005; Rawlinson et al. 2004). Except the antigen-based detection method, most of the existed detection methods, which require complicated instruments with trivial procedures, are time-consuming and high cost. For example, influenza (Flu) A detection based on nucleic acid analysis has to be completed with multiple steps including sample pre-treatment, nucleic acid extraction, nucleic acid amplification, labeling, and detection, which in totally even takes a couple of hours (Qiu et al. 2016; Tian et al. 2015). In contrast, more rapid, convenient, and low-cost influenza A detection can be achieved with antigen-based immunoassay detection.

Immunoassay is a detection method based on the interaction between antibody and antigen which can be implemented in a qualitative or quantitative way (Shepherd 2000; Harlow and Lane 1999). Because of the high sensitivity and low cost, enzyme-linked immunosorbent assay (ELISA) is a widely used immunoassay diagnosis method (Engvall and Perlmann 1971; Stanley et al. 1994). Normally, ELISA has to be performed on a complicated instrument by highly trained personnel. In contrast, lateral flow immunoassay is a simple, rapid, easy-to-use, and low-cost immunoassay method which is especially desired for POC diagnosis (Qiu et al. 2009, 2011a, b). Once the sample is applied onto the sample pad of the lateral flow strip, the target protein will be specifically captured by the immobilized protein on the test line area when it migrates along the strip because of capillary force. When the lateral flow strip is labeled with gold colloid, the detection result can be read by eyes without any instruments. Because of its natural limitation, it is difficult for the traditional ELISA to be performed at the point-of-care settings. As an optimized format, Dot-ELISA is developed to improve the sensitivity, specificity, and convenience (Piña et al. 2011; Diab et al. 2011). However, it takes at least 2 h for traditional Dot-ELISA to be manually completed with a disc made from NC membrane sitting in a micro-plate well. For POC diagnosis, Dot-ELISA or ELISA needs to be further improved to reduce the detection time and simplify the procedure.

Microfluidics provides solutions to integrate and automate the bio-analytical process based on lab-on-a-chip systems consisting of micro-valves, pumps, chambers, and channels (Chen et al. 2010; Qiu et al. 2011a, b). Until now, various diagnostics can be implemented on microfluidic chips, such as PCR amplification, nucleic acid extraction, cell sorting and capturing, and immunoassay detection. For most of lab-on-a-chip systems, the reagent has to be driven with integrated or external pumps and valves at different reaction steps, such as reagent dispensation and mixing, which limit the application of microfluidic systems. Since microfluidic paper-based analytical devices (µPADs) rely on capillary force or sometimes with gravity to drive the reagent without active pumping (Lankelma et al. 2012; Lópezmarzo and Merkoçi 2016), they are regarded as competent systems for diagnosis in POC settings. Once the sample or reagent is applied or discharged within the microfluidic system, it will join the reaction based on passive driving, which can significantly reduce the system complexity, size, and cost. Paper-based microfluidic systems for immunoassay or nucleic acid analysis have been developed to simplify the bio-analytical process (Hu et al. 2014; Gerbers et al. 2014). It can be anticipated that paper-based microfluidic systems will become more attractive in POC diagnosis.

With advanced embedded systems, smartphones can be regarded as desired devices to attain point-of-care diagnosis with simple operation, cost efficiency, high convenience, and portability (Mancuso et al. 2014). It is feasible for a smartphone to monitor the bio-reaction in real time or at the end point with its own camera by incorporating optical filters. The system complexity can be significantly reduced when a smartphone is incorporated properly because of its multiple functions. For software development, the standard function for camera control in Java environment is beneficial to simplify system design. If necessary, the detection results can be constantly transmitted with geographical record by smartphone to medical agencies for further analysis. Recently, various smartphone-based microfluidic analytical devices have been intensively studied and developed especially for POC diagnostics (Li et al. 2016; Yang et al. 2016).

In this paper, an integrated and sensitive paper-based microfluidic Dot-ELISA chip is developed to simplify the detection of influenza A with high efficiency. The integrated microfluidic chip consists of two modules. The storage module is developed for reagents storage and dispensation, while the other reaction module made of an absorbent pad and a functionalized NC membrane is developed for the detection of influenza A. With paper-based microfluidic chip, once the reagent is discharged onto the NC membrane, it will react with and flow through the NC membrane without any active pumping. Flu A detection result is attained by taking image from the NC membrane with a smartphone. The image is further analyzed to achieve qualitative analysis by an intelligent algorithm of custom application software developed with Java. With smartphone, the detection result can be conveniently displayed and transmitted if necessary. The utility and performance of the developed paper-based microfluidic system were systematically evaluated and demonstrated.

2 Materials and methods

2.1 Paper-based Dot-ELISA for Flu A detection

Similar to lateral flow strip, a rectangle functionalized NC membrane (16 × 16 mm2) is adopted as the solid substrate to capture the antigen of Flu A in the test sample with immobilized specific antibody (Ab). The test (ellipse) and control (circle) spots are fabricated by, respectively, immobilizing 14B11 antibody and another specific antibody for enzyme capturing, as shown in Fig. 1a.

Fig. 1
figure 1

Schematic depiction of paper-based Dot-ELISA

Full size image

The flow through protocol of paper-based Dot-ELISA is described as following. First, the lysed sample is applied onto the NC membrane, and the antigen of Flu A will be captured by the immobilized 14B11 on the test spot area. After that, HRP (labeled with the antibody 14F10) is applied, and it is captured by both the test and control spot areas. And then, the substrate is applied, and it reacts with the specifically captured HRP until the reaction is terminated by an applied stop buffer (Xu et al. 2008). After that, both the test and control spots become purple for a positive test. For a negative test, only the control spot becomes purple while the test spot is blank. For a failure test, the control spot always shows blank. As shown in Fig. 1b, when the sample or reagent is applied onto the NC membrane, it will flow through the NC membrane until it reaches the absorbent pad because of the capillary force and gravity. Compared with traditional Dot-ELISA or typical lateral flow immunoassay, the sensitivity of Dot-ELISA can be significantly improved based on the flow through working mode.

2.2 Paper-based microfluidic chip

As shown in Fig. 2a, in order to implement Flu A detection in POC settings, an integrated paper-based microfluidic chip is developed. The whole chip includes a reagent storage and reaction modules. As shown in Fig. 2b, d, the storage module is used for different reagent storage with totally six independent chambers which, respectively, for lysed sample, HRP, wash buffer #1, wash buffer #2, substrate, and stop buffer. Each reagent can be discharged by piercing its own storage chamber from a specific location with a very thin bottom layer. Each discharged reagent will be dispensed into the corresponding incubation chamber by following its own releasing channel. The incubation chamber is designed to temporarily hold the reagent to allow it to flow through the NC membrane for reaction. Totally three different incubation chambers are incorporated for different purposes. As shown in Fig. 2e, the chamber ‘a’ allows wash #1 and #2, substrate, and stop buffer to consecutively interact with the entire NC membrane through a big circular hole. The chamber ‘b’ allows HRP to interact with the entire NC membrane through a big circular hole. The chamber ‘c’ allows the sample to reach the NC membrane through a small ellipse and circle holes to, respectively, interact with the test and control spots on the NC membrane. With the incubation chamber ‘c,’ the lysed sample can only flow through the test and control spot of the NC membrane, which ensures the high sensitivity of Flu A detection. An independent incubation chamber ‘b’ is designed for HRP to prevent contamination inside the microfluidic chip. As shown in Fig. 2c, as a companion part of the storage module, a reaction module made of a NC membrane and an absorbent pad is fabricated. As shown in Fig. 2f, the NC membrane in the middle of the reaction module is used for Flu A detection, and an absorbent pad under the NC membrane is used for waste absorbing.

Fig. 2
figure 2

Paper-based microfluidic chip for Flu A detection. a Microfluidic chip. b Reagent storage module. c Reaction module. d Storage module. e Incubation chamber. f NC membrane

Full size image

2.3 Smartphone Flu A detection system

As shown in Fig. 3, a smartphone-based diagnosis system was developed to perform Flu A detection with a paper-based microfluidic chip. In Fig. 3a, a smartphone (Red rice Note 3, MI, China) is used to guide the microcontroller via the Bluetooth to actuate the microfluidic chip to perform multiple reaction steps, and at the same time, it collects the detection result from the NC membrane with its own camera.

Fig. 3
figure 3

a Schematic depiction of the smartphone detection system, b picture of the Flu A detection system, and c the reagent release module

Full size image

In Fig. 3b, c, sample or reagent can be released from the storage module when the bottom layer of the reagent storage chamber is pierced by a plastic needle. Multiple plastic needles with different lengths are mounted on a single linear motor, and each time only one of them will pierce the corresponding storage chamber according to the protocol of paper-based Dot-ELISA (see section of Paper-based Dot-ELISA for Flu A detection). To ensure each storage chamber to be pierced accurately, the needles are driven up and down by a linear motor along a predefined path which is reasonably aligned with the storage module based on a precise mechanical design (position error ≤ ±0.1 mm). Moreover, the diameter of each needle is 1.2 mm, while the diameter of the specific thin area for piercing in each storage chamber is 1.6 mm, which further improves the alignment robustness of the mechanical actuation part. Another linear motor is used to horizontally move and align the NC membrane to different incubation chambers consecutively according to the protocol. After the stop buffer flows through the NC membrane, the reaction module is moved to the position under the smartphone for image collection. The image of the NC membrane is analyzed by an intelligent algorithm to determine the detection result. Custom application software is developed with Java to direct the microcontroller, process the image, display the detection result, and transmit it if necessary. The power is provided by an AC/DC adaptor (12 V/1A). As a low-cost POC device, the Flu A detection system can be assembled with components costing less than $120 especially when a secondhand smartphone is adopted.

2.4 Image analysis for Flu A detection

Since the Flu A detection result is determined by judging the signals of the test and control spots on the NC membrane, it is important to first identify the control and test spots accurately and then extract their signal intensity amplitude. First, the original image (Fig. 4a) is processed to get a sub-image (Fig. 4b) covering the entire reaction area, and then, it is further converted into a gray image (Fig. 4c) which is defined as I_g.

Fig. 4
figure 4

a The collected original image, b sub-image for processing, c the gray image, d the converted gray image, e template for control spot searching, f the specific area for test spot searching, and g template for test spot searching

Full size image

For image processing, an edge detection algorithm is used to identify the boundaries of both the control and test spots. To improve the accuracy of boundary searching, a nonlinear amplitude converting method is adopted here. First, the maximal and minimal gray values in I_g are identified and defined as Max_g and Min_g, respectively. Then, a nonlinear converting function is proposed as follows (Li and Cai 2011),

$$d\left( {x, \, y} \right) \, = \exp \left( {\tfrac{{f\left( {x, \, y} \right) - {\text{Min}}\_{\text{g}}}}{a}} \right) - 1 ,$$
(1)

where f(x, y) and d(x, y) represent the original gray value and the converted gray value of the pixel in (x, y), respectively. Meanwhile, \(a = ({\text{Max}}\_{\text{g}} - {\text{Min}}\_{\text{g}})/5.545\). By doing this, gray value f(x, y) of each pixel in I_g is converted to a new value d(x, y) between 0 and 255, which generates a new gray image I_ng (Fig. 4d).

As shown in Fig. 4e, g, two templates, respectively, for control and test spot searching are designed. Based on the physical property of the NC membrane, the control spot can be regarded as a circle with a fixed radius of r pixels, while the test spot can be regarded as an ellipse with a fixed major axis of a pixels and a fixed minor axis of b pixels. Two templates are used to, respectively, determine the center point of the control and test spots. And then, the boundary of the test or control spot can be identified with the template.

Because of the physical property of paper-based Dot-ELISA, the gray value of those pixels between the area inside the control spot and the area outside the control spot is significantly different from each other. Therefore, the boundary of the control spot in I_ng can be identified when its center point is determined based on the following principle,

$$(x_{c} ,y_{c} ) = \mathop {\arg \hbox{max} }\limits_{(x,y)} \left( {G[r,r + 2,(x,y)] - G[r - 2,r,(x,y)]} \right)$$
(2)

where G[r1, r2, (x, y)] represents the mean gray value within a circle ring with a center point at (x, y) and the radiuses of the inside and outside circles are r1 pixels and r2 pixels, separately. The boundary of the control spot is identified based on the template after its center point \((x_{\text{c}} ,y_{c} )\) has been determined.

As shown in Fig. 4f, in order to identify the test spot quickly, the searching area is shrunk by switching the top boundary from y = 0 to y = y c  + r + 10 in I_ng based on the physical relationship between the control and test spots. Similarly, another principle for boundary searching of the test spot is defined as follows,

$$(x_{t} ,y_{t} ) = \mathop {\arg \hbox{max} }\limits_{(x,y)} \left( {G[a,b,a + 2,b + 2,(x,y)] - G[a - 2,b - 2,a,b,(x,y)]} \right) ,$$
(3)

where G[a1, b1, a2, b2, (x, y)] represents the mean gray value within an elliptical ring with a center point at (x, y), the two radii of the inside ellipse are a1 and b1 pixels, and the two radii of the outside ellipse are a2 and b2 pixels, respectively. The boundary of the test spot is identified based on the template after its center point (x t , y t ) has been determined. When the value max (G[a, b, a + 2, b + 2, (x, y)] − G[a − 2, b − 2, a, b, (x, y)]) at (x t , y t ) in I_g is larger than an empirical threshold 2, a positive test is recognized. Otherwise, a negative test is recognized. When the value max (G[r, r + 2, (x, y)] − G[r − 2, r, (x, y)]) at \((x_{\text{c}} ,y_{c} )\) in I_g is less than a threshold 2, a failure test is recognized.

3 Results and discussion

In principle, different subtypes of Flu A virus, such as H1N1, H1N2, H5N1 can be detected with the developed system. In this study, instead of using real blood sample or nasopharyngeal swab, H1N1 virus sample cultured with ‘A/Calliforia/4/2009’ (original concentration: 32HA, HA: hemagglutination titer) was detected as an example. Detailed procedure for paper-based Dot-ELISA is briefly described. To explore the sensitivity of the developed paper-based microfluidic system, the original sample was diluted with the lysis buffer in the experiments. First, 600 µL lysed sample is applied, and the sample incubation time is about 8 min when the sample flow through the NC membrane. The second step takes around 3 min to allow 200 µL HRP to flow through the NC membrane. The third step takes around 3 min to perform twice wash (each with 350 µL wash buffer). The fourth step takes around 3 min to allow 100 µL substrate to flow through the NC membrane. Finally, it takes 3 min to allow 50 µL stop buffer to flow through the NC membrane and read the detection result. Each test requires about 20 min.

3.1 Sensitivity evaluation with different operation modes

As shown in Fig. 5a, two types of microfluidic chips with different chambers for sample incubation were designed and compared. For type I, because there are two small holes on the bottom layer of the incubation chamber [similar to Fig. 2e (c)], the sample will flow through the NC membrane just from both the test and control spots simultaneously for a long-lasting time (~8–10 min), which contributes high detection sensitivity. For type II, because there is a big through hole on the bottom layer of the incubation chamber [similar to Fig. 2e (a)], the sample will flow through the entire NC membrane in a short time (~1–2 min), which results in low detection sensitivity. Flu A samples with different concentrations (100-fold and 1000-fold diluted from the original concentration) were used to evaluate the sensitivity of two different chips.

Fig. 5
figure 5

a Two types of incubation chambers, b and c two detection results, respectively, for type I and II with high concentration sample, d and e two detection results, respectively, for type I and II with low concentration sample. Insert numbers in c and e are mean gray values of different areas. Control and test represent, respectively, the control and test spot areas, and BK denotes the background of the NC membrane

Full size image

For the experiment with high concentration sample, as shown in Fig. 5b, c, the absolute signal amplitude values (after subtraction the background) for both the test and control spots of type I are significantly higher than those of type II. Similarly, for the experiment with low concentration sample, as shown in Fig. 5d, e, the absolute signal amplitude values (after subtraction the background) for both the test and control spots of type I are significantly higher than those of type II. Especially, for type II with low concentration sample, the detection signal of the test spot cannot be successfully differentiated from the background, which results in a false positive test. The experiment has been repeated for more than three times, and similar results were attained. Therefore, it is important to adopt type I incubation chamber in the developed microfluidic chip to ensure the high sensitivity.

3.2 Performance evaluation of the type I microfluidic chip

To evaluate the performance of Flu A detection with the paper-based microfluidic system, the virus samples with different concentrations from tenfold to 10,000-fold dilution were used. The sample was lysed outside the microfluidic chip, and then the lysed sample was introduced into the sample chamber of the storage module for detection. Other reagents were put into multiple storage chambers before the microfluidic chip was inserted into the device. Lysis buffer was used as the sample for negative control test. As shown in Fig. 6a, for each positive test, both the test and control spots show purple which is different from the background. However, for negative test, there has no difference between the test spot and the background although the control spot shows purple.

Fig. 6
figure 6

a Flu A detection result of the test sample with different concentrations (1/10,000 to 1/10 dilution) and b dose response curve of Flu A detection with the paper-based microfluidic chip (32 × 10−4 HA: a relative concentration unit, HA hemagglutination titer)

Full size image

As shown in Fig. 6a, the color of the test spot on the NC membrane becomes darker when the sample concentration increases. For negative test with lysis buffer, the color of the test spot is unchanged although a dash ellipse is put on the supposed reaction area for location marking.

For each original image of the NC membrane with Flu A detection, it was analyzed with the developed algorithm to not only identify the position of the test or control spot, but also determine the mean gray value within the test or control area. Furthermore, a differential gray value was attained by subtracting the previously achieved mean gray value with the gray value of the background to improve the accuracy of quantitative analysis. According to the Beer–Lambert Law, the relationship between the intensity of the transmission light (e.g., gray value in colorimetric detection) and the concentration of the test sample is represented by a logarithmic mathematical model. Therefore, log scale is used in the calibration curve. As shown in Fig. 6b, the dose response curve of Flu A detection is obtained based on the sample concentration and the differential gray value. The dose response curve can be fitted by \(y = 33.5786\log_{10} (x) + 16.917\) with \(R^{2} = 0.9764\), which establishes a reasonable linear model for Flu A detection. It can be anticipated that in principle, the developed paper-based microfluidic chip is able to perform quantitative Flu A detection. The experimental result shows that the sensitivity of the paper-based microfluidic chip meets the requirement for clinical diagnosis of Flu A. The experiments with the developed paper-based microfluidic chip were repeated for more than 20 times, and the coefficient of variation of the device is about 8.5%. Colorimetric detection with smartphone does partially depend on the smartphone types because of the performance of optics. To achieve the reliable detection result, two test samples, respectively, with low and high concentrations can be used for system calibration when different smartphones are adopted.

3.3 Dead volume analysis of the microfluidic chip

The dead volume of the reagent storage module was analyzed to evaluate the performance of the microfluidic chip. In the experiment, each type of reagent with a known volume was put into the storage chamber. The reagent storage module was weighted with a precise electrical balancer after the reagent was loaded, and it was weighted again after the reagent was discharged, and the dead volume can be figured out based on the difference of those two weights of the reagent storage module. Experiments have been repeated for multiple times. As shown in Fig. 7, the largest dead volume among all the storage chambers is under 13%.

Fig. 7
figure 7

Dead volumes for different storage chambers

Full size image

Principally, the dead volume is caused by the surface absorbing of the reagent to the chamber and channel surface. To reduce the effect of dead volume, additional amount of reagent can be added into the storage chamber for compensation to ensure the performance of the microfluidic chip. Alternatively, although not desirable, a micro-pump can be used to drive the reagent with less dead volume. As is well-known, the simplified flow control is always preferred in a POC microfluidic system to reduce the complexity and cost of the whole device.

4 Conclusions and outlook

A portable, low-cost, and integrated paper-based microfluidic system with smartphone has been developed for influenza A detection in POC settings. Paper-based Dot-ELISA is implemented on microfluidic chip in a flow through working mode to improve the sensitivity of Flu A detection. Functionalized NC membrane with a test and control spot is used for Dot-ELISA, and an absorbent pad under the NC membrane is used for waste absorbing. To implement POC diagnosis of Flu A, a reagent storage module with a couple of independent chambers is incorporated into the integrated paper-based microfluidic chip. When each reagent chamber is pierced by a plastic needle from its bottom, the reagent will be discharged with allowable dead volume and interacts with the NC membrane sitting on the reaction module. To ensure the detection sensitivity, a specific sample incubation chamber, which holds two small through holes to confine the reaction just on the specific areas, is designed to guide the sample to flow through the NC membrane at a desired low speed just from the test and control spots.

A portable electrical–mechanical device with low cost is developed to perform Flu A detection with paper-based microfluidic chip. A vertical motor is used to drive the plastic needle for reagent discharging, while another horizontal motor is used to move and align the NC membrane to different incubation chambers according to the multiple reaction steps. Once the reaction is completed, the detection result can be attained by taking image from the NC membrane with a smartphone camera. An optimized custom intelligent algorithm is developed to analyze the image to successfully extract the signal amplitude of the test and control spots. Custom application software developed with Java is used to guide the microcontroller of the portable electrical–mechanical device to control the motors to complete the multiple reaction steps for Flu A detection. Moreover, it is used to process the image, display the detection result, and transmit it to other medical agencies if necessary. Experimental results show that reasonable sensitivity for Flu A detection can be achieved with the developed paper-based microfluidic system. Furthermore, the dose response curve of Flu A detection shows that, in principle, quantitative detection can be achieved with the developed paper-based microfluidic system.

With paper-based microfluidics utilizing flow through working mode, the diagnosis system with high sensitivity can be significantly simplified by eliminating complicated fluid control. For POC test, it is desired to pre-store the required reagent inside the microfluidic chip to perform detection in enclosed environment with high efficiency and convenience without contamination. The complexity of the integrated paper-based microfluidic chip can be further reduced by a simple reagent discharging method relying on multiple plastic needs mounted on a single motor. A smartphone is incorporated into the diagnosis system properly because of its unique and combined capacities on taking and processing imaging with custom application software, device operation, and data analysis and transmission. In summary, a portable, easy-to-use, low-cost, and sensitive paper-based microfluidic system with the assistance of smartphone is successfully developed. Although, in this study, Dot-ELISA is implemented on this paper-based microfluidic platform for Flu A detection, in principle, this concept can be utilized to develop other similar immunoassay diagnosis system in POC settings.