PocketELISA: A Low-Cost Portable ELISA Reader Based on Image Analysis over PDA Platform for Clinical Diagnose in Medical Veterinary

Transcription

1 PocketELISA: A Low-Cost Portable ELISA Reader Based on Image Analysis over PDA Platform for Clinical Diagnose in Medical Veterinary Miguel A. Pérez 1, Martín Mera 2, Jose R. Arias 2, Beatriz G. Arza 1, Carlos E. Carleos 3, Rocío Muñiz 1, Cristina de la Torre 1, Dionisio Cifuentes 4, Juan C. Cueli 4, Jesús A. Baro 5 1 Dept. de Ingeniería Eléctrica y Electrónica. Oviedo University, Gijón (SPAIN) 2 Dept. de Informática. Oviedo University, Gijón (SPAIN) 3 Dept. de Estadística e Investigación Operativa, Gijón (SPAIN) 4 Sersia Cantabria S.L., Muriedas (SPAIN) 5 Dpt. de Ciencias Agroforestales. Valladolid University, Palencia (SPAIN) maperezg@uniovi.es Abstract- ELISA (Enzyme Linked Immuno-Sorbent Assay) diagnostic plates are widely used for the screening of viral infections. Currently, plates are measured either photometrically which is costly or manually which is unacceptably subjective and dependent on training and illumination. In this paper, it is proposed an instrument for reading-out plates by digital analysis of images in field applications. The main challenge is the design of a versatile platform using standard, low-cost elements, such as PDAs or embedded PCs, open libraries, and consumer-grade digital cameras. All these alternative solutions can provide similar I. INTRODUCTION Enzyme Linked Immuno-Sorbent Assay (ELISA), a useful tool used mainly in clinical immunology, is the first-choice test to detect the presence of certain pathogen in bodily fluids and it is used for routine screening of viral infections [1, 5, 6, 7, 9]. Based on the principle of specific antigen-antibody interaction, it makes no use of toxic or radioactive substances, and its outcome lies in a change in colour. ELISA is performed on 80 x 120 mm polystyrene plates with an 8 x 12 matrix of 10 mm wells, 7 mm in diameter. The plate wells are coated with pathogen antigens [1, 5]. Afterwards, each well is filled with a sample of suspected host fluids, then the plate is washed and all plate wells are filled with enzymeconjugated anti-immunoglobulin. After the second wash step, a dye is added and a change in colour is produced due to the presence of the enzyme. This change in the colour will be produced in the well only if a specific link between the coating antigen and the anti-immunoglobulin was produced in the sample, due to the presence in the sample of an immunoglobulin specific for the antigen, which constitutes evidence of infection. The colour reaction is proportional to the concentration of immunoglobulin in the sample. The absorbance of the plate well is measured photometrically at a wavelength specific for the dye, usually TetraMethyl Benzidine (TMB) at 450 nm. Spectrophotometers are used to read-out ELISA plates, but is a quite expensive piece of equipment and results with small differences in the final solution and costs. The purpose of this paper is the use of a PDA (Personal Digital Assistant) which includes a digital camera to constitute a portable and low-cost analyzer of ELISA plates, able to be used in field applications. Tests were performed taking several plates from routine monitoring in cattle of Brucellosis, red nose (Infectious Bovine Rhinotraqueitis, IBR) and Bovine Virus Diarrhoea (BVD). This work shows a technique to improve the efficiency of clinical interpretation ELISA diagnostic plates. in addition, is unable to repeat a read-out after short time because it cannot preserve original information about plate. Other alternative such as subjective determination by human operators is generally unacceptable because of its low precision, ambient dependence, and lack of repeatability. In the other hand, this paper presents an instrument to read automatically ELISA plates by means of a digital photograph, application of an image analysis algorithm, and display of the results. The PocketELISA application is an evolution of image processing [2] in traditional PC by means of OpenCV libraries [3, 4, 8].This technique analyses the results of the ELISA plates in a PDA (Personal Digital Assistant). The use of a PDA allows a large flexibility and mobility, so that field-tests can be carried out without delays because of waiting by the results. In addition, it introduces a significant reduction of equipment costs and is able to kept original information of plate (a picture). II. DESCRIPTION OF POKETELISA APPLICATION The developed technique lies in taking a picture of ELISA plate, in order to be analyzing by a PDA. The image can be loaded to the application in several ways: live capture by the on-board camera of the PDA, read from a digital camera flash room, etc. Previous tests that we have carried out showed that is better to obtain an image with a high contrast provided by a good illumination than an image with large amount of pixels. A typical PDA digital camera is able to give enough resolution,

2 but optics limitations and low profile characteristics impose strong restrictions on the illumination. Only 96 wells are read at each time, so that there is no need for high definition. Neither are stringent requirements for the original image format. Image processing is simple and can be built with any type of computer platform. An algorithm automates the measurement of the immune reaction at each one of the 96 wells of an ELISA plate. The developed algorithm takes the right colour component coordinate and calculates the colour level in a scale. This scale is obtained from reference wells; thus, minimum value is calculated by averaging of negative reference wells (filled with water) and a maximum value is calculated by averaging positive reference wells (filled with a standard reactive). In the Fig. 1, the general procedure of the PocketELISA application is shown. External Image capture image capture by PDA camera Image Load in PDA Calibrate Set reference wells Run Analysis Batch analysis queue Fig. 1. General procedure of PDA application for reading-out ELISA plates uses on-board camera capture or external input from other digital camera. When the image is in the PDA, program algorithm executes a process of spatial calibration. The calibration algorithm searches for the central point of each well of the ELISA plate. Then, the program will take colour measurement of each well. The program calibrates the image automatically and, when it ends, it asks the user about the founded wells in order to confirm the correctness of the calibration as it is show in the screen capture of Fig. 2. The next step is to set the reference wells in the ELISA plate. This is a manual work because there is not a common standard of placing the reference wells. The user must point on the image over each reference well for this specific plate and then, the program takes note of them for later use in the analysis. In this moment, the user can select two actions: to run the analysis of the test and get the results in a few seconds or to send the image to a batch queue. If the user chooses to use the batch queue, he can calibrate and set the references for another image. When he finalizes the pre-processing of all images, he can run the entire tests stored in the queue in batch. The general steps made by the algorithm which implements the ELISA plate analysis in the Pocket ELISA application is shown in the Fig. 3. Calculation of reference wells median Obtention of calibration line for analysis Input data: - Well positions - Reference wells - Image ELISA value 10 0 Output data: - Positive ref. value - Negative ref. value Colour level Calculation of wells median 10 POSITIVE NEGATIVE Well classification 0 Yes More wells? No Report results Fig. 2. Screen capture of an ELISA plate, after ending the spatial calibration process. It shows a dark circle in each well and the program is waiting for user validation of the process. In the low right corner, a small image of original capture is also shown. Fig. 3. General algorithm of ELISA plate analysis.

3 The algorithm needs a spatial calibrated image with the entire wells of the plate located and the reference (positive and negative) wells of each specific analysis defined by the user. The program uses the RGB components of the image to calculate the intensity of the colour in each well. The algorithm calculates the statistical value of the median for each well and this value will be the intensity value assigned to that well. The use of the median value is preferred over the mean in order to avoid the extreme values of light or shadows due to a deficient illumination when the image was taken. The ELISA analysis is constituted by an algorithm in two steps. In the first one, it is necessary to obtain the plate calibration parameters, and only the reference wells will be taken into account: for each reference well median and dominant RGB component is calculated. Thus, each reference well is defined as positive or negative reference and a colour scale to classify all wells is obtained. Then, the second step of the algorithm starts. In this step, for each well of the plate, the dominant RGB component found in the reference set is studied and the median calculated. This value is compared to the parameters of the reference set and the well is classified like positive, negative or indecisive. The proposed application presents the obtained results in different ways. The first one draws the result value of each well over the original image. Positive wells are drawn in green, negative wells, in red and, indecisive, in yellow as it is shown in Fig. 4. Fig. 5. Results of an ELISA plate analysis, showing the density of colour over each well in a formatted table. The quality of the captured picture affects to the results of the image analysis; thus, low quality images without light uniformity, shadow or shining areas introduce additional difficulty in the manual procedures such as confirmation of spatial calibration or definition of reference wells and as a result, produce poor results of analysis. The above mentioned problems can be solved by using a uniform retro-illumination as the one provided by EL (Electro- Luminescent) films. A simple illuminator with EL film at an appropriate emission wavelength is able to provide enough illumination uniformity without shadows or shining areas. The EL films use simple inverter to produce a 100 V 400 Hz drive signal as it is shown in Fig. 6 from a primary battery pack. Fig. 4. Results of an ELISA plate analysis, showing the density of colour over each well. Final results of analysis can be displayed as it is shown in Fig. 5: a formatted table that can be imported in most spreadsheet software with an image capture of the ELISA plate. Fig. 6. EL-based illuminator circuit for ELISA plates analysis. There are several types of ELISA plates depending on the objectives of analysis, that is, depending on illness (IBR, BVD, and Brucellosis). Manufacturers of ELISA plates introduce additional differences in final colour, from blue to red. Moreover, there are two main ELISA methods: direct

4 ELISA in which negative reference has higher colour density than the positive reference, and indirect ELISA plates, where positive reference has a more intense colour that the negative. The developed application overcomes these issues and searches for the dominant colour, calculating an ad hoc classification scale and as a result, the user interface is reduced. There is only a requirement: the EL illuminator must contain wavelengths used in the analysis, but it is not a problem if a white (near RGB) EL film is used. III. EXPERIMENTAL RESULTS An extensive test has been carried out to verify the behaviour of the developed algorithm. This process involves several steps: a) Filling of ELISA plates with blood samples from animals under test b) Developing ELISA plates by following standard procedures c) Reading-out ELISA plates in standard equipment (usually, a spectrophotometer) d) Reading-out ELISA plates in developed system without time delays to guarantee similar color density in steps, c and d e) Comparing results from steps c and d Fig. 7 shows the comparison between results provided by standard method (above step c) and proposed algorithm (above step d). Comparison returns total concordances in positive and negative identification but establish some limitations due to saturation in both methods but it does not affect to well classification. results. method based on spectrophotometry. As can be seen, there is a high concordance between the results of both methods. Case 1 Case 2 Case 3 A B C D E F G H A B C D E F G H A B C D E F G H A B C D E F G H A B C D E F G H A 0,30 0,34 0,27 0,27 0,29 0,66 0,38 1,62 B 0,26 0,32 1,82 0,33 0,36 0,84 0,34 0,32 C 1,57 0,28 0,28 0,29 0,27 0,29 0,80 0,37 D 0,26 0,32 0,37 0,77 0,28 0,30 1,23 0,33 E 0,27 0,34 0,30 0,28 0,34 1,05 0,26 1,68 F 0,36 0,26 0,25 1,68 0,28 1,42 0,34 0,33 G 0,40 0,29 0,36 0,29 0,28 1,27 0,27 0,32 H 0,36 0,30 0,29 0,32 0,91 0,32 0,69 0,33 Fig. 8. Experimental comparison between results provided by proposed method (above) and results from standard equipment for ELISA reading-out (bottom) for three cases of ELISA plates. Positive references are E12, F12, negative references are G12, H12 for Cases 1 and 2 and positive references are C1, E8 wells, and negative references, A1 and B1 for Case 3; highlighted (by reverse colour) wells in both tables denotes a positive well, grey denotes indecisive wells and white, negative well in all cases. As it can be seen, there is a high concordance between both methods. IV. CONCLUSIONS Fig. 7. Experimental comparison between results provided by proposed method and results from standard equipment for ELISA reading-out: total concordance in negative and positive classification. Some saturation appears in both systems but it does not affect to well classification results. Fig. 8 compares the results of well classification as positive or negative by using the proposed algorithm and standard This paper proposes an automatically method to read ELISA test plates. An algorithm estimates the concentration of specific immune products in the sample used to measure the degree of exposition to the pathogenic agent by means the colour intensity level. Overall, the method achieves very low costs, allows long-term conservation of the readings along with original data and high portability. This method has been implemented in a PDA with on-board camera, providing excellent results.

5 An extensive experimental verification returned a complete concordance in wells classification between results provided by standard equipment (and high-cost) ELISA reader and results from proposed method. REFERENCES [1] Cifuentes, D., Crespo, R., Prieto, M. and Relancio, A.. Enzyme immuno-assay (ELISA) as a routine serological test for the diagnosis of bovine brucellosis. XV World congress on cattle diseases. Dublin, [2] González J.C et al. A new low-cost reader system for ELISA plates based on automated analysis of photographic pictures, Proc. Of Intrernational research conference on brucellosis in small ruminants, Skopje [3] Open source computer vision library - reference manual, INTEL Corporation, , OpenCV website, [4] Shapiro, L. and Stockman, G.. Computer vision. Prentice Hall [5] Crowther, J.R.. Methods in Molecular Biology, Volume 42: ELISA, Theory and Practice. New Jersey. Humana Press, [6] Desphande, S.S.. Enzyme Immunoassays: From Concept to Product Development. New York. Chapman & Hall, [7] Diamandis, E.P. and Christopoulos, T.K. (Eds.). Immunoassay. New York. Academic Press, [8] Daugman, J. G., Computer science tripos: computer vision. [9] Ferencik, M.. Handbook of Immunochemistry. New York.Chapman & Hall, 1993.