gms | German Medical Science

Introduction: In order to prevent a further increase in health care costs, there is a growing need for cheap and disposable point-of-care tests (POCT), which could for example enable the rapid detection of the spread of pandemics. These tests should be simple enough to be administered and evaluated by the patient itself. A first approach is to use classic Lateral Flow Tests (LFT). These are already ubiquitous in the form of pregnancy or simple drug tests. Several groups have proposed microfluidic paper-based analytical devices (µPads) as an even less expensive alternative. They have the advantage of a cheap substrate material and a rather high number of structuring techniques, including straightforward cutting of the substrates to selective plasma etching or polymerization steps [1]. This enables the integration of multiple detection zones for multiparametric detection without cross contamination as well as complex fluidic networks for multistage detection reactions.

However, both, LFTs and µPads share a few unresolved challenges, including restricted capabilities for quantitative read-out and multiparametric detection without the use of dedicated readout devices. This limits the use for sophisticated point-of-care testing.

Methods: We propose to address these limitations with a laser-structured nitrocellulose membrane strip with multiple parallel channels. To allow for one-step test execution, the test is based on the sandwich assay principle. The disadvantage of this principle is the Hook effect, meaning false inaccurate and false negative results at high analyte concentrations. Thus, the fluidic design includes three reaction zones per channel (test, antigen, and control) [2]. This enables the reliable detection of wide concentration ranges, such as 1 ng/mL upto 500 μg/mL for C-reactive protein (CRP).

In order to keep the costs low and simultaneously allow for a quantifiable readout, the evaluation is carried out by an automated image-processing algorithm running on an (low-cost) smartphone. Here, no expensive dedicated equipment, such as camera add-ons (e.g. lenses), or additional chemicals are needed. The algorithm incorporates multiple steps including (1) strip localization within the frame, (2) reaction zones detection, and (3) gray scale intensity calculation. For constant results, regardless of ambient conditions (mainly lighting) and the device used, multiple correction steps have been added, including camera independent exposure and white balance (Grayworld algorithm), shadow correction using adaptive thresholding as well as contrast limited adaptive histogram equalization (CLAHE algorithm).

The algorithm was developed using C++ and the popular image-processing library OpenCV, which is highly optimized for weak hardware. It was integrated in both Google Android and Apple iOS apps with native graphical user interfaces. Overall, this allows for the widest distribution possible.

Results: The average processing time per frame is 441ms per frame with an average of 669ms until all intensities are obtained. The measured values are constant and comparable between devices; the standard deviation is in the range of ± 3%.

Discussion: First results have shown that the presented approach, consisting of laser-structured test and an advanced image-processing algorithm, can overcome previous limitations. Future work includes calibration with defined analyte concentrations. Thus, a conclusion of intensity to concentration becomes possible.

Die Autoren geben an, dass kein Interessenkonflikt besteht.

Die Autoren geben an, dass kein Ethikvotum erforderlich ist.