Medical Device Link.

IVD Technology Magazine
IVDT Article Index

Image analysis for rapid-flow diagnostics

Thomas C. Tisone, Michelle Rodriguez, and Paul Queeney

Over the past decade, membrane-based lateral-flow assays have become important tools in medical diagnostics. Using this test format, researchers have demonstrated the clinical utility of more than 150 different analytes, and many of these have been introduced as products.

Segmented analysis of the reagent lines of a rapid-flow diagnostic can help control manufacturing quality.

R&D for such devices often involves significant testing, which is sometimes conducted as an orderly design-of-experiments process, but is just as often performed on a trial-and-error basis. Process development and quality control (QC) are equally rough-and-ready, with experience and guesswork often substituting for hard data.

Advances in technology can help to streamline these phases of product development and take guesswork out of the equation. This article discusses the use of automated imaging analysis methods as a tool for both R&D and QC. Applications of this method to studies of reagent application, dilution, process development, and device kinetics are also presented.

The Principles of Rapid-Flow Testing

Rapid-flow immunochromatographic test devices are made up of a number of components, commonly including a sample pad, a conjugate pad, a membrane that incorporates capture reagents, and an absorbent pad (see Figure 1). In practice, the user dispenses a patient sample (usually urine or whole blood) onto the sample pad. The sample then flows through the sample pad into the conjugate pad, where it mixes with and releases the detector reagent. This mixture then flows across the membrane, where it binds with the test and control reagents. When the mixture binds to the reagent that forms the test line, a positive result is indicated. The color intensity of the test line is proportional to the concentration of analyte in the sample. Excess sample that flows beyond the test and control lines is taken up in the absorbent pad.

Figure 1. Top and side views of a typical rapid-flow immunochromatographic test device, including a sample pad, a conjugate pad, a membrane that incorporates capture reagents, and an absorbent pad.

Although such tests appear simple, complex interactions among their various components lead to a number of challenges in both the development and manufacturing environments. These challenges are even greater for quantitative tests, where the color intensity of the test line must be repeatable between production lots. From a developmental perspective, creating a successful test system means optimizing the interactions between its raw materials, component design, and manufacturing techniques.

Whether the developer is creating a qualitative test that provides only a positive or negative response, or one that offers a quantitative measurement, optimizing every aspect of the test system requires detailed information. It is easy to detect gross problems caused by nonspecific binding, for example, but it can often be very difficult to eliminate this problem without tools that provide quantitative information about the functioning of the test system. To optimize a test's design and manufacturing process, in other words, the manufacturer must move beyond qualitative criteria and guesswork to the use of quantitative measurements. Such detailed measurements can assist the product designer in verifying that the system acts in an accurate and repeatable manner.

Many of the attributes of an immunochromatographic test can be displayed and measured using a test-strip reader. For instance, such a reader can display the capture lines at magnification and provide an accurate graph of the background and signal levels along the axis of flow. It can also display graphically and quantitatively (numerically) the maximum and actual strength of reagent capture, asymmetries in the capture line, and the rate of change in the formation of the capture line as the test is performed.

Image Processing

An image processing system suitable for analyzing rapid-flow diagnostics typically consists of three main components: an electronic camera, a frame grabber, and image processing software installed on a personal computer (PC). A variety of low-cost electronic cameras based on charge-coupled device (CCD) technology are commercially available. Instead of focusing an image onto film, as with a normal camera, an electronic camera focuses the image onto an array of CCD elements or pixels (picture elements). The pixels convert light to electrical signals and increase their charge or voltage in proportion to the amount of light they receive. The resolution of the camera is a function of the image size relative to the number of pixels. Typical low-cost CCD cameras have arrays with about 250,000 pixels. Most electronic cameras also feature automatic gain control (AGC), which adjusts the camera's gain based on the amount of light the camera receives. Because this feature must be disabled in order to calibrate the system, the camera incorporated into an image processing system for analyzing rapid-flow diagnostics must have an option to turn off the AGC.

A frame grabber is a printed circuit board that can be added to a PC as an interface between the camera and the software. The frame grabber incorporates a converter that enables it to read picture information provided by the camera, specialized hardware that can perform image processing functions very quickly, and a high-speed memory for storing those images.

The image processing software uses picture data stored in the frame grabber. Commercially available software packages range from general-purpose image processing programs to those tailored specifically for rapid-flow tests. The software should display a magnified image of what the camera sees and allow the user to perform various functions and measurements. For instance, the software should have the ability to subtract the background, automatically find and measure lines, calibrate to current light levels, export measurement data to a spreadsheet, and save and retrieve images. The better software packages have an integrated image acquisition and processing program.

The user interface of such programs is also important, because they are typically complex and have a large number of settings and options. At a minimum, the user interface should display the image, have a scaleable processing window superimposed on the image, and display a graph of average intensity versus the x-axis (see Figure 2).

Figure 2. The user interface of the BioDot SR3000 test-strip reader, showing its multiple windows.

The Test-Strip Reader Concept

The first requirement for any system designed specifically to image and analyze rapid-flow tests is the ability to position the test under the camera and illuminate it with controlled lighting. The light level should be relatively constant across the camera's entire field of view. The SR3000 from BioDot (Irvine, CA), for instance, has a light box that blocks out ambient light and a multiple-element lighting system that provides uniform light across the field of view (see Figure 3). Note that this system also uses monochromatic (single wavelength) light instead of broadband light. The wavelength of the light is selected to increase the contrast of the developed test and control lines with the background (membrane). Selecting the correct wavelength is crucial for imaging very low levels of signal. This system also has an adjustable cartridge to hold tests in place, and spacers to keep the top surface of a device at the same distance from the camera (at the required magnification level, the camera has a very restricted depth of field).

Figure 3. The BioDot SR3000 light box is a self-contained system with device holder, camera, lighting, and power supply.

Software. Two sets of operations are required to turn images of rapid-flow tests into useful data. The first is an image processing and measurement program, and the second is a data analysis program.

The image processing software used for the SR3000 was developed to analyze and measure test strips with lines. The user interface consists of three main windows, which display a graph showing the density profile (versus the x-axis) along with the peaks and edges of any lines found, an image of the strip, and measurement results (see Figure 2). The vertical lines in the graph and image windows show where the edges of the lines occur. The operator can select areas of interest by overlaying a measurement window directly onto the image of the strip. This measurement window can be moved or sized as required. The program operates on the values within the measurement window and averages the density along the vertical or y-axis. This average density is displayed as a graph. All further mathematical operations are carried out on the graph data as opposed to the image data.

Figure 4. The ideal density profile for a reagent line, a near-Gaussian waveform, indicating uniform binding of the reagent to the membrane.

Figure 5. An asymmetrical waveform indicating a reagent line with varying density and nonuniform binding.

Calibration and Repeatability. One requirement of a test-strip analysis system is to be able to compare image density (darkness) with engineering standards. Another important requirement of the imaging system is the ability to compare the measurements made on one test strip with those made on strips manufactured at different times or even at different locations. The feature that allows these things to happen is calibration. Two types of calibration are required: spatial and density. Spatial calibration is used to calibrate the imaging system for distances and locations. Density calibration correlates the brightness or darkness seen by the camera with optical standards. Thus, all measurements are correlated to industry standards (a step-density chart such as the Electronic Industry Association gray scale is typically used). On the SR3000, the repeatability of measurements on a single system or between multiple systems is approximately 99%—that is, the system's coefficient of variance (CV) is about 1%.

Sensitivity. To be useful in analyzing a rapid-flow test, the imaging system has to be sensitive enough to distinguish between different levels of density in a quantitative test. The BioDot system has a resolution of about 1% across the full range of density values, meaning that it can distinguish approximately 100 shades of gray between black and white. Sensitivity is determined by the step size or difference between one shade of gray and the next; if the range is reduced, finer shades of gray can be distinguished. Resolution and range can be traded off for one another by adjusting the camera's iris.

Application Studies

In practice, use of an automated image analysis system can enable test developers to measure and evaluate a number of variables that can affect the performance of their products. The sections below suggest some of the more common applications that can help manufacturers to reduce product development time or improve manufacturing quality.

Line Shape Measurement. One measure of a controlled process is the symmetry of the reagent dispensed onto the membrane. Figures 4 and 5 show different types of tests and the resultant density profiles of their reagent test lines. The differences in the density profiles are a result of different reagent chemistries and methods of application. Figure 4 shows the ideal density profile, a near-Gaussian waveform, whereas Figure 5 shows a skewed or asymmetrical waveform.

Quality Control for Manufacturing. Typical in-process quality control measurements include the locations of the lines, line widths, maximum darkness, and integrated optical density (IOD). IOD is the product of the average density of the line times the area of the line; it is essentially the area under the curve in a graph of density versus x-axis. The BioDot system can measure the average width of the reagent line as well as variations in the width along the axis of the line. Figure 6 shows typical data for multiple samples taken during a production run.

Kinetic Studies. The flow rate of a conjugated patient sample through a test membrane can affect the density of the colored test line that develops as the test is performed. Although the intensity of the color is proportional to the concentration of analyte in the sample, the effective concentration of all reactants decreases linearly as flow rate increases. In general, the background will also drop with increased flow rate. Thus, flow rate affects the ultimate color of the test line as well as the sensitivity of the device. Device sensitivity decreases with the square of the increase in flow rate.

The BioDot test-strip reader is a useful tool for accurately and reproducibly measuring the darkness (color) of the test line and background—both for the finished test and while the test is developing. Although many components and properties might be measured to determine their influence on the outcome of the test (e.g., the pore size of the membrane; the concentrations of surfactants, buffers, and sample fluids), these offer only qualitative information. The test-strip reader provides quantitative measurements of the interactions of these components and their processing methods, and thereby the resultant performance of the entire system.

Figure 6. Typical in-process quality control data for multiple samples taken during a production run.

As an example, the BioDot test-strip reader was used to compare two ways of processing a conjugate pad. Conjugate was applied to one pad using an airbrush, and to another using a BioJet dot printer. The tests were then developed using 200 µl of sample reagent, and were imaged every 10 seconds for 15 minutes. The images were then processed to measure the development of line width, maximum darkness, and IOD as a function of time (see Figure 7). The analyses indicated that each conjugate reached the same level of intensity, as expected, but that the rate of the conjugate release was faster for the dot-printed pad.

Sample Dilution Kinetic Studies. The test-strip reader can also be used to quantify test-strip color as a function of sample concentration. A set of identical rapid-flow tests was prepared. Solutions with increasing concentration of sample were prepared and used to develop the tests. Sample volume was 200 µl. Images were taken of the tests as they developed. Figure 8 shows the integrated intensity versus time for 50, 75, and 100% concentration levels.

Figure 7. Comparison of conjugate application methods and test line development as a function of time, as measured by the BioDot SR3000.

Sensitivity Studies. The BioDot test-strip reader incorporates an image processing program that enables it to detect very low level signals. The program is capable of finding the test line for signal concentrations as low as 3% in a 200-µl sample. Like the human eye, the test-strip reader distinguishes test lines by seeing the contrast between lines and their background. Unbound conjugate elevates the background signal and reduces contrast, making it difficult for the system to discern the test lines. If the unbound conjugate is completely washed through, the system can find signal levels lower than 2% (see Figure 9).

Figure 8. Graphs of integrated intensity versus time for 50, 75, and 100% concentration levels of a sample solution.

Data Processing. The BioDot system has an interface to commonly used spreadsheets and databases such as Lotus, Excel, and Access. Once the data have been exported to one of those programs, the user can perform standard statistical functions and create graphs such as those used for this article. The system also includes a spreadsheet template that calculates averages, standard deviations, and CVs from exported data.

Figure 9. The integrated intensity of fully developed test strips versus sample concentration of hCG. The image processing program of BioDot's SR3000 enables it to consistently measure signal concentrations as low as 3% in 200 µl of sample, or to measure signal levels lower than 2% when unbound conjugate has been removed.

Conclusion

Using an automated test-strip reader reduces the time required to measure and record performance data, and can also improve the accuracy of such measurements. Where manual measurements commonly have CVs in the range of 6 to 8%—in part because of operator fatigue and subjectivity—automated systems can reduce this to a mere 1%. Moreover, automated systems can measure variables that are virtually impossible to determine manually, including integrated optical density, maximum density, contrast, line shape, and flow kinetics.

Such commercial image processing systems are valuable tools for both developers and manufacturers of diagnostic test strips. They offer the accuracy and sensitivity required to quantify the shape and density levels of developed test and control lines. And they provide a cost-effective means of reducing development time and improving the overall manufacturing quality of rapid-flow tests.

Thomas C. Tisone is vice president for R&D and engineering, Michelle Rodriguez is applications specialist, and Paul Queeney is a consultant at BioDot Inc. (Irvine, CA).

Back to the IVDT Sept. 99 table of contents