Flowify AI®: Lateral-Flow-Test Evaluation using Machine Learning
Flowify AI® is a software solution from Liftric for the quantitative evaluation of lateral flow assays (LFAs) powered by AI.
LFAs are used as a diagnostics device in a wide variety of fields including healthcare, nutrition, veterinary, aviation and more. The most well-known examples of LFA tests are probably pregnancy tests and COVID-19 tests. In most cases, those tests are fairly easy to evaluate with the human eye as they are only qualitative (negative or positive). However, LFAs are also able to indicate the concentration of an analyte and be used as (semi-) quantitative tests. A widely distributed example would be a vitamin D test that measures a specific concentration (quantitative: e.g. 50 ng/mL) or classifies the concentration level (semi-quantitative: e.g. deficient, insufficient, sufficient).
An LFA test has a control line (band), which shows if the test is working properly, and one or more test lines that indicate the presence or absence of a target analyte. Depending on the concentration of the analyte, the test line is not visible or has a varying shade of red if the test uses gold nanoparticles.
The human-eye cannot reliably determine the strength of a test line (e.g. perceived brightness). Hence, a reader device is required in order to evaluate a quantitative LFA test. So far, dedicated hardware reader are typically used for this purpose. Those readers have the disadvantage that they are rather costly, often only work for a specific LFA, and are difficult to configure.
What is Flowify AI?
Flowify AI® is an image analysis algorithm for the quantitative analysis of LFAs. In a nutshell, Flowify AI® processes images of LFAs and provides a signal value (band intensity) for each line of an LFA (i.e. control line and test lines) that refers to its perceived brightness. Hence, it is applicable to LFAs having color labels (e.g. gold nanoparticles).
The computed signal values do only depend on the perceived brightness, which is independent from the target analyte or LFA. The signal value would be zero if the line is not visible and close to 1 if it has a dark shade of red. Based on the signal values, it is possible to obtain the following types of test results:
- Quantitative: The predicted signal values can serve as input of typical calibration functions (e.g. linear, 4PL, 5PL). For instance, it is also possible to compute the ratio of a test line and the control line or to sum the values of all test lines.
- Semi-quantitative: Multiple thresholds have to be set in order to obtain a semi-quantitative result.
- Qualitative: A single thresholds has to be set to distinguish positive and negative sample.
Flowify AI® is already compatible to a wide range of tests and not limited to a specific test cassette. Moreover, it can easily be adapted to new tests. Flowify AI® is designed to run on mobile devices, but it can obviously process any type of images or image streams and also run in the cloud.
How does Flowify AI® work?
Flowify AI® uses artificial neural networks in order to process the images and to obtain the test results. In a first step, the different test cassettes and their components get detected. Subsequently, the quantitative analysis (regression) of the signals takes place.
1. Input: The input of Flowify AI® is a camera stream or a single image.
2. Object Detection & Segmentation: We detect and segment test cassettes and their components that appear in an image. In order to ensure, a high quality of the test cassette images, we employ various validation checks.
3. Regression: After the detection and segmentation, we employ a regression model that infers the signal value for each line of an LFA. The regression model is trained on a very large and continuously growing dataset, which is based on more than 400k (as of 2022/12) test cassette images.
4. Concentration / Thresholds: The signal values can serve as input for arbitrary concentration functions. For qualitative and semi-quantitative tests, it is possible to set appropriate thresholds. It is also possible to set a specific threshold for the control line.
5. Test Result: The test result is provided for each test that appears in the image.
The first three steps are generic and should work with most types of tests out of the box. If a test uses an uncommon housing, the segmentation models can be easily adapted to the new tests. Only the fourth step does depend on the characteristics of a specific product (LFA).
Flowify® — Lab in your Pocket
Flowify AI® is also part of the Flowify® app that transforms the smartphone into a reader device for LFAs. Hence, it simplifies data measurement and documentation of LFA tests. Flowify® is a convenient solution for everyone looking for a mobile read-out for rapid tests as well as a platform that collects all testing details.
TL;DR
Flowify AI® is a solution for the quantitative evaluation of LFA tests. It is designed to run on mobile devices and enables real-time point-of-care testing. Out of the box, it is compatible with a wide variety of tests and can be adapted easily.
