This technology is a smartphone application with several deep convolutional neural networks for quality control and classification of images from lateral flow immunoassays (LFIAs).
Current methods for interpreting lateral flow immunoassays (LFIAs), such as COVID-19 rapid tests and pregnancy tests, involve the user assessing the results by eye. Although the results are usually readily observable, LFIAs may produce an ambiguous result, particularly if the analyte being probed is in low abundance. Therefore, there is a need for an unbiased and automated method to interpret LFIAs.
This technology is a smartphone application that uses machine learning to read and interpret LFIA data using the phone’s built-in camera. The pipeline consists of several deep convolutional neural networks for quality control and classification of LFIA images. The quality control network ensures the image is of high quality and the classification network outputs a diagnosis as either positive, negative, indeterminate, or invalid. This system can be easily adapted for quantification of other tests with visually represented results (i.e., pregnancy tests).
Training, validating, and testing of the network has been done on approximately 20,000 different LFIA images captured on various phone models, including iPhone X and Samsung Galaxy Note 10.
IR CU21157
Licensing Contact: Joan Martinez