Machine Learning-Assisted Lateral Flow Assay for Detecting COVID-19 and Influenza

In the pandemic and endemic era, lateral flow assay (LFA) is a great candidate for point-of-care testing, and especially as a tool for companion diagnostics. For example, fast detection and immediate treatment are key factors in companion diagnostics for oral antiviral treatments, such as the use of nirmatrelvir/ritonavir (Paxlovid) and oseltamivir (Tamiflu) for treating coronavirus disease (COVID-19) and influenza, respectively. However, several limitations are exhibited when untrained individuals (i.e., during self-testing) analyze the results with the naked eye. Herein, we propose a method to achieve enhanced assay performance via the use of machine learning (ML)-assisted smartphone image processing and ML-assisted LFA readers. The use of ML-assisted LFA readers demonstrated a seven-fold increase in the sensitivity slope when detecting influenza A virus with enhanced R2 values (0.90). Moreover, when using smartphone-based digital images, we achieved an eight-fold enhancement in the sensitivity slope when detecting severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) with good reliability (R2 values of ~0.96). Further, with the ML-assisted LFA assay, we acquired up to a 12-fold enhanced colorimetric signal (average ~ six-fold) using real patient samples from COVID-19 positive patients, leading to point-of-care-testing (POCT) using smartphone without the use of any external attachments, especially for the low-income countries and self-testing by untrained individuals.