Machine Learning-Driven and Smartphone-Based Fluorescence Detection of Covid-19 infection co-developed by Sertis

It’s been over 2 years since we were hit by Covid-19 and it isn’t going away any time soon. There have been a number of medical development technologies so far, however, one of the ideal technologies to carry out Covid-19 infection tests is Real Time PCR detection, which is expensive, laborious and requires costly equipment and highly trained personnel. Many developing countries and health care settings have limited access to such resources, restricting their capacity to sufficiently test their populations.

The Covid-19 pandemic is the worst crisis since World War II as the virus has threatened every nation by infecting over 179 million people and killing over 3.8 millions around the world. The virus can spread and transmit via an air, even individuals with asymptomatic infections, can be carriers of the virus and transmit it to others. Therefore, being able to extensively test individuals and quarantine those shown to be infected early on in the disease progression is crucial to slow the rate of transmission.

A novel CRISPR-based diagnostic method is an attractive option due to its rapid, sensitive, specific, and easy to implement characteristics but the disadvantage is that its readout mostly relies on a simple visual inspection, which is prone to errors, particularly for discriminating samples with low viral counts. However, this quantification can be performed using fluorescence intensity readers which is typically costly and inaccessible to many health professionals.

To solve that problem to achieve rapid and sensitive testing, a team of researchers, including Aubin Samacoits, Senior AI Researcher at Sertis, jointly researched and developed an AI based Computer Vision application coupled with machine learning-driven software on smartphones designed to help researchers detect fluorescence intensity in medical experiments and tests through the CRISPR-based diagnostic platform.

The custom software in a smartphone is equipped with a binary classification model that has been developed by the researcher team, to quantify the acquired fluorescence images and determine the presence of the virus from laboratory samples. Then using Computer Vision technology to indicate the fluorescence intensity of the liquid, and estimate the possibility of infection that reporting results will be shown in the application immediately.

The result shows that the method is almost as accurate as Real Time PCR detection, and the system can also detect Viral Load to offer valuable information of the infection in individuals that cannot be obtained from current readout methods.

Read a full research at

Leading big data and AI-powered solution company