Machine Learning and a Low-Cost Nanosensor Could Beat PCR, Rapid Antigen for COVID-19 Testing
Spectroscopy-based sensing device could be used "on any surface" or even integrated directly into a wearable platform.
A team of engineers at Johns Hopkins University, working with the support of the US National Science Foundation, has developed a new sensor for COVID-19 infections, which addresses limitations in current PCR and rapid antigen test approaches — and that could even be used in wearables.
Polymerase chain reaction (PCR) tests are recognized as one of the most accurate and sensitive means of detecting an active COVID-19 infection in an individual, but have an uncomfortable sample retrieval process and require said samples to be submitted to a lab for processing. Rapid antigen tests do away with the need for a lab, offering at-home personal testing, but the results are less accurate.
The sensor developed at Johns Hopkins, the engineers behind it claim, offers close to the sensitivity of a PCR test with the convenience of rapid antigen testing. "The technique is as simple as putting a drop of saliva on our device and getting a negative or a positive result," says Ishan Barman, one of the senior authors of the paper detailing the work.
"The key novelty is that this is a label-free technique, which means that no additional chemical modifications like molecular labeling or antibody functionalization are required. The sensor could eventually be used in wearable devices."
The sensor is based on a machine learning system fed data from surface-enhanced Rama spectroscopy on metal-insulator-metal nanostructures as the sensor itself. Results are provided within 25 minutes from a simple saliva sample, and the test can distinguish between a number of different respiratory and non-respiratory viruses. Better still, its creators claim it could be produced as a "relatively inexpensive mass testing technology" — though stop short of putting a dollar value to that.
"Label-free optical detection, combined with machine learning, allows us to have a single platform that can test for a wide range of viruses with enhanced sensitivity and selectivity," lead author Debadrita Paria claims, "with a very fast turnaround."
"We can use this for broad testing against different viruses, for instance, to differentiate between SARS-CoV-2 and H1N1, and even variants," adds Barman. "This is a major issue that can't be readily addressed by current rapid tests."
The team's work, which is undergoing further development and testing ahead of commercialization, has been published under open-access terms in the journal Nano Letters.