Many of us have seen or been subjected to a COVID-19 test. Frequent screening, like the pandemic itself, has become a way of life. As SARS-CoV-2 remains a powerful foe, our strategies to identify and classify the virus must stay agile and smart.
A new study, published by the Enter Beckman researcher Gabriel Popescu, a UIUC professor of electrical and computer engineering, and his interdisciplinary team in the Light: Science and Applications-Nature, have paired microscopy with AI to develop a COVID-19 test that’s fast, accurate, and cost-effective.
The researchers’ initial step, as is typical of a Beckman team, was to identify an opportunity for innovation; they saw that, while several approaches exist to test for SARS-CoV-2, none use a label-free optical approach.
Even with a microscope, the tiny size of a single particle makes depending on sight alone nearly impossible. Electron microscopy can be used to image the structure of a particle, but substantial preparation is required to assure the sample’s visibility. Though required, this method has the potential to obscure the target image.
Popescu’s team used a technique developed at Beckman for seeing cells: spatial light image microscopy, which allows for chemical-free (or label-free) imaging.
Although an electron microscope produces a crisp image, it necessitates substantial sample preparation. Using SLIM for virus imaging is analogous to staring at something without spectacles. Because the viruses are smaller than the diffraction limit, the image is hazy. However, because of the high sensitivity of SLIM, we can not only detect viruses, but also distinguish between different types, according to the study authors.
Fortunately, the researchers discovered a novel approach to identify viruses using SLIM data: artificial intelligence. An advanced deep neural network can be trained to recognise even the most blurry photos with the correct training.
They fed the AI programme two images: a labelled SARS-CoV-2 particle that fluoresced and a phase image acquired with a fluorescence-SLIM multimodal microscope. The AI was already utilized to understand these photographs as identical. The fluorescence-stained image, which is easily identifiable, acts as a training wheel; with enough repetition, the system begins to recognize viruses directly from the SLIM, label-free photos without the extra help.
Following detection comes differentiation: distinguishing SARS-CoV-2 from other viruses and particles.
The AI was taught to distinguish from other viral infections such as H1N1, or influenza A; HAdV, or adenovirus; and ZIKV, or Zika virus. The preclinical experiment was exceedingly effective, with a 96 percent detection and classification success rate for SARS-CoV-2.
While clinical validation is pending, researchers believe that a COVID-19 test using this technology would look like this: the subject would wear a face shield attached to a clear glass slide; they would then execute an activity that causes their breath to get permanently bonded to the slide (like reading a paragraph out loud). The slide, along with any particles attached, would be photographed and examined for the presence of viruses.
From a clinical perspective, the impact of such innovative diagnosing capabilities is pronounced.
This highly adaptive AI programming could help address future pandemics, not just COVID-19.
“We need fast detection of diseases,” said the study authors.
“Not only COVID, but others. We can and should put our efforts together, both in terms of optics and AI, to try and find out just how far we can go.”
Image Credit: Getty