Frequency comb breathalyzer can "accurately detect COVID-19" with machine learning

Researchers Jun Ye and David Nesbitt and their team at JILA have been working on an exciting new diagnostic tool that could revolutionize healthcare. JILA was created in 1962 as a research lab jointly operated by NIST, the National Institute of Standards and Technology (which is part of the U.S. Department of Commerce), and the University of Colorado, Boulder. JILA used to be an acronym for "Joint Institute for Laboratory Astrophysics" but currently just goes by JILA (no acronym).

The project involves using a frequency comb breathalyzer to scan human breath—which contains more than 1,000 different trace molecules—for markers of specific health conditions, including COVID-19. Frequency combs measure different colors of light, and are able to identify chemical signatures of different molecules based on color and on how much or how little infrared light is absorbed or reflected. Ye and Nesbitt collected breath samples from 170 students and staff at University of Colorado, Boulder, and used that data to determine whether and with what accuracy the frequency comb could detect COVID-19. Turns out the machine made COVID-19 predictions with pretty high accuracy—about 85 percent. But the machine's potential goes well beyond detecting COVID-19. succinctly explains the project:

NIST/JILA fellows Jun Ye, David Nesbitt and their colleagues have demonstrated that a breathalyzer based on Nobel Prize-winning frequency-comb technology combined with machine learning techniques can accurately detect SARS-CoV-2 infection in human breath. The frequency-comb technology, recognized in the 2005 Nobel Prize in physics, has the potential to detect many other diseases, such as COPD, lung cancer, and kidney failure. Frequency combs can precisely measure different colors of light, including the infrared light absorbed by molecules in a person's breath. Combined with machine learning, this technique can detect the presence of specific combinations of molecules that are signatures of disease.

In the first video featured in the NIST article, Nesbitt and Ye provide more context and information about their research. Here's part of that video that I transcribed:

David Nesbitt, Chemical Physicist, NIST & JILA Fellow: "Molecules are amazing things. They're sort of like atoms held together sort of by springs—springs that allow the atoms to move and vibrate. And molecular spectroscopy, as used by Jun and myself here, are basically looking at the frequencies of the motions of these atoms in a molecule."

Jun Ye, Physicist, NIST & JILA Fellow: "Atoms and molecules—they emit light, they absorb light. In fact, light is the tool we use to communicate with the microscopic world of atoms and molecules. We had developed a tool called Optical Frequency Comb, for the sake of optical atomic clocks, for time keeping. We can use this to measure molecules—and wouldn't it be funny if we could use this to smell people's breath and tell the health conditions of a particular person?"

Nesbitt: "What we've been doing is looking at something on the order of 150 COVID positive and COVID negative participants in this study, and looking at their breath samples, and looking at allll the different frequencies that are absorbed or not absorbed, using this frequency comb."

Ye: "I can collect hundreds of thousands of detection channels, all at once. That's why we can detect many different kinds of molecules. Each molecule leaves a bunch of different fingerprints, and you can just count how many fingerprints you have there. You can have really unambiguous identification of the molecules. But before the machine learning tool was introduced, I really wasn't sure how we were going to process all that data."

They ended up harnessing the power of machine learning to process and analyze the data they collected. The results of their research could be revolutionary for the healthcare industry, as the frequency comb breathalyzer could potentially provide a quick, accurate, easy, non-invasive diagnostic tool to detect many more health conditions than other breath-based analysis tools. NIST explains that "Efforts are already under way to miniaturize and simplify the technology to make it portable and easy to use in hospitals and other care settings." To read more about the project and watch several videos explaining more, check out this article on the NIST website.