Ohio State engineers combine ultrasound and AI for faster COVID-19 detection

When Professor Alper Yilmaz learned about the need for better, faster COVID-19 diagnostic tools for the masses, he jumped into action.

Working since mid-March, the Ohio State civil, environmental and geodetic engineering professor and PhD student Shehan Perera developed technology that combines ultrasound scans of the lung with deep learning technology to identify COVID-19. Diagnosis takes less than 10 minutes, including scanning time, and sanitization is as simple as changing a plastic sleeve that covers the device.

“My lab typically works on machine learning and artificial intelligence problems in a number of domains, including medicine, nuclear engineering, navigation and visual surveillance,” Yilmaz said. “So I said, let's use that expertise and some of the results and new technologies we have to foster new ideas to solve this problem.”

Turnaround times and accessibility can be an issue with existing COVID-19 diagnostic tests, he explained.

The most popular COVID-19 diagnostic test—the PCR test—uses a swab to obtain a mucus sample, typically from the nose or throat. But getting results can take four to six hours or longer when mass quantities are involved.

Another option for diagnosing COVID-19 in sick patients is via a CT scan of the lungs, Yilmaz said. While results can be read in minutes by a trained radiologist, the acquisition times are long and the machines are extremely expensive. Thus, CT scanners are not available everywhere, particularly in rural hospitals. Plus, sanitizing the machine and the room after each use can take over an hour.

Unlike CT, ultrasound devices are more widely available and portable, allowing the engineers’ solution to be used in nursing homes and home care settings.

Since COVID-19 is not as visible on ultrasound, experts have to be trained in how to spot it, Yilmaz explained. So the researchers built a large library of COVID-19 results with data mined from the web and created a novel artificial intelligence architecture capable of diagnosing COVID-19 or helping medical professionals make a diagnosis.

The ultrasound diagnostic technology has roots in one another of Yilmaz’s current research projects, a tool that uses CT to help doctors decide if a patient has had a stroke.

“In the end, we came up with a pretty accurate tool,” he said. “Over time the model has evolved and become more complicated, in the sense that it started seeing those [COVID-19] signatures much more clearly. We now hit about 100% accuracy on data that the system hasn't seen before.”

Yilmaz is working to commercialize the technology. He launched a spinoff company through Rev1 Ventures, which is a startup incubator in Columbus. The researchers’ beta product runs on a personal computer and they are converting it to be compatible with mobile platforms. It could also be integrated into existing ultrasound machines.

While interest in the technology is high, obtaining full FDA approval could take six to 12 months, Yilmaz said. Although he noted there are COVID-19 tests currently in use that haven’t completed the full approval process.

This is not Yilmaz’s first entrepreneurial effort. He launched startup Ubihere two years ago to commercialize his tracking and localization technology, which will soon be installed in several hospital systems to help locate critical resources like ventilators in real-time.

“There are available tracking technologies, but they do not exactly work the way hospitals want them to work,” Yilmaz said. “A lot of things get lost in the hospital system. During this pandemic we are seeing the necessity of having this kind of system to actually find all the tools that they need at the right time.”

In addition to helping during the current crisis, the COVID-19 diagnostic project is special to Yilmaz because his student has been involved in every step of the process, including commercialization.

“This is what I love about being an academician, because students also want to learn how to move forward in the startup environment,” Yilmaz said. “Shehan is really interested in learning, maybe he'll start his own companies in the future. When you see you can help them, it makes things fun. It's great to have him as a part of it.”

by Candi Clevenger, College of Engineering Communications, clevenger.87@osu.edu