Bio-Stream Diagnostics uses machine learning to design faster, better COVID tests

Two years into the coronavirus pandemic, the world is still struggling to safely navigate back to normal. Many workplaces have had to find ways to protect their employees from the spread of the virus, and rapid COVID testing has been a big part of that.

For the past two years, an Edmonton-based biotech company Bio-Stream Diagnostics has been working on a new approach to COVID testing. By combining advanced biosensors with artificial intelligence, their work has the potential to be faster, cheaper and more accurate than current rapid tests.

“Companies are trying to kind of get back to some semblance of normal. But, to do that, they need to be able to actually have their employees feel safe to be productive,” says John Murphy, co-founder and CEO of Bio-Stream.

Many companies already use extensive testing to prevent coronavirus outbreaks. However, current tests have their limitations. Polymerase chain reaction (PCR) tests, which look for the virus’s genetic material, have high accuracy. But they require nasal or throat swabs, which aren’t exactly pleasant. And those samples usually have to be sent to a centralized lab for analysis. As such, results can take days.

Rapid antigen tests are faster but much less accurate. Instead of genetic material, they detect the proteins that make up the coronavirus. This method does carry the risk of false positives. And if someone is in the early stages of infection, there might not be enough protein to show up on the test.

If people are going to feel safe in their workplaces again, Murphy says, another solution is needed.

“It had to be really accurate, and then it had to actually be very low cost, and it had to be really easy to take, convenient and comfortable,” he says.

Viral voltage

Bio-Stream instead went with a testing device that uses organic electrochemical transistors. These transistors can capture biochemical information from a sample, like saliva, and translate it into electrical signals.

The test is made up of a handheld reader and disposable test strips, much like a blood sugar monitor. Murphy says a person would put saliva on the strip and then insert it into a reader, which then measures the electrical signal of the sample.

A sample with the coronavirus will have a particular change in voltage, which the reader uses to give a positive or negative result within a matter of minutes. That speed is necessary for organizations that rely on testing to prevent outbreaks.

“Usually it’s 20 minutes, maybe an hour before you get results, maybe a lot more. That introduces just a lot of awkwardness in the flow of how people really work in an organization. So we saw that as an opportunity,” he says.

Cutting through the noise

When Bio-Stream first developed the testing platform, it had a problem. The data was noisy, and it was difficult to detect the signals that pointed towards a COVID infection. Machine learning allowed the company’s researchers to analyze the results to ignore the irrelevant data, making the test more reliable and accurate.

“It was just invaluable for us. AI to really be able to iterate quickly through research data and through the test data, which allows us to actually create a usable product,” he says.

The company’s initial results are promising — based on their in-house testing, Biosteam’s voltage-based system has accuracy rivalling that of PCR tests, without the need to send the samples off to a lab. However, Murphy stresses that clinical trials are still needed to confirm those results and are currently underway. Still, he is confident that the platform will prove better than current rapid antigen tests.

In addition to improving accuracy, Murphy says artificial intelligence could examine the data the machines record while testing. The metadata is anonymous, he says, and wouldn’t link individual people to their results. But it could allow a company to spot trends that could inform their COVID policies and warn them about possible outbreaks in their workplaces.

The voltage testing method isn’t necessarily limited to detecting COVID-19. Murphy says adjustments to the test strips and readers could allow it to potentially detect other infectious diseases, like tuberculosis or influenza. They are also investigating whether the approach could be used in the future to detect certain types of cancer.

Artificial intelligence is nothing new for Bio-Stream; the company is a spin-off of Stream.ML, which offers a platform for businesses to build machine learning models. Still, even with that previous expertise, the company knew that collaboration would allow them to develop their testing method faster. So they partnered with the Prairies Canada Regional Innovation Ecosystems (RIE), a program that provided support to Alberta companies using AI to develop innovations in healthcare and wellness.

Since the need for reliable COVID testing is so great, Bio-Stream knew it would have to act fast. Without the support of the ecosystems project, Murphy says it would have been difficult to find the talent they needed to make rapid progress.

The company is now working towards clinical trials for its testing platform. It is also in talks with manufacturers to build the readers and test strips.

This project was part of the Western Economic Diversification Canada Regional Innovation Ecosystems (RIE) program. The initiative brought together nine organizations from non-profit, business and academia to establish viable uses for artificial intelligence and machine learning in health and data analytics.

Learn more about how artificial intelligence and machine intelligence can transform your business. Discover 5 ways AI can be your business’ competitive advantage.

Let's get started!

Interested in learning more about how Amii can help you harness the potential of AI in your business?

The first step is easy: contact us! We're here if you want to float an idea or to answer any burning questions you might have. We're always happy to chat with a fellow AI enthusiast (it's absolutely our favourite topic of discussion).