fb-pixel
Dispatches from the Edge

Can AI predict a coronavirus infection before doctors even do a test?

Building a platform to analyze millions of data points can take months. This Boston startup did it in three weeks.

Kuldeep Singh Rajput, the founder and CEO of Boston startup Biofourmis.
Kuldeep Singh Rajput, the founder and CEO of Boston startup Biofourmis.

Editor’s Note: This story is part of a Globe Magazine special report, appearing in print on Sunday, March 29.

It was nearly midnight on Valentine’s Day and Kuldeep Singh Rajput, the 28-year-old founder and CEO of Boston startup Biofourmis, was catching up on e-mails when the phone rang. On the other end of the line were four professors and clinicians from the University of Hong Kong, where Biofourmis’s wearable technology and software is used to remotely analyze heart failure patients. The unscheduled call wasn’t cardiac related.

Instead, Singh Rajput found himself talking with infectious disease experts, including doctors who lived through the 2003 SARS epidemic. They had an urgent question: Could the Biofourmis system be used to help stem the burgeoning global coronavirus crisis?

Advertisement



“We essentially are able to detect these subtle physiology changes, precursors to symptoms,” Singh Rajput says. “And specifically in this case, things like shortness of breath, fever, or cough, et cetera.” The company’s artificial intelligence software finds patterns in the data and makes it available to clinicians in an understandable fashion, then helps them intervene remotely, including through phone or video calls.

The hope, he explains, is to be able to detect COVID-19 infection markers in two or three days — less than the median five days it takes for infected people to show symptoms. That means potentially infected people could be tested earlier, and medical professionals could help treat some patients remotely, lessening their own chance of infection.

The Hong Kong group on the phone – which included doctors David Chung Wah Siu, Kwok-yung Yuen, and Harmony Medical’s CEO, Dr. Raymond Tong — had a two-page proposal ready to go. Singh Rajput, who was at home in the Seaport, hung up and got to work. “I reached out to our teams in India and Singapore, where most of our developers are,” he says. “We were on the phone until 2 a.m. Eastern Time. We mapped out all the requirements. And the team started working.”

Advertisement



Over the next three days, 14 Biofourmis employees on three continents worked nonstop to modify the system, creating a mock-up design. By day seven, “we had a full, functional platform,” Singh Rajput says. By the end of the second week, they had the app ready to test the system with patients and clinicians.

Hong Kong had begun compulsory 14-day quarantines on February 8. That meant Biofourmis employees couldn’t go there without risk. Instead, they figured out a way to develop, ship, and test the system remotely. Typically, just building the app for such a system could take two to three months. On March 4, exactly three weeks after the midnight call, Biofourmis announced its COVID-19 system would be tested at the University of Hong Kong.

With the new system designed, it was time for the Hong Kong clinicians to find people already quarantined to test it. About 100 to 200 patients, with and without symptoms, have started wearing the Biofourmis sensor-based armbands as part of the study, says Dr. Fan Ngai Ivan Hung, another University of Hong Kong clinician involved in the project. (They aim to test it with as many as 1,000 patients.)

Wearable monitors send patient data automatically through an app to the cloud for secure processing. Patients also self-report symptoms. Every four hours, clinicians check the data on their own systems, from which they can instantly order a COVID-19 test. Clinicians also learn more about disease progression through patterns highlighted by the software.

Advertisement



Biofourmis’s system collects 2 million data points per patient per day and feeds it into the AI system, Singh Rajput says. “Given the massive amount of data, the pattern recognition that is needed would be far beyond human physician experiences,” explains Siu. “We plan to monitor and collect data from as many patients as possible to help us to build predictive models and also learn more about the infection.”

Biofourmis’s technology for heart failure, approved by the Food and Drug Administration last year, is already being tested by Boston’s Brigham and Women’s Home Hospital program and at the Yale University-Mayo Clinic, among other places. Biofourmis also has partnerships with companies like Novartis.

Now, Biofourmis is close to announcing agreements with US health systems for the COVID-19 version, Singh Rajput says. Because the system monitors and analyzes physiological changes similar to the heart projects, it doesn’t need new FDA approval, he adds. That could mean more people exposed to the virus in the US get tested sooner, hospital workers can safely monitor more patients remotely, and drug companies can speed development.

It was during a year working at MIT’s Media Lab that Singh Rajput became familiar with Boston’s talent pool: specialized data scientists who focus on clinical information. Last year he moved the company from Singapore to Arch Street in downtown Boston. Some 30 of the company’s 120 employees now work there.

Advertisement



“Everyone comes to work with a mission,” he says. “This was a very special project, and everyone is proud of it. Because right now, we have patients in Hong Kong being monitored and managed. That’s a great feeling.”

____________

Rachel Layne is a Boston-based freelance journalist who has written for CBS MoneyWatch, USA Today, and Bloomberg. Send comments to magazine@globe.com.