AyrFlo team hopes to save lives with device that could predict respiratory distress

Guelay Bilen-Rosas, Humberto “Tito” Rosas and Irene Ong developed a noninvasive device to help patients before it's too late.
Irene Ong, Guelay Bilen-Rosas and Humberto “Tito” Rosas
Photo by Romulo Ueda
The AyrFlo team (left to right), Irene Ong, Guelay Bilen-Rosas and Humberto “Tito” Rosas, rents space inside Forward BIOLABS, a nonprofit coworking life sciences lab at University Research Park.

On an otherwise ordinary day in August 2019, three people held their breath in the corner of a University Hospital operating room. They were pediatric anesthesiologist Guelay Bilen-Rosas; her husband, radiologist Humberto “Tito” Rosas; and Irene Ong, a Ph.D. data scientist and family friend. On the operating table, a patient, sedated for an unrelated routine procedure, had consented to participate in the trio’s study. Taped to the patient’s neck was a sensor from a modified ultrasound machine. If at any point his airway became compromised — if he had a bad reaction to sedation medication, for example — the three colleagues hypothesized they’d be able to see it on their computer screen before any of the standard machines, such as a pulse oximeter, alerted the health care providers that something was wrong.

And then it happened.

“That person isn’t breathing, right?” Ong whispered to Bilen-Rosas and Rosas, who stared at the signal on the monitor that only they could see. They looked at the patient, who seemed fine. They looked at the caregivers, who were calmly going about their procedure.

“He isn’t breathing,” Rosas confirmed.

More than 60 seconds later, the pulse oximeter alarm went off.

As mandated by protocol, the team was forced to keep their poker faces. But when those alarms finally sounded, they let out a breath themselves — one they’d been holding for years. They were finally collecting data, and it seemed to support the prototype. In a situation where every second counts and sedation medications could be reversed with enough warning, the implications were thrilling.

Rosas also felt vindicated on behalf of his wife, who sometimes came home from work in tears. She’d been the one who’d gone looking for a solution to this problem in the first place. The one who kept trying after everyone told her no, that the machines they already had were good enough. In that moment, he couldn’t help himself; he pictured the high-ranking colleague who’d told Bilen-Rosas that she didn’t have the chops, that she wasn’t smart enough.

“There’s still quite significant sexual discrimination in health care,” Bilen-Rosas says. “I think the lesson learned for all of us, you know, having come from an immigrant background, you don’t have to buy into what they think.”

Still, that’s part of why Rosas took a 20% pay cut to devote Fridays to supporting his wife’s side project. It’s why he’d been picking up the extra slack at home — the cooking, cleaning and general shuttling of their two kids. He knew better than anyone what she was capable of. Sure, he wanted to help produce a potentially lifesaving intervention. He also wanted her to prove them all wrong.


Rosas and Bilen-Rosas first met almost 23 years ago in Boston. Walking home late at night from rotations at Brigham and Women’s Hospital, they bonded over their surprisingly similar backgrounds. She was born in Turkey before her family immigrated to Germany as refugees. His mom was born in Tangier, Morocco, and his dad in Mexico. Both came to America because of the tremendous sacrifices of their parents, none of whom were able to go to college. “We are the vision of our parents,” Bilen-Rosas says. “We became, because they couldn’t.” Ten months after they met, they got engaged, then married.

Years later, after Rosas came to Madison in 2007 and Bilen-Rosas followed in 2008 to work at University Hospital and raise two kids, Bilen-Rosas, now double-boarded in pediatric and adult anesthesiology, had grown frustrated by a phenomenon she felt helpless to control. It was something that Rosas saw in his radiology practice as well, as a proceduralist using conscious sedation: Patients sedated without intubation, or post-operative patients who’d had their breathing tube removed, sometimes took a turn without warning.

“Our currently available monitors are not breathing monitors,” Bilen-Rosas explains. “They show changes after a patient is falling off the cliff. So we’re reacting to resuscitating this event, rather than proactively intervening so it doesn’t get there.”

graphic of a doctor helping a patient

Sources Getty Images

Moreover, besides downstream measures like the pulse oximeter, providers can only rely on subjective tools like instinct and experience — which they felt left too much room for human error or bias. They needed an objective measure — a common language all health care providers could speak, especially the younger ones who came through their teaching hospital.

“Guelay used to come home and say, ‘I have all these years of training and I can see things happening almost before they happen, but I don’t know how to teach that to someone else. If I could just find something that would be able to detect this early respiratory compromise, I would buy it,’ ” Rosas says. But there was no such device on the market.

Bilen-Rosas couldn’t let it go. For about six months in 2015, the couple brainstormed, dismissing dozens of ideas. One October morning, while cooking pancakes, Rosas brought up the possibility of using ultrasound. He was only half-serious — but his wife was immediately hooked. “That’s it!” she yelled. “That’s it.”

Bilen-Rosas bought an old ultrasound machine and they began tinkering. She knew it had to be noninvasive; she’d seen too many toddler-aged patients try to rip every tube or device from their bodies. But as she sought buy-in from others, she only hit more roadblocks. Colleagues were busy, often skeptical. Some helped where they could — in particular a veterinary anesthesiologist at the University of Wisconsin Veterinary School who connected her and Rosas with cadaver dogs on which they rigged a foley catheter to mimic different states of airflow — but most couldn’t or wouldn’t.

“People are on tenure track, people have their own labs, and this was something so outside,” Bilen-Rosas says. “We started describing a novel approach of quantifying airflow … so it was a very outlandish way of thinking. It’s never been done before.”

The couple even practiced on each other, and the anecdotal evidence began to pile up — but what they really needed was true data.

They filed their first patent in March 2016. On Oct. 11 — the birthday of Rosas’ father, who had just died — they learned they’d won the Wisconsin Alumni Research Foundation’s 2016 Innovation Award. They were the first medical school team to ever do so and, at the time, the first female-led group. It was enough to buoy their faith, but they still needed help.

Then in January 2017, Bilen-Rosas scheduled lunch at La Brioche True Food with Ong, whose two kids went to school with hers. The busy moms had been casual acquaintances for years, attracted by shared backgrounds: Ong, too, was an immigrant. Born in Malaysia, her single-parent mother brought her to New York when Ong was just 13, then sacrificed to send her to college. Sitting across from Bilen-Rosas that day, Ong innocently asked how her friend was doing. “I’m fine,” Bilen-Rosas said, then burst into tears.

She told Ong all about her frustration at work — how much potential there was. What she really needed, she said, was data so she could test her hypothesis. Ong was moved. She was also astonished. Reaching for her hand, Ong realized that Bilen-Rosas didn’t understand what she did professionally. “I work with data,” said Ong, who is now a faculty member of the University of Wisconsin Carbone Cancer Center’s Cancer Informatics Shared Resource. “I can help you.”


From there, things moved swiftly. Ong brought on a graduate student to join the all-volunteer team. In 2017, they won WARF Accelerator Phase I Funding , making Bilen-Rosas and Ong the first women from the medical school to ever win. In July 2019, they received a Phase II award from WARF’s accelerator program. This allowed them to take their prototype to sedation patients — and into the operating room that fateful day. Sixteen more patients followed — then COVID-19 hit, effectively halting the case studies.

Faced with the forced “downtime” of the pandemic (still navigating virtual schooling with their kids, still working hands-on with hospital patients during a global health crisis, still putting all their extra energy into this project on a volunteer basis), they took advantage of the “break” to finally analyze the data they’d been collecting, to look closer at the machine learning potential and determine whether they could develop predictive models. For example, if they could stretch that 60- to 90-second head start on detecting a patient’s breathing constriction to three or four minutes, that could be enough time for a sedation medication to be reversed, averting the crisis. It also gave them time to officially form a company — the only way their theories would have the chance to save patients in practice.

In May 2020, they incorporated as AyrFlo Inc. “A” for their son’s name, “Y” for their daughter’s name, “R” for Rosas — and, of course, the wordplay on “airflow.” In September 2020, they rented a bench at life sciences coworking space Forward BIOLABS. They filed two more patents during the pandemic and earned a spot in the competitive National Science Foundation I-Corps entrepreneurship training program. They were able to conduct market research, putting them in touch with health care providers across the country who alerted them to the additional implications this technology could have for things like asthma and other chronic breathing conditions, including sleep apnea and complications faced by so-called COVID-19 “long-haulers.”

“The fact is, the use case is for anybody who is walking and breathing,” Bilen-Rosas says.

In August 2021, they were able to resume data collection in the operating room. Now they’re in the process of finding manufacturing collaborators and making their first hires: data scientists and engineers. They’ll wrap up academic data collection funded by the Phase II award within the next few months; parallel to that, they’ll work toward creating their first custom-manufactured, miniaturized, self-adhesive prototype with signals they can manipulate. Then the team will start the regulatory process toward FDA approval. This can take years, but they’re hoping for a “breakthrough designation,” which would speed the process. They don’t mind. They want it to be the best product it can possibly be.

“These are people, so we’re gonna do it right,” Bilen-Rosas says, not only of the future patients, or the parents of those children she still mourns, but of their caregivers as well. “Doctors, nurses, technicians — those are the people we need to empower.”

Above all, they don’t want to get ahead of themselves. They’re scientists and they don’t want to overstate their claims. But they’re also human.

“If we can actually save one person,” Bilen-Rosas says, through tears, “if I’m an old lady with lots of gray hair and somebody is gonna be saved, that’s gonna be worth it.”

Until then, the AyrFlo team will take it one data point at a time, keep putting one foot in front of the other — deep breaths in, deep breaths out.

Maggie Ginsberg is an associate editor of Madison Magazine.