Flannery Dean writes about technology for MaRS. CTVNews.ca has partnered with MaRS to highlight Canadian innovations in health care.
Salim Kandedi believes premature babies should get the same standard of neonatal intensive care regardless of where they’re born, whether it’s in a fully staffed Toronto hospital or a resource-strapped NGO in a remote northern community where health-care inequities often perpetuate maternal and fetal health crises.
Dwindling access to labour and delivery services in rural Northern Ontario means pregnant people are often forced to travel hours to get such care, added stress that, according to a 2024 report from the Northern Policy Institute, can increase the risk of pre-term birth by 130 per cent and is associated with longer stays in neonatal intensive care units (NICUs).
There are about 32,000 babies born prematurely in Canada each year. Those infants, and others born with health complications, need more than incubators to keep them warm; as Kandedi explains, their brains, hearts and lungs are still developing, which can make them uniquely vulnerable to things like respiratory disorders and neonatal encephalopathy. Neonatal care can play out among 30 different machines, including respirators, vitals monitoring and a full suite of sensors to keep them healthy. To provide that equipment for all 32,000 preterm patients, he says, “is impossible.”
If you ask Kandedi, inequitable access to neonatal care isn’t solely an economic issue — it’s also an engineering puzzle begging to be solved. Machines are expensive, often unwieldy and take up a lot of space. Not every hospital is equipped with the right machine or specialist, which means many infants need to be transferred to places that do have them at additional expense.
The Toronto-based biomedical engineer is working on a potential solution to address these compounding access and logistics issues. In 2024, he founded Corvita Biomedical, a biotech startup dedicated to developing the ARK incubator, a portable NICU that not only consolidates many of the mechanical and monitoring components that help keep preemies alive, but also integrates machine learning to assess the data those machines collect.
“A baby placed into an ARK incubator would receive all the care he needs within that one device,” Kandedi explains.
The very existence of the ARK Incubator reflects the value of expanding access. Back in 2002, Kandedi was himself a micro preemie, born at 23 weeks gestational age. Babies who arrive that early can spend months in hospital supported by specialized equipment including ventilators and feeding tubes as their organs develop.
Although recent strides in neonatal care have improved outcomes for micro preemies, survival rates can still hover around 50 to 60 per cent. (Pre-term births overall account for almost two-thirds of infant deaths in Canada.) Kandedi credits his own survival to the fact that he was able to access thorough neonatal intensive care in his home country of Gabon, a stroke of luck his mother couldn’t help but bring up when he started researching the ARK.
“She told me, ‘If someone has to do something about those babies, it’s you, because you’ve been there, and without those machines, you wouldn’t have survived,’” he says.
Here, Kandedi talks about how he found his way into biomedical engineering and where he thinks his AI-powered innovation may lead.
Not every preemie winds up designing portable NICUs. How did you come to this field?
Up until the age of 16, I wanted to be a pilot. I passed all the tests … except the vision test. That’s when I discovered I was colour blind. I was pretty sad I couldn’t be a pilot, but after that I thought, “What if I could create some kind of technology that would allow colour blind people to follow their dreams?” I looked into biomedical engineering. It was the best thing that ever happened to me that I didn’t pass that eye test.
Where did the idea for the ARK come from?
In 2024, my colleague Leah Ritchie, who’s worked in maternal and fetal medicine, told me her aunt, who runs an NGO in northern Manitoba, was having a hard time getting the medical devices she needed to help babies survive premature birth. She was complaining about the lack of NICU access for babies born far from major cities and asked if I could build something portable and easy to use.
I dove into researching the problem and became fascinated with the diagnostics, the life support and other biomedical engineering to improve care for infants. I reached out to a doctor at SickKids, who explained that they were also struggling to keep up with increasing demands for NICU care — including the demand for labour. Doctors in Vancouver said the same thing. You need 30 different machines to care for the most acute babies; those machines need four specialists. If you don’t have one of those machines or specialists, the baby has to be transferred to another hospital. I realized this isn’t just a problem for remote areas; it’s a technology issue.
Tell me a bit about the technology, then. What issues did you have to solve?
When babies are born prematurely, they lose heat quickly because they don’t have body fat, and if you don’t stabilize their temperature, they could die of hypothermia. Incubators keep babies warm at a steady temperature. But once you’ve stabilized the baby, temperature-wise, you need to provide respiratory support with a respirator. If the baby starts having other problems, you might need to change the ventilation mode to a CPAP. You could have about 30 machines hooked up at the baby’s bedside. A portable incubator isn’t enough; you need a portable NICU that can monitor for all those complications. The ARK consolidates those machines and sensors within the same module.
How did you determine which components the ARK needed to contain?
I looked into why premature babies die and found about 80 per cent of all fatalities were caused by six distinct conditions. I learned about how you diagnose those conditions, the devices you need, the tools and the processes that need to run. That’s how I reverse-engineered the design of the ARK Incubator.
The complete list is proprietary for now, but the families of devices are incubation, respiration (CPAP and ventilators), vitals monitoring — which includes EEG, ECG, and so on. It has the temperature control of a traditional incubator as well as blood pressure, bilirubin and cerebral oxygen level monitoring. It captures about 30 monitors and machines that, combined, would be worth about $300,000.
The ARK also includes artificial intelligence capabilities. Why was that essential to your design?
When people hear about AI, they think about generative AI — that’s the stuff you interact with on your phone. Generative AI is probabilistic: it guesses the next word based on data it’s been trained on. We built what you call a hybrid model — those are deterministic and probabilistic, and they’re the best ones to mimic the human mind.
AI is a primary component of the ARK. We automated all the scoring tools and systems that are run in a hospital manually and we computerized them using machine learning intelligence. Our model doesn’t guess the next step; it follows a standardized American Association of Pediatrics diagnostics process. Recently, our tool was able to diagnose a disease as well as a doctor using data only. That’s huge. Right now, we’re working with about 80 per cent accuracy. We want to push that further, but that moment really cemented our work.
When you say you want to push that further, to bring the accuracy up beyond 80 per cent, what kind of work does that entail?
Babies come in all shapes and colours and conditions. It’s a very heterogeneous population, and the technology has to be able to work perfectly with a baby born in India as well as a baby born in North America. What we’re training our models to do right now is: you can classify ethnic background or other features, and adjust the coefficient (a numerical value that informs the relative importance of a specific feature) so that it’s tailored to that specific baby. That’s where we want to go, because that’s what doctors do in real life. It will treat a baby born in Brazil very differently than a baby born in Denmark. Beyond the simple physiological aspect, you have to understand geographical concepts, specific diseases that only occur in certain areas of the world. So that’s the work happening right now to improve those models.
When the ARK Incubator captures data, how does that information get relayed to the health-care team?
A nurse would receive a detailed notification on her phone that the baby is showing early signs of X because of specific vitals. She’d walk to the machine and work with the massive user interface to see how that information was found and decide what to do. For example, if the machine indicates the baby has early signs of jaundice, she could click a button to turn on the phototherapy light.
What is the potential impact of this technology?
Health-care workers often complain about the burden of documentation, with some studies suggesting nurses can spend nearly 30 per cent of their time on paperwork alone. We are giving them that time back so they can focus on care. The second thing, and the thing that I’m most proud of, is that a baby born on Baffin Island will receive the same monitoring and early diagnostics as a baby born in New York or here in Toronto. And there’s the financial impact: hospitals can save millions of dollars in capital and operating expenses by improving the quality of their work.
What are your next steps?
We’ve built the AI tool. We’ve confirmed the demand from hospitals that are not only willing to research with us but want to fund this machine for their patients. The next milestone is to take the minimum viable product and build a clinical version that can run pilots in hospitals. We’re raising about $2 million this fall to fund that development and start pilot testing in September 2027.
We’re planning on starting pilots in southern Ontario for urban centres, and northern Canada for rural centres. We also have partners in Zambia, Gabon, Guyana and Panama. They’ll start simultaneously and run for three months. In total, we’re hoping to pilot between 50 and 200 babies. With the data from those pilots, we’ll make any necessary improvements, and then we’ll present the data to Health Canada and the FDA (U.S. Food and Drug Administration) and launch the product.
