Backed by Yoshua Bengio, Ubenwa is using AI to diagnose neurological issues through baby cries

Montréal’s Ubenwa foresees a world where the cries of a newborn baby can predict the presence of neurological conditions.

It’s a world that is not far off. The Mila spin-out has been analyzing baby cries for the last few years and claims to have a high accuracy rate of detecting a condition that can often result in brain damage.

Using machine learning, Ubenwa – which means the cry of a child in the Igbo language – claims it can decipher baby cries to detect and diagnose medical conditions that have traditionally been difficult to diagnose. With a particular focus on neurological conditions caused by perinatal asphyxia (a lack of oxygen before, during, or after birth), the company is gearing itself up in preparation for seeking U.S. Food and Drug Administration (FDA) approval, and is also launching an app to help parents decipher their baby’s cries.

“It’s about creating a world in which cry analysis becomes a standardized assessment that everybody receives.”
– Charles Onu, Ubenwa

Ubenwa co-founder Charles Onu told BetaKit in an interview that the goal of the company is to create a world where baby cries are no longer viewed as “noisy sounds that we are trying to quiet,” but as ways to better understand health.

The five-year-old company has piqued the interest of leading machine-learning expert Yoshua Bengio, who has served as an advisor for Ubwena over the past three years, and recently became an investor in the startup’s pre-seed funding round.

“Ubenwa’s AI technology has the potential to save the lives of newborns and significantly support the medical and research communities, demonstrating the importance of responsible AI innovation for all,” Yoshua Bengio, Mila’s founder and scientific director, said in a statement to BetaKit.

The $3.24 million CAD ($2.5M USD) round includes a spattering of artificial intelligence leaders. It was led by Radical Ventures and saw participation from AIX Ventures, which is a collective of experienced AI stakeholders, as well as Hugo Larochelle, the Montréal lead for Google Brain, and Marc Bellemare, who also works with Google Brain (and previously DeepMind).

The all-equity, pre-seed round marks Ubenwa’s first institutional capital to date. Previously, the startup funded its research and development through a collection of grants, including taking home two prizes from the Massachusetts Institute of Technology (MIT) in 2020. Ubenwa also made it through a couple of rounds of IBM’s Watson AI XPrize, which saw Montréal’s Aifred Health take home second place.

Ubenwa is led by Onu, who first heard about perinatal asphyxia while volunteering for the global non-profit organization Enactus in his home country of Nigeria.

A computer engineer by trade, but with a passion for medicine, Onu was taken by the health complications that these newborn babies were facing. The idea stayed with Onu as he moved to Montréal to study for his Master’s degree and Ph.D. in computer science at McGill University.

It was there that he met Innocent Udeogu and Samantha Latremouille, and the trio launched Ubenwa.

Analyzing the pitch of baby cries is not a new concept. In an interview with BetaKit, Onu noted that studies on cry sounds took place in the 1970s and ’80s that found a correlation between the pitch of a cry and the central nervous system. In more recent years, machine learning has made it easier to analyze large sets of data, and there has been a smattering of studies on how cry acoustics can point to medical conditions.

UCLA’s health group launched an app in 2018 called ChatterBaby that helps parents to decipher their baby’s cries. Data collected on the app was used in a study that claimed 90.7 percent accuracy for identifying pain cries, and 71.5 percent accuracy in discriminating between cries related to fussiness, hunger, or pain.

Ubenwa is taking a more specific approach, hoping to help identify neurological conditions as early as when babies are born. The startup touts itself as the first technology for rapid detection of neurological conditions in infants using only their cry sounds.

“It’s about creating a world in which cry analysis becomes a standardized assessment that everybody receives everywhere they’re born almost, as de facto as blood pressure,” said Onu. That kind of test could allow for what Onu called “a cry stamp,” essentially giving medical practitioners or parents a baseline for what their child’s cries mean.

Radical Ventures investor Sanjana Basu argued that Ubenwa’s focus on the pediatric space is what sets the company apart from others studying speech biomarkers.

“The pediatric market has been traditionally underserved but is seeing rapid growth,” she said in a statement to BetaKit, noting that Radical Ventres was attracted to Ubenwa because its technology is built on “a unique and diverse database” of infant cry sounds, and based on research developed in collaboration with MILA, the Montréal Children’s Hospital, and pediatric hospitals in South America and Africa.

Over the last two years, a select group of doctors in six hospitals spanning Canada, Nigeria, and Brazil have used Ubenwa’s medical app as part of a clinical study. The doctors collect and record the various cry sounds of babies with neurological conditions and those without, with Ubenwa using machine learning to analyze what they mean. Onu said the data has shown that Ubenwa can detect asphyxia with an 88 percent accuracy rate.

Ubenwa had originally focused on creating a solution for the medical community, but Onu noted that through conversations with medical and nonmedical individuals many people assumed the company was also developing an app for parents.

That led Onu and his founders to see a gap in the market, and the trio decided to split their focus between a medical app and a consumer one.

Now, the startup is looking to launch its app for parents later this year or in early 2022. The app is currently available for parents to sign up to join a beta test. The consumer app gives Ubenwa a quicker way to market as the startup has a long journey ahead to get FDA and Health Canada approval for its medical app.

Ubenwa is still a long way from Onu’s goal of having cry assessment become a de facto test for newborns, but he is hopeful that the work his startup is doing will help change the mindset of how we hear a baby’s cry.

“The infant cry, I believe, is such a vital sign that a baby should not leave the hospital without their cry stamp being analyzed and added to the medical information that is being looked at by the physician,” said Onu.

Feature image by Minnie Zhou on Unsplash