Mobile app could save lives by using AI to detect respiratory distress in newborns

By Diane Weidner

Growing up in Nigeria, Charles Onu was torn between deciding whether to pursue studies in mathematics or medicine, his two passions. “I wanted to be a doctor, to help people,” explains Onu, “but I also loved mathematics and was very good at it. Math came naturally to me. So I had to make a choice.”

He decided to pursue a BEng degree in Electrical Computer Engineering at the Federal University of Technology in Owerri, but his desire to care for others remained strong. In 2008, Onu began volunteering for Enactus, a global non-profit organization, first as a member then as president. That’s how he first learned about perinatal asphyxia, a medical condition that deprives the brain of oxygen in newborn babies, resulting in serious complications and possible death. “This is a very big problem for resource-limited communities in developing countries like Nigeria where sophisticated detection tools are not available,” explains Onu.

It occurred to him that he could combine his passion for math and medicine to address a global health problem. His idea was to develop a smart tool to identify babies who are showing signs of respiratory distress by analyzing their cry patterns. In 2015, Onu joined forces with Innocent Udeogu, a former schoolmate, and they began to work on this problem together.

Cry of a baby

Innocent Udeogu

Fast forward five years and Onu is now the Founder and AI Research Lead of a promising start-up, Ubenwa, which means “cry of a baby” in the Igbo language of Nigeria. Udeogu leads the Software Engineering Development and is the contact person on the ground in Nigeria while Samantha Latremouille, a PhD student in Experimental Medicine at McGill, is the Clinical Lead.

The Ubenwa team has developed a fast, easy-to-use mobile app that uses artificial intelligence to analyze the sound of a newborn baby’s cries. This cost-effective and non-invasive tool allows clinicians to flag the risk of newborn asphyxia, which, if detected early, is easy to treat. It has the potential to save lives and improve outcomes for millions of newborns each year.

Looking for the right people

Onu first came to Montreal in 2015 on a Jeanne Sauvé Foundation fellowship as a visiting researcher. This public leadership program allowed him to access resources at both McGill and Concordia, where he made important connections. At the time, Ubenwa was still a nascent idea.

“I started initial work to develop this concept but at one point, it became clear to me that I needed more skills. I needed to immerse myself in the environment and work with researchers who have experience and expertise on the intersection of medicine and AI.”

Samantha Latremouille

He applied to McGill, where he completed a Master’s Degree in Computer Science under the supervision of Doina Precup, Associate Professor at the School of Computer Science. During this time, while working on a project with the Montreal Children’s Hospital, he met Latremouille, a Montreal native who had earned her BSc in Physiology at McGill in 2013. With her unique background in neonatology, clinical research, innovation, biomedical engineering and medical device development, she was the perfect fit to fill the role of Clinical Lead at Ubenwa.

“When Charles told me about Ubenwa, I was immediately captivated and wanted to help,” says Latremouille. “I have seen the devastation that birth asphyxia can cause through the course of my PhD, and I was shocked to hear that the lack of resources in developing countries meant that so many of these babies were not identified early so they missed out on time-sensitive life-saving treatments. We are fortunate to live in a place with the systems and resources to provide the best care for asphyxiated babies, but many people in the world do not have this, and the consequences are dire.”

Leveraging mobile technology

The idea of correlating infant cry patterns to medical conditions is not new. “The cry of the newborn is this pure signal,” explains Onu. “It’s an involuntary response to stimuli, so the infant has no control over how loud or soft the cry is. They have no control over their vocal cords like we do as adults. Their cry pattern is directly coordinated by the Central Nervous System (CNS). The hypothesis in the clinical community for some time now has been that when babies suffer from conditions that affect the CNS, they will have altered cry patterns. Asphyxia is such a condition. If you go back to the 1960s, you can find research attempts to try to correlate these cry patterns, before we had advanced tools to do so.”

Fortunately, we are now able to leverage advances in mobile technology with audio and machine learning. “Almost everyone has a smart tool, a device with huge processing capacity that can record audio,” explains Onu. “And we have advanced AI tools that allow us to decipher patterns in cries. With coding, we can convert all that data into actionable tools.”

Collecting the cries

The process of designing, validating, and testing a clinical app is very complex, and the Ubenwa team would soon realize the complexities of the journey they were embarking on.

One obstacle involved finding clinical data to work with. “AI algorithms need data to work, and you can’t just record any baby on the street,” explains Onu. “The data needs to be clinically validated, to be categorized as either normal or pathological. Next, you need to train the algorithm with that data.” Fortunately, Onu was able to contact a researcher who had done work on analyzing infant cry patterns in Mexico and obtain access to their database. This allowed the team to set up the first set of analysis and research ideas, then build a prototype and make a strong case for investment.

Supporting student entrepreneurs

Onu is currently working on his PhD in Computer Sciences at McGill under the mentorship of Professor Precup, who is also Co-Director of the Reasoning and Learning (RL) Lab at McGill and Director of the DeepMind Lab in Montreal. “Ubenwa is a very strong example of the kind of good that machine learning can have in our world. It has been a pleasure to support Charles and his team over these years towards realizing the vision of saving babies with AI,” says Professor Precup.

Onu is also part of Mila, an AI research centre that evolved from a partnership between the Université de Montréal and McGill. He is grateful for the ecosystem of programs that support student entrepreneurs at McGill, including the McGill Dobson Cup and the McGill Clinical Innovation Competition (CLIC). “All of these opportunities have prepared us, have allowed us to work on our narrative and improve our product,” says Onu.

Photo of Charles and advisor Urbain Kengni
Charles Onu (left) and Ubenwa advisor Urbain Kengni

As a finalist team in the 2020 McGill CLIC, the team received personalized coaching from McGill’s SKILLSETS program with pitch specialist Dr. Andrew Churchill on how to captivate the audience’s attention and keep them engaged. Steven Arless, Lead Judge and Professor of Practice at McGill’s Faculty of Medicine and Health Sciences, congratulated Ubenwa for their outstanding presentation, and remarked that in his many years of supervising competitions, he had never seen such a compelling presentation.

Gathering clinical data

Since 2017, Ubenwa has obtained funding from Mila, from the Ministère de l’Économie et de l’Innovation (MEI) de Québec, and from the District 3 Innovation Centre to support clinical studies in Nigeria. They are currently working in partnership with health care professionals to gather clinical data and create a database of the cries of 2,500 newborn infants.

More recently, Ubenwa won three awards at the MIT Solve 2020 Global Challenge, including the Health Workforce Innovation Prize. It plans to use the prize money to prepare for clinical trials and lay the groundwork for regulatory approval process with the FDA and Health Canada.

“If there’s one thing that babies do, they cry,” says Onu. “It’s not something you have to look for, or find, or place a sensor in the right location to capture it. In places where the more complicated tests are not available, the cry analysis will really be a saving grace.”

The Ubenwa team is working hard to get the app to the point where it starts saving lives, and are relying on McGill’s credibility and resources to help move the vision forward.



January 13, 2021