In Africa, AI is enabling experts to detect medical anomalies in baby cries

“Like a special language, there is lots of health-related information in various cry sounds,” says Lichuan Liu, director of the Digital Signal Processing Laboratory at Northern Illinois University.

In 2019, Liu alongside some of her colleagues used a specific algorithm based on automatic speech recognition to detect and recognise the features of infant cries, this allowed them to distinguish the meanings of both normal and abnormal cry signals in a noisy environment. “The differences between sound signals actually carry the information. These differences are represented by different features of the cry signals. To recognize and leverage the information, we have to extract the features and then obtain the information in it,” she said.

Prior to Liu's research, a Nigerian tech entrepreneur and medical practitioner, Charles Onu co-founded Ubenwa, a software-as-a-service startup that is developing technologies, such as a mobile app and API, that is helping the medical interpretation of infants’ cries, using artificial intelligence and machine learning algorithms to extrapolate patterns and insights.

Charles, Ubenwa’s CEO was inspired to build the startup when his cousin was born with birth asphyxia and later developed a hearing condition. This desire deepened after he worked with a health department at Enactus in Nigeria, a global youth organisation, and he saw cases similar to his cousin’s.

When he relocated to Canada to work with neonatal experts, Charles continued to see the challenges of communication between experts and infants. “There’s a lot of guesswork doctors have to do with babies,” he said. “But a lot of (infant care) is really making informed guesses as to when to make this action or where to make that action, and we’re hoping to close some of these gaps.”

After about five years of research at Mila, an AI hub in Quebec, Canada. He founded Ubenwa in 2017 with the objective to “...bring the world to a point where infant cries are considered to be a vital sign just as much as we would consider their heart rate to be a vital sign,” Charles told TechCrunch. “Ultimately, our goal is to be a translator for baby cry sounds, providing a non-invasive way to monitor medical conditions everywhere you find a baby: delivery rooms, neonatal and paediatric intensive care units, nurseries, and even homes.”

Currently, the Ubenwa software identifies early signs of birth asphyxia, and can potentially determine learning milestones based on cry triggers. In its early days, the lack of access to existing samples of babies' cries was a challenge, this led the Montreal-based healthtech startup, to establish a “close collaboration” with six hospitals, including two from Nigeria—Enugu State University Teaching Hospital and Rivers State Teaching Hospital in Nigeria.

Ubenwa says that it has the largest and most diverse clinically annotated database of infant cry sounds, an essential asset for the development of audio-based biomarkers. The startup’s first pilot on detecting neurological injury due to birth asphyxia demonstrated about 40% improvement over APGAR scoring, the most common physical exam at birth.

“The significance of the app lies in the fact that it is fast, cost-effective, non-invasive, and easy to use,” says Guilherme Sant'Anna, Neonatologist at Montreal Children’s Hospital and Professor at McGill University.

“Today, doctors use physical assessment to look at eyelids, look at the skin tone, and so on and so forth,” Onu said. “If [doctors] are really worried it could be with an MRI or a brain MRI machine because that’s the ultimate standard, but we don’t live in an MRI machine every day. That is costly. With simple cry analysis, you can track neurological biomarkers on an ongoing basis, non-invasively.”

Last year, the company raised $2.5 million in pre-seed funding in a round led by Radical Ventures to scale its operations.

“Supported by a strong clinical foundation, Ubenwa has developed a proprietary innovation for an underserved and important market,” Sanjana Basu, an investor at Radical Ventures said. “Deciphering a baby’s cry using machine learning can open up a range of possibilities in the consumer and clinical paediatrics market where demand for better digital products is only growing.”

Aside from developing the product, Ubenwa is also investing in training medical practitioners in underserved communities. Last year, the health-tech startup trained pediatricians at the Lagos State University Teaching Hospital, on how to conduct the Bayley Assessment.

Bayley is an exam for assessing neurodevelopmental delays in children. Although it is important for clinical studies, several medical facilities in low- and middle-income countries are not equipped to conduct the exam.

“By upskilling the health professionals, they can effectively collaborate on our clinical studies but also deliver better care to their patients,” Charles wrote in Ubenwa's 2022 end-of-year review. “Growing up in Nigeria, I experienced first-hand the positive impact of having access to medical specialists as well as the tragedy of not having this. So it's really a gift to be able to contribute in this way.”

Charles is working with more than 13 other experts in clinical development, machine learning, software engineering, and business operations, to build Ubenwa. The healthtech startup also has about six advisors including Edward Alikor, a professor of Paediatric Neurology at the University of Port Harcourt Teaching Hospital in Nigeria.

In 2019, Ubenwa was recognised by the World Health Organisation as one of the top 30 healthcare innovators in Africa.