Maternal & Newborn Health
Ubenwa Health
Analyzing infants' cry sounds using machine learning to detect early signs of birth asphyxia.
One-line solution summary:
Analyzing infants’ cry sounds using machine learning to detect early signs of birth asphyxia.
Pitch your solution.
Ubenwa is a smartphone app which screens for birth asphyxia by analysing a newborn’s cry sounds. Birth asphyxia is a major cause of neonatal disability and death, resulting from shortage in oxygen supply to the brain. The Ubenwa algorithm works by classifying acoustic biomarkers of brain injury using machine learning.
Every year, the lack of specialised expertise or equipment to rapidly screen babies for asphyxia, leads to about 2 million casualties, mainly in low- and middle-income countries. Using Ubenwa, health workers involved in newborn delivery will be able to assess patients at risk and refer them on time for potentially life-saving treatment.
By deploying Ubenwa as a smartphone app, we aim to increase global access to this critical screening as it is: 1) cost-effective 2) easy to use 3) fast, and 4) non-invasive.
Film your elevator pitch.
What specific problem are you solving?
We are addressing the problem of inadequate access to early screening for birth asphyxia leading to high newborn mortality in developing countries.
Birth asphyxia is when a newborn lacks adequate oxygenation at birth, resulting in harm to the brain which can lead to death or severe life-long disabilities, such as cerebral palsy, deafness, and paralysis. It is one of the top 3 causes of newborn mortality in the world. Every year, it is responsible for the disability and death of up to 2 million newborns.
Developing countries carry most of this disease burden and have the highest rates of newborn mortality. Early diagnosis of birth asphyxia is critical for survival and for reducing brain injury, but this requires expensive equipment and specialized personnel which are scarce in low-resource settings. The result is that many babies only get referred for treatment when they are extremely sick and may have suffered irreversible neurological damage.
Evidently, there is an urgent need for a cost-effective, easy to use, and non-invasive solution to quickly identify newborns with birth asphyxia to reduce mortality and disability, such that even the lowest resource birthing centers can afford to use it.
What is your solution?
Ubenwa is deployed as a smartphone app that analyzes a baby’s cry to make a quick, reliable and non-invasive diagnosis of birth asphyxia.
At the core of Ubenwa are two machine-learning-based detection algorithms: first, a cry detection algorithm to identify a newborn’s cry from other sounds; second, an asphyxia detection algorithm to identify the distinguishing characteristics between cries of healthy and asphyxiated babies.
The app allows rapid detection of birth asphyxia in 3 simple steps: the birth attendant first records the newborn crying, the app then detects and isolates the cry sounds, and finally the app provides the risk of birth asphyxia.
We have been working with neonatologists in Nigeria and Canada to collect clinically-annotated cry data, to develop and to validate our algorithms.
A demo of the Ubenwa app can be found here: https://youtu.be/b32zvjFAivU
Who does your solution serve, and in what ways will the solution impact their lives?
Ubenwa is directly targeted at health workers in the delivery room in resource-constrained settings. They usually lack the tools to give their patients the highest quality of care. Ubenwa will make possible rapid screening for brain injury, thereby enabling better patient outcomes.
Ultimately, Ubenwa serves newborns and their families, given that a delayed diagnosis could lead to life-long disability and a consequent impact on the quality of life of affected individuals, or in the worst case, death.
A survey we conducted with health professionals from 14 federal medical centres in Nigeria confirmed the need. Nigeria has one of the highest rates of newborn mortality in Africa. Most public birth centres do not own a blood gas analyzer (used to detect a low-oxygen insult to the newborn) and many babies are born in centers without specialized care. The appropriate level of care for newborns with birth asphyxia is usually available in no more than 3 to 5 centres in a state serving between 1.5 - 5 million residents. So early detection is utterly crucial for knowing who to refer and when.
We have been engaging neonatologists, nurses and other paediatricians in the development and validation of Ubenwa through clinical studies.
Which dimension of the Challenge does your solution most closely address?
Expand access to high-quality, affordable care for women, new mothers, and newbornsExplain how the problem, your solution, and your solution’s target population relate to the Challenge and your selected dimension.
This MIT Solve Challenge asks the question of how newborns can access the care they need to “survive” and “thrive”. Ubenwa is ensuring that newborns get access to rapid diagnosis and treatment which could save their lives and decrease the chances of long-term disability.
By providing a solution in app form, all levels of birthing centers, regardless of resources, will be able to access this cost-effective screening tool for birth asphyxia.
Who is the primary delegate for your solution?
Charles C Onu
What is your solution’s stage of development?
Prototype: A venture or organization building and testing its product, service, or business modelIn what city, town, or region is your solution team headquartered?
Montreal, QC, CanadaIf you have additional video content that explains your solution, provide a YouTube or Vimeo link here:
Which of the following categories best describes your solution?
A new technologyDescribe what makes your solution innovative.
Our solution harnesses a newborn’s cry as a vital sign using artificial intelligence, creating an accurate, cost-effective, easy-to-use, and non-invasive screening tool for birth asphyxia. Current standards for diagnosis of birth asphyxia includes blood gas analysis and neurological examinations, with additional assessments of brain injury using electroencephalography (EEG) and magnetic resonance imaging (MRI).
All of these methods require expensive, specialized equipment and training for the testing procedure and interpretation of results, whereas our solution is a simple smartphone app that is completely contact-free and requires no clinical expertise to use.
Describe the core technology that powers your solution.
Ubenwa is powered by a combination of audio signal processing and machine learning algorithms. Using data we are collecting in conjunction with our clinical partners in Canada and Nigeria, we are identifying acoustic biomarkers of healthiness and of pathology. Then we are leveraging neural networks to map these acoustic features to accurate predictions.
Our current results indicate performance of almost 90% sensitivity and specificity in correctly identifying newborns that have suffered birth asphyxia.
Ubenwa has been integrated into a mobile app which our clinical collaborators are now using to acquire more data and to validate our models in neonatal units.
Below is a block diagram of our classification model:
Provide evidence that this technology works.
Ubenwa is based on scientific research led by our founder Charles, whose work has appeared in major scientific venues including: Neural Information Processing Systems (NeurIPS), Engineering in Medicine and Biology Conference (EMBC), International Conference on Learning Representation (ICLR) and Conference of the International Speech Communication Association (INTERSPEECH).
Below are a few publications for reference:
C. C. Onu, J. Lebensold, W. L. Hamilton et al, “Neural Transfer Learning for Cry-based Diagnosis of Perinatal Asphyxia”, 20th Annual Conference of the International Speech Communication Association INTERSPEECH 2019.
C. C. Onu, I. Udeogu, E. Ndiomu et al, “Ubenwa: Cry-based Diagnosis of Birth Asphyxia”, Machine Learning for Development workshop, 31st Conference on Neural Information Processing Systems (NeurIPS) 2017.
C. C. Onu, “Harnessing infant cry for swift, cost-effective diagnosis of perinatal asphyxia in low-resource settings,” Machine Learning for Healthcare workshop, 29th Conference on Neural Information Processing Systems (NeurIPS), 2015.
Watch the video below for a live demo of the app by Charles: https://youtu.be/b32zvjFAivU
Please select the technologies currently used in your solution:
Select the key characteristics of your target population.
Which of the UN Sustainable Development Goals does your solution address?
What type of organization is your solution team?
For-profit, including B-Corp or similar modelsHow many people work on your solution team?
Our team consists of 3 full-time members, 1 part-time, 8 advisors and 10 partnering neonatologists.
How many years have you worked on your solution?
5 years
Why are you and your team well-positioned to deliver this solution?
Our core team has complimentary skills in computer science, software engineering, clinical research and social innovation.
Charles Onu is our Artificial Intelligence Lead. He is a PhD candidate currently carrying out research in AI and biomedical engineering at Mila - the Québec AI Institute and McGill University. His research has been published at major AI venues. He has industry experience in software engineering and machine learning; as well as in implementing social ventures.
Samantha Latremouille is our Clinical Research Lead. She is a PhD candidate in Experimental Medicine at McGill University, with additional studies in Translational Biomedical Engineering and Surgical Innovation, and knowledge of biodesign processes, medical device patents, clinical trials, and regulatory affairs. She is currently conducting clinical research in the newborn intensive care unit investigating biological signal analysis as the future of neonatal care.
Innocent Udeogu is our Software Engineering Lead. He is a graduate of the Meltwater Entrepreneurial School of Technology (MEST Africa). Innocent has headed and completed several production software applications including building applications at Andela. As a Yunus and Youth fellow, he also brings to the table his experience in developing strategies for effective social innovation.
Furthermore, our team of advisors span the fields global health, healthcare product development and leading experts in artificial intelligence (Prof. Yoshua Bengio and Prof. Doina Precup) and biomedical engineering (Prof. Robert Kearney).
What organizations do you currently partner with, if any? How are you working with them?
Our partners have been a core part of our successes till date:
- Clinical Studies: McGill University Health Centre (MUHC) in Montreal, Canada and Enugu State University Teaching Hospital (ESUTH) in Nigeria have strong reputations for newborn health research and are the primary sites for our clinical studies.
- Technical and Operational Support: Mila (Québec Artificial Intelligence Institute) provides Ubenwa with office space, access to compute power, and research collaborations. The Québec Ministry for Economy and Innovation awarded Ubenwa our first ever seed grant of $25,000. District 3 Innovation Centre provided incubation, office space, social venture coaching, legal advice and a family of like-minded dreamers.
Why are you applying to Solve?
Our next major milestone is to validate Ubenwa on thousands of patients internationally, in both high and low-income countries. We believe that Solve can accelerate our progress by providing multiple types of support:
Partners: We are seeking new clinical partners to expand the geographical span of our ongoing studies.
Mentors: We are interested in bringing on new advisors and mentors from the Solve community in the areas of regulatory approval of medical devices and global distribution, especially within the unique environment of developing countries.
Prize: Prize money from Solve would enable us to acquire to pursue our expansion plans for our ongoing clinical studies. It would also enable us to direct resources to product development and intellectual property protection which are critical in the highly regulated space that we work in.
In which of the following areas do you most need partners or support?
What organizations would you like to partner with, and how would you like to partner with them?
There are 2 sets of partners we aim to work with over the next 2 years:
- Clinical partners: we are looking to expand our clinical network of university hospitals to increase the size and diversity of the data with which we validate the Ubenwa algorithms. We are already working with 6 hospitals and clinics in Nigeria and Canada, and are looking to expand into the US and South Africa.
- Distribution partners: we are further looking to connect with organisations, such as the Gates Foundation, UNICEF, and other newborn-health-oriented non-governmental organizations or initiatives. When we move to field trials, these partners could help with access to their network of clinics and hospitals in resource-constrained settings, as well as their wealth of experience in scaling social innovations.
Solution Team
-
Samantha Latremouille Clinical Lead, Ubenwa Health
-
Charles Onu AI Research Lead, Ubenwa Health
-
Innocent Udeogu Software Engineering Lead, Ubenwa Health
to Top
Solution Name:
Ubenwa Health