Ubenwa Health - The 2018 AI Journey to tackle Asphyxia

January 30, 2019

By Charles C Onu, on behalf of the Ubenwa team

At Ubenwa, we set for ourselves the task of addressing the problem of perinatal asphyxia - one of the top 3 causes of newborn mortality today. We are applying machine learning to detect this breathing deficiency using only the infant cry as input. This has enabled us to develop an extremely cost-effective tool - a smartphone app - that is accurate and fast. I encountered first-hand the problem of asphyxia during the years I spent volunteering with Enactus and Fisher Foundation in Nigeria. The journey to develop, test and clinically validate Ubenwa has been exciting but difficult most of the time, with many lessons learned. I would like to take the chance of a new year to look back, reflect on how far Ubenwa has come and take a peek at where we are headed.

Partnerships

Building a research-grade database of carefully annotated infant cries is central to the quality of our machine-learning approach. As a growing organization working at the crossroads of healthcare and artificial intelligence, we have focused much of our efforts on developing partnerships with health research organizations and pioneers in AI.

In early 2018, we officially launched a project with the Montreal Children’s Hospital (MCH), under the McGill University Health Centre. Our plan was to carry out further data acquisition, testing and validation there. Despite the tireless efforts of several clinicians and hospital administrators, we underestimated the time required to finalize ethics and clinical study procedures for gathering data. In late 2018, and in part, to help address some of the issues encountered, we entered a partnership with MILA, the Quebec Artificial Intelligence Institute. MILA will be providing us with consultation, AI expertise, access to compute infrastructure, office space, financial support, among others. We are presently on track to getting the work off the ground in the coming weeks.


While we have accelerated our research and growth in Canada, we must also resume work where our tool will have the most significant impact. Ubenwa was birthed in Nigeria. A key component of our plan in 2019 is to return to Nigeria. In 2018, we received confirmation from paediatricians at two major teaching hospitals - Enugu State University Teaching Hospital, ESUTH (which also happens to be in my home state) and University of Port Harcourt Teaching Hospital, UPTH. Both institutions are high capacity referral centres in their respective cities, with surrounding rural and urban areas depending on them. We are continuing talks with doctors at other hospitals, while working with ESUTH and UPTH to develop a plan for evaluating what impact Ubenwa can have in the clinical setting in Nigeria.

Funding

While Ubenwa has been largely driven by volunteers who have invested valuable time and expertise over the last 2 years, our planned clinical studies will involve significant costs that we cannot bear. Fortunately, in view of the study at the Montreal Children’s Hospital, we have grant commitments of $25,000 CAD from Quebec Ministry of Economy, Science and Innovation (MESI) and $95,000 CAD from MILA.

We are very grateful to Osler who provided us with pro bono legal services and District 3 Innovation Center who has provided us invaluable support till date by paying interns to work with us, providing free office space as well as access to countless resources.

We are actively seeking funding for testing, data acquisition and validation with all our clinical sites.

Symposia

Sharing our work at conferences and public events have been an important way to receive feedback on Ubenwa, find new collaborators, and more generally share our vision of the possibilities of artificial intelligence in healthcare. 2017 was the year our team got a chance to speak about Ubenwa at our first big stages. Eyenimi (clinical lead) presented Ubenwa at United Nations’ AI for Good Global Summit (see YouTube video), Innocent (software engineering lead) on the Nigerian delegation at ITU Telecom World and I, at the Montreal AI Symposium (see Facebook video).

Last year, we got several opportunities to showcase Ubenwa at the world stage. I will highlight a few.

  • January, keynote at AI on a Social Mission conference (Montreal),
  • February, a talk on the show - This Week in Machine Learning and AI (TWiML),
  • June, exhibition at the UN AI for Good conference (Geneva),
  • July, presentation at UNESCO Technologies for Development conference (Lausanne),
  • December, at Neural Information Processing Systems (NeurIPS) the largest gathering of academic and industrial machine learning researchers, we presented a new research direction we are taking applying neural transfer learning from adult to newborn speech (Montreal).

Overall, these opportunities helped underscore the belief that the community has in us, leaving us with a deeper feeling of responsibility over the work we do.

AI XPrize

In December 2018, we qualified for the 3rd round of the $5million AI XPrize competition. The AI XPrize is a 4-year long global competition challenging teams to develop and demonstrate how AI could tackle the world’s grand challenges. It kicked-off in 2016 with 152 teams. As of December 2018, Ubenwa is now among only 30 teams remaining in the competition.

Media

The media has picked up on our work and helped us spread the word. Ubenwa was featured in:

Team and advisors

I cannot thank the team enough for their commitment. First, Ubenwa would simply not exist but for Innocent’s push and persistence since the birth of the idea over 4 years ago. He developed our proof-of-concept app and remains a solid co-founder and engineering lead. Dr Opara has played an invaluable role in steering our clinical efforts. And Jon, who has been with the team for over 1 year now, has brought in his wealth of experience in building startups. I would also like to thank Smita Rao and Marina Ayrapetyan for their drive and passion they gave during the time spent working with us.

The Ubenwa Team

The Ubenwa Team

We have been blessed with a set of advisors whose guidance has in no small way contributed to the path we are on - Prof Precup, Prof Alikor, Prof Kearney, Dr Ndiomu, Urbain, and Gautam; as well as Prof Sant’Anna and Dr Shalish at the MCH. We would like to specially thank Prof Precup whose advice and concrete efforts have helped us overcome several operational challenges over the last 2 years.

Supporters

Beyond our team and advisors, there have been several individuals and organisations who we have to thank. Jeanne Sauve Foundation for providing us with our first home. Sydney Swaine-Simon and District 3 for critical support, exposure and access to resources. Yoshua Bengio for his advice and help in realizing the Mila collaboration. Hugo Larochelle who persistently continues to spread the word about us at Google and elsewhere. We thank you all.

2018 was a productive year for us. Many persons contributed and continue to contribute to our progress. We begin 2019, ready to tackle the many challenges and look forward to expanding our work from Canada to Nigeria. Ultimately, we hope to start demonstrating the impact Ubenwa can have in Nigeria and beyond.