Meet the Kenyan AI Scientist Bringing AI to Healthcare
Image Credit: Mbithe Nzomo
Author: Naliaka Odera
In 2018, the World Health Organization released a report titled The State of Health in the African Region, and its findings were dire. Access to healthcare in most African regions remains one of the most significant challenges with only 32% of the continent able to access essential services. As the clock ticks to meeting the SDG 2030 vision, many healthcare innovators are turning to AI to improve African healthcare access. Homegrown initiatives such as the Deep Learning Indaba annual conference of Africa’s AI researchers, to Data Science Africa, an African training organisation dedicated to bridging the gap between higher learning institutions, and the ever evolving knowledge required by researchers in deep learning, have been instrumental in ensuring that AI solutions explored bear an African foundation.
“Only 32% of Africa is able to access essential services.”
2023 Mawazo Fellow Mbithe Nzomo, a Kenyan researcher at the University of Cape Town, is one such innovator, keen to apply AI to healthcare systems. We caught up with her to understand the potential for AI in Africa’s development, as well as to grasp her big idea on shaking up Africa’s healthcare ecosystem.
Upon completion of her Masters in Advanced Computer Science at the University of Manchester, where she worked on a project using deep learning to help a social robot learn to interact with people, Mbithe was drawn to research in a field where she could have more direct positive influence in African development. In the first year of her PhD, her supervisor introduced her to the potential of wearable sensor data for health monitoring and she was off to the races.
Mbithe’s research stresses the importance of melding machine learning with knowledge-driven methods such as ontologies and knowledge graphs. Machine learning is a subset of AI where computer systems use algorithms to continually learn from data to make predictions. She explains that the use of machine learning is practical, since it allows systems to detect patterns in data without explicit programming. On the other hand, Mbithe states that “knowledge-driven methods can formally and unambiguously capture expert domain knowledge. They are also known to be explainable and easily interpretable, which means any outputs made by the system is easier to understand by humans, an essential quality in a high-stakes domain like health.” In using the two methods, Mbithe is leveraging both of their strengths.
“Health is one of the most significant determinants of the quality of human life and it made sense to shift my focus to the health domain”
Mbithe is evaluating the real-world applicability of her approach through two use cases, atrial fibrillation and stress monitoring. She began by conducting a mapping study of what is already in use in the field, and discovered a number of deficiencies in the current sensor-based health monitoring systems, especially those taking a knowledge-driven approach. She found that there was a lack of what she refers to as sufficient explainability. “Although knowledge-driven AI methods tend to be inherently interpretable, most systems lack explicit and targeted explanations for their outputs,” Mbithe says. Another issue she highlighted is the generation of automated recommendations without the human-centered approach of allowing users to cognitively engage in decision-making. She believes that complete automation limits user agency and can lead to over- or under-reliance on decision support tools, which can ultimately have a negative impact on health outcomes.
Mbithe’s work has significant practical implications for how machine learning and knowledge-driven approaches can be successfully integrated in AI systems for healthcare. “From a societal perspective, health monitoring systems have the potential to increase access to healthcare beyond clinical settings.” When her work is put into the African context of alarmingly low rates of healthcare access, Mbithe is uniquely positioned as an African woman in an already underrepresented field, putting African concerns in the forefront of her problem-solving efforts.
She is acutely aware of the low support for African women researchers, especially those in STEM fields, and remains grateful to her supervisor, her supportive environment at the University of Cape Town, and the Mawazo Institute. Through Mawazo, she has received training in topics not typically covered in PhD programmes, such as applying for grant funding, communicating research findings to non-experts, and translating research into policy. “Mawazo has been incredibly impactful in my growth as a researcher,” she declares.
As for Mbithe’s research, it has the possibility to greatly enhance the development of sensor-based health monitoring systems. She believes that this approach to improving healthcare access has the potential to turn things around for healthcare on the continent. She encourages public actors and governments to increase their investment in research such as hers, as a significant “way towards the reduction of healthcare inequalities.”