Our Research
Advancing the frontiers of AI in healthcare through rigorous research focused on East Africa's unique challenges and opportunities.
Research Focus Areas
Language Adaptation
Developing techniques to adapt AI models to understand and process East African languages, enabling more natural and effective healthcare communication.
Privacy-Preserving AI
Researching federated learning and other techniques that enable AI model training without centralizing sensitive patient data.
Cultural Context in AI
Exploring how cultural context affects AI performance in healthcare settings and developing methods to create culturally-adaptive AI systems.
Low-Resource Computing
Developing techniques to optimize AI models for low-bandwidth environments and resource-constrained devices common in rural healthcare settings.
Medical Imaging AI
Researching computer vision techniques for medical imaging analysis, with a focus on conditions prevalent in East Africa and images from diverse skin tones.
AI Ethics & Governance
Developing frameworks for ethical AI deployment in healthcare settings, with a focus on fairness, transparency, and accountability in the East African context.
Recent Publications
Adapting Medical AI Models to East African Languages: Challenges and Solutions
Dr. Sarah Mwangi, James Omondi, Dr. Thomas Okello
Journal of Healthcare Informatics • October 2023
This paper explores the challenges of adapting medical AI models to understand and process East African languages, with a focus on Swahili, Amharic, and Luganda. We present novel techniques for transfer learning and data augmentation that significantly improve model performance in low-resource language settings.
Federated Learning for Privacy-Preserving Healthcare AI in Rural Settings
James Omondi, Dr. Amina Kimani, Michael Odhiambo
AI for Global Health Conference Proceedings • August 2023
We present a novel federated learning approach that enables AI model training across distributed healthcare facilities in rural East Africa without centralizing sensitive patient data. Our approach addresses challenges of intermittent connectivity and device heterogeneity while maintaining model accuracy.
Cultural Context in Medical AI: Improving Diagnostic Accuracy Through Cultural Adaptation
Dr. Amina Kimani, Grace Achieng, Prof. David Otieno
Ethics in Artificial Intelligence • June 2023
This study examines how incorporating cultural context into medical AI systems affects diagnostic accuracy and patient trust. Through a series of controlled experiments in diverse East African communities, we demonstrate that culturally-adapted AI systems achieve significantly higher accuracy and user acceptance.
Low-Bandwidth Optimization Techniques for Healthcare AI in Remote Areas
Daniel Muthoni, Thomas Okello, Dr. Sarah Mwangi
IEEE Transactions on Medical Imaging • April 2023
We present a suite of optimization techniques that enable sophisticated medical imaging AI to function effectively in low-bandwidth environments. Our approach combines model compression, progressive loading, and adaptive resolution to deliver diagnostic assistance even in areas with limited connectivity.
Our Research Partners
University of Nairobi
Kenya
Makerere University
Uganda
University of Dar es Salaam
Tanzania
MIT Media Lab
USA
Stanford HAI
USA
DeepMind Health
UK
Join Our Research Efforts
We're always looking for research partners, collaborators, and talented researchers to join our team.