We build open, clinically validated datasets and benchmarks from African hospitals, starting with chest radiology in Nigeria. AI trained elsewhere does not reliably work here. We are changing what it learns from.
The work so far
What we do
Batlak AI Lab works at the foundation of medical AI. Before a model can help a patient, it needs data that reflects that patient. We build that data, and the methods to validate it.
Structured datasets and natural language methods for chest radiology, built from real Nigerian clinical reports.
Curation, annotation, and validation of clinical data that reflects African patients and practice.
Methods to test whether medical AI actually performs on local data.
Growing local expertise in medical AI research.
Featured research
Our work runs as ongoing research programmes. Each one produces an open dataset and peer reviewed research.
FMCEB-CXR is a structured chest X-ray benchmark built from 5,517 anonymized radiology reports at Federal Medical Centre Ebute-Metta, Lagos. Labels are extracted with a hybrid natural language pipeline and validated by consultant radiologists.
Approach
How we work is part of the point.
The data is sourced, labeled, and held in Nigeria. Ownership stays local, because the value and the responsibility both should.
Radiologists and clinicians validate the work. The data reflects real practice, not assumptions.
Released for the research community, so others can build on it and benchmark against it.
Built to comply with Nigeria's Data Protection Act, with anonymization and documented safeguards.
Roadmap
A future initiative to track datasets, benchmarks, and medical AI research across Africa. A public knowledge hub for researchers, clinicians, students, and policymakers.
Work with us
We welcome research collaborators, clinical partners, and data contributors.