Naija-CXR dataset-JPG-A Nigerian Chest X-ray Dataset for Equitable AI and Vision-Language Research

Naija-CXR is a curated chest X-ray (CXR) dataset collected from Nigerian healthcare institutions to support the development and evaluation of artificial intelligence (AI) models for medical imaging in low-resource settings. The dataset was created to address the underrepresentation of African populations in publicly available chest X-ray datasets and to facilitate research on domain adaptation, fairness, and equitable AI for medical diagnosis. The dataset contains de-identified frontal chest X-ray images paired with expert radiologist annotations and diagnostic reports. Where available, accompanying demographic and clinical metadata have been anonymized in accordance with institutional ethical approval and applicable data protection guidelines. Naija-CXR is intended to support research in: Chest X-ray disease classification Automated radiology report generation Vision-language models (VLMs) Domain adaptation and transfer learning AI fairness and bias mitigation Foundation models for medical imaging Medical image representation learning The dataset complements existing public datasets such as MIMIC-CXR by providing imaging data from an underrepresented African population, enabling the development of more robust and equitable AI systems. Dataset Contents De-identified frontal chest X-ray images Free radiologist reports Dataset documentation Label definitions Intended Use The dataset is released exclusively for non-commercial scientific research and educational purposes. It should not be used as a substitute for professional medical diagnosis or clinical decision-making. Ethics All personally identifiable information has been removed prior to release. Data collection and sharing were conducted under approval from the relevant Institutional Ethics Committee. Users must comply with all applicable ethical and legal requirements when using the dataset.