This dataset contains 302 anonymised T2-weighted abdominal pelvic MRI DICOM slices from 8 female patients referred for pelvic MRI at Crestview Radiology Ltd, Nigeria. The dataset was collected to support research into artificial intelligence-assisted endometriosis screening in Sub-Saharan Africa, where annotated medical imaging datasets are critically scarce.
Dataset characteristics:- 8 patients, 302 DICOM slices total- Modality: T2-weighted MRI, 320x320 pixels- Scanner: 1.5T MRI- Acquisition dates: 2021-2024- No clinical labels available
Ethical approval was obtained from the Babcock University Health Research Ethics Committee (BUHREC) and the Crestview Radiology Ltd Research Ethics Committee. All data are fully anonymised in compliance with the Declaration of Helsinki. Informed consent was obtained from all participants.
This dataset is the foundation for the research pipeline described in the following manuscript currently under review:
Fatade, O. B. et al. (2026). GAN-Augmented Graph Neural Network Framework for Endometriosis Detection in Pelvic MRI: A Pilot Study from Sub-Saharan Africa. Dataset DOI:
doi.org
The dataset is also available on Kaggle at:
kaggle.com