Brain tumor segmentation in Sub-Saharan Africa patient population: The BraTS-Africa challenge.
Authors
Affiliations (23)
Affiliations (23)
- Medical Artificial Intelligence Laboratory (MAI Lab), Lagos, Nigeria.
- College of Medicine, University of Lagos, Lagos, Nigeria.
- Department of Radiology, University of California, San Diego, California, USA.
- Crestview Radiology Limited, Lagos, Nigeria.
- Department of Electrical Engineering, University of British Columbia, Vancouver, British Columbia, Canada.
- Department of Biomedical Engineering, McGill University, Montreal, Quebec, Canada.
- Department of Radiology, Lagos University Teaching Hospital (LUTH), Lagos, Nigeria.
- Department of Radiology, Lagos State University Teaching Hospital (LASUTH), Ikeja, Lagos, Nigeria.
- Department of Radiology, NSIA-Kano Diagnostic Center, Kano, Nigeria.
- Department of Radiology, National Hospital Abuja, Abuja, Nigeria.
- Department of Radiology, Nnamdi Azikiwe University Teaching Hospital, Nnewi, Anambra State, Nigeria.
- Department of Radiology, Federal Medical Centre, Umuahia, Abia State, Nigeria.
- Medhub Africa, Umuahia, Abia State, Nigeria.
- Department of Radiology, Federal Medical Centre, Abeokuta, Ogun State, Nigeria.
- Department of Radiology, Benue State University Teaching Hospital, Makurdi, Benue State, Nigeria.
- Lily Hospital Limited, Benin, Nigeria.
- Cornell University, Ithaca, New York, USA.
- Department of Radiation Oncology, Duke University Medical Center, Durham, North Carolina, USA.
- Department of Biomedical Engineering, Emory University, Atlanta, Georgia, USA.
- Department of Pathology and Laboratory Medicine, Indiana University, Indianapolis, Indiana, USA.
- Department of Quantitative Biomedicine, University of Zurich, Zurich, Switzerland.
- Department of Radiology, University of Pennsylvania, Philadelphia, Pennsylvania, USA.
- Montreal Neurological Institute, McGill University, Montreal, Quebec, Canada.
Abstract
Automated brain tumor segmentation on multi-parametric magnetic resonance imaging (mpMRI) is crucial in assessing patient outcomes and remains a challenge across Sub-Saharan Africa (SSA). Since 2012, the Brain Tumor Segmentation (BraTS) Challenge has evaluated state-of-the-art artificial intelligence (AI) methods to detect, characterize, and classify tumors. However, it is unclear if these methods can generalize, and hence be widely implemented, in SSA populations. To address this, the BraTS-Africa challenge was conducted in 2023 and 2024 to evaluate population-specific AI models for automated segmentation of brain tumors on mpMRI from the region. Preoperative T1-weighted, T1-contrast enhanced, T2-weighted, and T2-FLAIR brain MRI scans of 115 patients diagnosed with adult diffuse glioma were curated, annotated, and utilized for the Challenge. Participants were invited to develop and validate their methods. The participating teams were evaluated using weighted Dice Score Coefficient (DSC) and 95% Hausdorff Distance (HD95). In the 2023 challenge a total of 9 teams with participants from 12 countries met the submission criteria and were ranked. The 2024 challenge featured 6 teams from 7 countries. Compared to 2023, the six top ranked methods from 2024 had higher DSC (9.4%) and lower HD95 (57.21%). Together, the 2023 and 2024 challenges provided a unique opportunity to include brain mpMRI glioma cases from SSA in global efforts to develop and evaluate AI methods for the detection of glioma boundaries and their quantification, where the potential of AI solutions to transform healthcare into resource-limited settings is more likely.