Bridging the Gap: Generalising State-of-the-Art U-Net Models to Sub-Saharan African Populations

A critical challenge for tumour segmentation models is the ability to adapt to diverse clinical settings, particularly when applied to poor quality neuroimaging data. The uncertainty surrounding this adaptation stems from the lack of representative datasets, leaving top-performing models without exposure to common artefacts found in MRI data throughout Sub-Saharan Africa (SSA). We replicated a framework that secured the 2nd position in the 2022 BraTS competition to investigate the impact of dataset composition on model performance and pursued four distinct approaches through training a model with various combinations of the BraTS-Africa and BraTS-Adult Glioma datasets. Notably, training on the smaller low-quality BraTS-Africa dataset alone yielded subpar results, and training on the larger high-quality BraTS-Adult Glioma dataset alone struggled to delineate oedematous tissue in the low-quality validation set. The most promising approach involved pre-training a model on high-quality neuroimages and then fine-tuning it on the smaller low-quality dataset. This approach took second place in the MICCAI BraTS Africa global challenge external testing phase. These findings underscore the significance of larger sample sizes and broad exposure to data in improving segmentation performance. Furthermore, we demonstrated there is potential for improving such models by fine-tuning them with a wider range of data locally.