Deep Ensemble Approach for Enhancing Brain Tumor Segmentation in Resource-Limited Settings

Abstract

Brain tumor segmentation is a critical step in treatment planning, yet manual segmentation is time-consuming and subjective — a challenge magnified in Sub-Saharan Africa by overburdened medical systems and limited access to advanced imaging modalities and expert radiologists. We develop a deep learning ensemble combining UNet3D, V-Net, and MSA-VNet for automated glioma segmentation, training initially on the BraTS-GLI dataset and fine-tuning on the BraTS-SSA dataset. The ensemble achieves Dice scores of 0.8358 (Tumor Core), 0.8521 (Whole Tumor), and 0.8167 (Enhancing Tumor), outperforming individual model implementations and underscoring the potential of ensemble methods for automated brain tumor segmentation in resource-limited settings.

Jeremiah Fadugba (Jerofad)

Data Scientist | Machine Learning Engineer

My current research interests include Medical Imaging, Machine Learning, Deep Learning, and Trustworthy ML.