EMedNeXt: An Enhanced Brain Tumor Segmentation Framework for Sub-Saharan Africa using MedNeXt V2 with Deep Supervision

2025-07-31

AI Summary

In resource-constrained settings of sub-Saharan Africa, brain tumor segmentation suffers from severe MRI image degradation, scarcity of expert annotations, and shortage of specialized personnel—leading to low accuracy and high inference latency. To address these challenges in low-quality, few-shot scenarios, we propose EMedNeXt: a robust segmentation framework specifically optimized for such conditions. Its key contributions are: (1) enlarged input region-of-interest to mitigate artifact interference; (2) an enhanced nnU-Net v2 backbone built upon MedNeXt V2, incorporating deep supervision and adaptive post-processing; and (3) a lightweight model ensembling mechanism to improve generalizability. On an implicit validation set, EMedNeXt achieves a lesion-wise Dice Similarity Coefficient (DSC) of 0.897, with Normalized Surface Distance (NSD) scores of 0.541 (at 0.5 mm tolerance) and 0.840 (at 1.0 mm tolerance)—substantially outperforming existing methods. This work delivers a highly robust, low-dependency automated solution tailored for primary healthcare settings.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Semi-Supervised Learning
• Search and Optimization: Distributed Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Brain cancer affects millions worldwide, and in nearly every clinical setting, doctors rely on magnetic resonance imaging (MRI) to diagnose and monitor gliomas. However, the current standard for tumor quantification through manual segmentation of multi-parametric MRI is time-consuming, requires expert radiologists, and is often infeasible in under-resourced healthcare systems. This problem is especially pronounced in low-income regions, where MRI scanners are of lower quality and radiology expertise is scarce, leading to incorrect segmentation and quantification. In addition, the number of acquired MRI scans in Africa is typically small. To address these challenges, the BraTS-Lighthouse 2025 Challenge focuses on robust tumor segmentation in sub-Saharan Africa (SSA), where resource constraints and image quality degradation introduce significant shifts. In this study, we present EMedNeXt -- an enhanced brain tumor segmentation framework based on MedNeXt V2 with deep supervision and optimized post-processing pipelines tailored for SSA. EMedNeXt introduces three key contributions: a larger region of interest, an improved nnU-Net v2-based architectural skeleton, and a robust model ensembling system. Evaluated on the hidden validation set, our solution achieved an average LesionWise DSC of 0.897 with an average LesionWise NSD of 0.541 and 0.84 at a tolerance of 0.5 mm and 1.0 mm, respectively.

Problem

• Automates brain tumor segmentation in low-resource MRI settings
• Addresses poor segmentation due to low-quality MRI scans
• Optimizes model for sub-Saharan Africa's healthcare constraints

Innovation

• Larger region of interest for segmentation
• Improved nnU-Net v2-based architectural skeleton
• Robust model ensembling system for accuracy