Training Beyond Convergence: Grokking nnU-Net for Glioma Segmentation in Sub-Saharan MRI

2026-01-30

AI Summary

This study addresses the challenges of scarce MRI data and poor survival outcomes for glioma patients in sub-Saharan Africa by proposing a two-stage training strategy based on the nnU-Net architecture using the BraTS Africa 2025 dataset. The approach first establishes an efficient baseline through short-cycle training under resource constraints, followed by ultra-long overconvergent training to elicit the “grokking” phenomenon—a behavior previously unexplored in medical image segmentation. This work is the first to validate and leverage grokking in this domain, significantly enhancing model generalization without additional annotations. Experimental results demonstrate superior performance in segmenting the tumor core and enhancing regions, achieving Dice scores of 90.1% and 90.2%, respectively, and 92.2% for the whole tumor, outperforming conventional training protocols.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Gliomas are placing an increasingly clinical burden on Sub-Saharan Africa (SSA). In the region, the median survival for patients remains under two years, and access to diagnostic imaging is extremely limited. These constraints highlight an urgent need for automated tools that can extract the maximum possible information from each available scan, tools that are specifically trained on local data, rather than adapted from high-income settings where conditions are vastly different. We utilize the Brain Tumor Segmentation (BraTS) Africa 2025 Challenge dataset, an expert annotated collection of glioma MRIs. Our objectives are: (i) establish a strong baseline with nnUNet on this dataset, and (ii) explore whether the celebrated"grokking"phenomenon an abrupt, late training jump from memorization to superior generalization can be triggered to push performance without extra labels. We evaluate two training regimes. The first is a fast, budget-conscious approach that limits optimization to just a few epochs, reflecting the constrained GPU resources typically available in African institutions. Despite this limitation, nnUNet achieves strong Dice scores: 92.3% for whole tumor (WH), 86.6% for tumor core (TC), and 86.3% for enhancing tumor (ET). The second regime extends training well beyond the point of convergence, aiming to trigger a grokking-driven performance leap. With this approach, we were able to achieve grokking and enhanced our results to higher Dice scores: 92.2% for whole tumor (WH), 90.1% for tumor core (TC), and 90.2% for enhancing tumor (ET).

Problem

• glioma segmentation
• Sub-Saharan Africa
• MRI
• automated diagnosis
• limited imaging resources

Innovation

• grokking
• nnU-Net
• glioma segmentation
• Sub-Saharan MRI
• training beyond convergence