Citation

Barakat, Mohtady Ehab and Salah, Omar and Yasser, Ahmed and Ahmed, Mostafa and Arief, Zahirul and Khan, Waleed and Zhang, Dongdong and Iorumbur, Aondona and Raymond, Confidence and Barakat, Mohannad and Magdy, Noha (2026) Training Beyond Convergence: Grokking nnU-Net for Glioma Segmentation in Sub-Saharan MRI. In: Brain TumorS Lighthouse Cluster of Challenges, and the Automated Identification of Moderate-Severe Traumatic Brain Injury Lesions Challenge, BraTS 2025 and AIMS-TBI 2025, held in Conjunction International Conference on Medical Image Computing and Computer, 23 - 27 September 2025, Daejeon.

Text Training Beyond Convergence_ Grokking nnU-Net for Glioma Segmentation in Sub-Saharan MRI _ Springer Nature Link.pdf - Published Version Restricted to Repository staff only Download (855kB)

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).

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Diagnostic imaging,glioma
Subjects:	R Medicine > R Medicine (General) > R856-857 Biomedical engineering. Electronics. Instrumentation
Divisions:	Faculty of Artificial Intelligence & Engineering (FAIE)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	05 Jun 2026 03:13
Last Modified:	08 Jun 2026 09:27
URII:	http://shdl.mmu.edu.my/id/eprint/15994

Downloads

Edit (login required)