Towards Reliable Pediatric Brain Tumor Segmentation: Task-Specific nnU-Net Enhancements

TL;DR

This paper introduces task-specific enhancements to the nnU-Net framework for pediatric brain tumor segmentation, achieving state-of-the-art results on the BraTS 2025 dataset by addressing unique pediatric imaging challenges.

Contribution

The paper presents novel nnU-Net modifications tailored for pediatric brain tumor segmentation, including a widened residual encoder with SE attention, depthwise convolutions, and regularization techniques.

Findings

Achieved top performance on BraTS 2025 Task-6 validation leaderboard.

Attained high lesion-wise Dice scores across multiple tumor regions.

Demonstrated the effectiveness of task-specific nnU-Net enhancements.

Abstract

Accurate segmentation of pediatric brain tumors in multi-parametric magnetic resonance imaging (mpMRI) is critical for diagnosis, treatment planning, and monitoring, yet faces unique challenges due to limited data, high anatomical variability, and heterogeneous imaging across institutions. In this work, we present an advanced nnU-Net framework tailored for BraTS 2025 Task-6 (PED), the largest public dataset of pre-treatment pediatric high-grade gliomas. Our contributions include: (1) a widened residual encoder with squeeze-and-excitation (SE) attention; (2) 3D depthwise separable convolutions; (3) a specificity-driven regularization term; and (4) small-scale Gaussian weight initialization. We further refine predictions with two postprocessing steps. Our models achieved first place on the Task-6 validation leaderboard, attaining lesion-wise Dice scores of 0.759 (CC), 0.967 (ED), 0.826 (ET), 0.910 (NET), 0.928 (TC) and 0.928 (WT).