迈向可靠的儿科脑肿瘤分割：针对任务的 nnU-Net 增强
Towards Reliable Pediatric Brain Tumor Segmentation: Task-Specific nnU-Net Enhancements
AI总结：
针对儿科脑肿瘤分割数据有限等挑战，提出专为 BraTS 2025 Task-6 定制的增强型 nnU-Net 框架，引入加宽残差编码器、3D 深度可分离卷积等改进及后处理，在验证排行榜上获第一名。
摘要：
多参数磁共振成像（mpMRI）中儿科脑肿瘤的准确分割对诊断、治疗规划和监测至关重要，但由于数据有限、解剖变异性高以及各机构成像异质性而面临独特挑战。在本研究中，我们提出了一个专为 BraTS 2025 Task-6 (PED) 定制的先进 nnU-Net 框架，这是最大的治疗前儿科高级别胶质瘤公开数据集。我们的贡献包括：(1) 带有挤压与激励（SE）注意力的加宽残差编码器；(2) 3D 深度可分离卷积；(3) 特异性驱动的正则化项；(4) 小规模高斯权重初始化。我们进一步通过两个后处理步骤优化预测结果。我们的模型在 Task-6 验证排行榜上获得第一名，病灶级 Dice 得分为 0.759 (CC)、0.967 (ED)、0.826 (ET)、0.910 (NET)、0.928 (TC) 和 0.928 (WT)。
英文摘要：
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).
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
