Assessing nnU-Net Generalization across Brain Tumor Populations in BraTS-GoAT 2026
- 类型:arxiv
- 标识:2609.15524
- 链接:https://arxiv.org/abs/2609.15524
- 主分类:llm-infra
- 形态:method
- TLDR:BraTS-GoAT evaluates tumor segmentation across heterogeneous populations. We trained a conventional 3D nnU-Net on 1,351 labeled cases using five-fold cross-validation and 1,000 epochs per fold. The final predictor averaged all folds and applied test-time mirroring. On pooled official validation, global DSC values were 0.7805, 0.8288, and 0.8854 for enhancing tumor (ET), tumor core (TC), and whole tumor (WT). Under matched fold-0 inference, mean regional Dice decreased from 0.9058 on source out-of-fold (OOF) cases to 0.8310 on pooled validation (difference--0.0747). Mirroring gave small single-
- 待LLM分类:否
- 标题中文:在 BraTS-GoAT 2026 中评估 nnU-Net 跨脑肿瘤人群的泛化能力
- TLDR中文:BraTS-GoAT 在异构人群上评估肿瘤分割。我们使用 1,351 例标注病例训练了一个标准 3D nnU-Net,采用五折交叉验证,每折 1,000 个 epoch。最终预测器对所有折取平均,并应用测试时镜像增强。在汇总的官方验证集上,增强肿瘤(ET)、肿瘤核心(TC)和全肿瘤(WT)的全局 DSC 分别为 0.7805、0.8288 和 0.8854。在匹配的 fold-0 推理下,平均区域 Dice 从源域 OOF 病例的 0.9058 下降到汇总验证集的 0.8310(差异 -0.0747)。镜像带来小幅单次...
- 来源文件:
- /inbox/tom/_candidates/2026-09-17-agent-rag-longcontext-candidates.json