Segmentation Labels for the Pre-operative Scans of the TCGA-GBM collection

This data container describes both computer-aided and manually-corrected segmentation labels for the pre-operative multi-institutional scans of The Cancer Genome Atlas (TCGA) Glioblastoma Multiforme (GBM) collection, publicly available in The Cancer Imaging Archive (TCIA), coupled with a rich panel of radiomic features along with their corresponding skull-stripped and co-registered multimodal (i.e. T1, T1-Gd, T2, T2-FLAIR) magnetic resonance imaging (MRI) volumes in NIfTI format. Pre-operative multimodal MRI scans were identified in the TCGA-GBM collection via radiological assessment. These scans were initially skull-stripped and co-registered, before their tumor segmentation labels were produced by an automated hybrid generative-discriminative method, ranked first during the International multimodal BRAin Tumor Segmentation challenge (BRATS 2015). These segmentation labels were revised and any label misclassifications were manually corrected by an expert board-certified neuroradiologist. The final labels were used to extract a rich panel of imaging features, including intensity, volumetric, morphologic, histogram-based and textural parameters, as well as spatial information and diffusion properties extracted from glioma growth models. The generated computer-aided and manually-revised labels enable quantitative computational and clinical studies without the need to repeat manual annotations whilst allowing for comparison across studies. They can also serve as a set of manually-annotated gold standard labels for performance evaluation in computational challenges. The provided panel of radiomic features may facilitate research integrative of the molecular characterization offered by TCGA, and hence allow associations with molecular markers, clinical outcomes, treatment responses and other endpoints, by researchers without sufficient computational background to extract such features.

本数据集包含经计算机辅助生成并经人工校正的分割标签，对应癌症基因组图谱（The Cancer Genome Atlas, TCGA）多形性胶质母细胞瘤（Glioblastoma Multiforme, GBM）队列的术前多中心扫描数据，该扫描数据公开收录于癌症影像档案（The Cancer Imaging Archive, TCIA）；同时本数据集还附带丰富的放射组学特征集，以及与上述标签匹配的经颅骨剥离（skull-stripped）、配准（co-registered）后的多模态磁共振成像（magnetic resonance imaging, MRI）序列影像，涵盖T1、T1-Gd、T2、T2-FLAIR序列，所有影像均采用NIfTI格式存储。本研究通过放射学评估从TCGA-GBM队列中筛选出术前多模态MRI扫描数据，所有扫描数据首先完成颅骨剥离与配准处理，随后采用在2015年国际多模态脑肿瘤分割挑战赛（International multimodal BRAin Tumor Segmentation challenge, BRATS 2015）中夺冠的自动化混合生成-判别式方法生成肿瘤分割标签。此后，经认证的神经放射学专家（board-certified neuroradiologist）对上述分割标签进行复核，并手动修正所有标签分类错误。最终修正后的标签被用于提取丰富的影像特征，涵盖强度、体积、形态学、基于直方图的纹理参数，以及从胶质瘤生长模型中提取的空间信息与扩散特性。本数据集提供的计算机辅助生成并经人工修订的分割标签，可免去研究者重复开展人工标注的工作，支持量化计算与临床研究，同时便于不同研究间的结果对比；该标签集同时可作为经人工标注的金标准标签（gold standard labels），用于计算类挑战赛中的算法性能评估。本数据集附带的放射组学特征集，可帮助不具备足够计算能力的研究者整合TCGA提供的分子表征（molecular characterization）数据，进而探索分子标志物（molecular markers）、临床结局（clinical outcomes）、治疗响应（treatment responses）及其他研究终点之间的关联。