Angelou0516/HNTSMRG
收藏资源简介:
HNTSMRG 2024数据集是MICCAI挑战赛的公开训练数据,主要用于头颈部肿瘤的MRI图像分割。数据集包含150名患者的T2加权MRI图像,每个患者有两个时间点的数据(放疗前和放疗中),并提供了专家共识的分割标签。数据集的模态为T2加权解剖MRI(混合脂肪抑制和非脂肪抑制),身体部位为头颈部(口咽癌和原发灶不明的癌症)。标签包括背景(0)、原发肿瘤(GTVp,1)和受累淋巴结(GTVn,2)。注释由3-4名专家医师完成,并通过STAPLE共识融合(在极端分歧情况下由教师仲裁)。数据集总大小约为15 GB,许可证为CC-BY-NC-4.0,来源为Zenodo。数据集支持两个分割任务:放疗前分割(基线MRI上的GTVp和GTVn)和放疗中分割(治疗中MRI上的GTVp和GTVn,并提供放疗前数据作为先验)。
The HNTSMRG 2024 dataset is the public training data for the MICCAI Challenge, primarily used for MRI image segmentation of head and neck tumors. The dataset includes T2-weighted MRI images from 150 patients, with data from two timepoints (pre-radiotherapy and mid-radiotherapy), and provides expert-consensus segmentation labels. The modality of the dataset is T2-weighted anatomical MRI (a mix of fat-suppressed and non-fat-suppressed), and the body part is the head and neck (oropharyngeal cancer and cancer of unknown primary). Labels include background (0), primary tumor (GTVp, 1), and involved lymph nodes (GTVn, 2). Annotations were performed by 3-4 expert physicians and fused via STAPLE consensus (with faculty arbitration in extreme disagreement cases). The total size of the dataset is approximately 15 GB, licensed under CC-BY-NC-4.0, and sourced from Zenodo. The dataset supports two segmentation tasks: pre-RT segmentation (GTVp and GTVn on baseline MRI) and mid-RT segmentation (GTVp and GTVn on intra-treatment MRI, with pre-RT data provided as a prior).




