MMIS2024

这一挑战通过专注于从MRI数据中分割多样化和个性化的总肿瘤体积(GTV)来解决多重问题，这对于鼻咽癌和胶质母细胞瘤的放疗计划至关重要。任务1数据集：给定三个MRI序列(T1, T2和T1- contrast)，并将一个样本的不同序列刚性注册到一个共同的受试者空间。对于每个样本，来自不同地区的4名资深放射科医生(经验约5-10年)分别注释了鼻咽癌的GTV。我们将170个主题分为单独的训练集、验证集和测试集(即分别为100、20和50个)。任务2数据集：来自三个公开可用的存储库(即LUMIERE, RHUH, UPENN-GBM)的胶质母细胞瘤成像数据集，并进行了术后和复发MRI扫描。每个MRI研究包括四个序列:t1加权，t2加权，FLAIR, t1加权钆造影剂。四名注释者(两名放射肿瘤学家和两名放射学住院医师)在T1对比增强序列上对肿瘤进行分割。训练集:120次扫描在注释器之间平均分配，即每个注释器30次扫描。评估集:由每个注释者标记另外20个扫描。

This challenge addresses multiple clinical problems by focusing on segmenting diverse and personalized gross tumor volumes (GTV) from MRI data, which is critical for radiotherapy planning of nasopharyngeal carcinoma (NPC) and glioblastoma (GBM). Task 1 Dataset: Given three MRI sequences (T1, T2, and T1-contrast), the distinct sequences of a single sample are rigidly registered to a common subject space. For each sample, four senior radiologists (with approximately 5–10 years of experience) from different regions independently annotated the GTV of NPC. We split 170 subjects into separate training, validation, and test sets, with 100, 20, and 50 subjects respectively. Task 2 Dataset: Glioblastoma imaging datasets sourced from three publicly available repositories, namely LUMIERE, RHUH, and UPENN-GBM, which include post-operative and recurrent MRI scans. Each MRI study consists of four sequences: T1-weighted, T2-weighted, FLAIR, and T1-weighted gadolinium-enhanced. Four annotators (two radiation oncologists and two radiology residents) segmented the tumors on the T1 contrast-enhanced sequence. Training set: 120 scans were evenly allocated across the four annotators, with 30 scans per annotator. Evaluation set: An additional 20 scans labeled by each annotator.

搜集汇总

背景与挑战

背景概述

MMIS2024数据集专注于鼻咽癌和胶质母细胞瘤的肿瘤体积分割，包含多序列MRI数据和多位专家的注释，分为训练、验证和测试集以提高模型的泛化能力。