Duke Liver Dataset (MRI)

This dataset contains images and supporting files for two corresponding liver MRI applications: (1) Liver segmentation application: 310 MRI image series (4 different contrast types) across 95 subjects are provided. Roughly 75% of series are accompanied by mask files corresponding to manual segmentations of the liver. Participants can use these images and masks to train an automated liver segmentation model. The remaining 25% of series can be used as the test set for these models. (2) Series classification application: 2146 MRI image series across 105 subjects are provided. An alphabetical series label is provided for roughly 75% of the series and withheld for the other 25%. Participants can use these images and labels to train an automated series classification model for liver MRI. The models can be tested on the unlabeled 25% of series.

本数据集包含面向两类肝脏磁共振成像（MRI）应用的图像及配套文件：(1) 肝脏分割应用：涵盖95名受试者的310个MRI图像序列（含4种不同对比类型），其中约75%的序列配有对应肝脏手动分割的掩码文件。参与者可利用此类图像与掩码训练自动化肝脏分割模型，剩余25%的序列可作为此类模型的测试集。(2) 序列分类应用：涵盖105名受试者的2146个MRI图像序列，其中约75%的序列配有按字母顺序编排的序列标签，剩余25%的标签未予公开。参与者可借助这些图像与标签训练面向肝脏MRI的自动化序列分类模型，并可在未标注的25%序列上完成模型测试。

搜集汇总

背景与挑战

背景概述

Duke Liver Dataset (MRI) 是一个专注于肝脏磁共振成像（MRI）的数据集，发布于2020年，包含两个主要应用：肝脏分割和序列分类。数据集提供310个MRI图像系列（来自95名受试者，涵盖4种对比类型）用于分割任务，其中75%带有手动分割掩码；同时提供2146个MRI图像系列（来自105名受试者）用于分类任务，75%带有序列标签。数据集受限制访问，需请求许可，适用于非商业研究或通过自定义许可用于商业用途。