Multi-center Paired 2D Ultrasound and CT Dataset

该数据集包含了来自9家医院的真实世界数据，包括严格配对的多视角二维超声和CT扫描，以及伪配对数据。具体来说，数据集包含了130个用于多视角重建的严格配对样本，以及1,379个用于隐式自动编码器训练的心脏模型。这些数据是在超声和CT检查间隔少于10天内前瞻性收集的。数据集的规模为891个配对案例（其中96个为ECG门控，795个为非严格配对）。该数据集的任务是生成心脏解剖学的双胞胎模型。

This dataset contains real-world data from 9 hospitals, including strictly paired multi-view 2D ultrasound and CT scans, as well as pseudo-paired data. Specifically, the dataset includes 130 strictly paired samples for multi-view reconstruction and 1,379 cardiac models for implicit autoencoder training. All these data were prospectively collected with an interval of less than 10 days between ultrasound and CT examinations. The dataset consists of 891 paired cases, among which 96 are ECG-gated and 795 are non-strictly paired. The task of this dataset is to generate twin models of cardiac anatomy.

搜集汇总

背景与挑战

背景概述

该数据集包含50个患者病例的配对2D超声和CT扫描数据，超声图像为256×256像素PNG格式，CT扫描为1mm³各向同性NIfTI格式，数据已匿名化并在Zenodo平台公开。