Synthetic 3D Echocardiography Dataset
收藏资源简介:
本研究通过GAN技术生成了一个包含27个样本的3D心脏超声图像数据集。该数据集由真实的3D心脏超声图像和对应的手动标注的解剖标签组成,用于训练生成对抗网络以合成逼真的3D心脏超声图像。数据集的创建过程涉及使用高分辨率的CT心脏模型作为输入,通过GAN生成具有相应解剖标签的合成图像。该数据集主要应用于训练深度学习模型,特别是用于心脏结构的分割任务,以解决现有3D心脏超声图像数据集的稀缺问题。
In this study, a 3D cardiac ultrasound image dataset with 27 samples was generated via GAN technology. This dataset consists of real 3D cardiac ultrasound images and their corresponding manually annotated anatomical labels, and is designed for training generative adversarial networks to synthesize realistic 3D cardiac ultrasound images. The dataset development process employs high-resolution cardiac CT models as inputs to generate synthetic images with corresponding anatomical labels using GANs. This dataset is mainly used to train deep learning models, particularly for cardiac structure segmentation tasks, to address the scarcity of existing 3D cardiac ultrasound image datasets.

- 1A Data Augmentation Pipeline to Generate Synthetic Labeled Datasets of 3D Echocardiography Images using a GANGE Vingmed Ultrasound - GE Healthcare · 2024年



