US-43d
收藏资源简介:
US-43d是由斯特拉斯堡大学和斯特拉斯堡图像引导手术研究所创建的大型公开超声分割数据集,包含43个公开数据集,共计280,000多张图像和分割掩码,覆盖50多个解剖结构。数据集内容丰富,包括2D和3D扫描图像,涉及心脏、胎儿头部、甲状腺和乳腺病变等多种临床应用。数据集的创建过程涉及从多个平台爬取数据,并进行预处理以去除标签背景重叠。US-43d旨在解决超声图像分析中的自动分割问题,提供了一个强大的基础模型,适用于多种下游任务,如分类和分割。
US-43d is a large-scale publicly available ultrasound segmentation dataset created by the University of Strasbourg and the Strasbourg Institute for Image-Guided Surgery. It encompasses 43 publicly available datasets, with over 280,000 images and segmentation masks, covering more than 50 anatomical structures. The dataset is rich in content, including 2D and 3D scan images, and is applicable to various clinical applications such as heart, fetal head, thyroid, and breast lesion analysis. The creation process of the dataset involved data crawling from multiple platforms and preprocessing to remove label background overlaps. US-43d aims to address the automatic segmentation problem in ultrasound image analysis, providing a powerful foundation model suitable for a variety of downstream tasks, such as classification and segmentation.

- 1UltraSam: A Foundation Model for Ultrasound using Large Open-Access Segmentation Datasets斯特拉斯堡大学,CNRS,INSERM,ICube,UMR7357,斯特拉斯堡,法国;斯特拉斯堡图像引导手术研究所,斯特拉斯堡,法国 · 2024年



