SAT-DS 大规模 3D 医学图像分割数据集

该数据集为上海交通大学团队于 2024 年构建的当前最大规模的 3D 医学图像分割数据集。它汇集了 72 个公开数据集，来自 CT 、 MR 和 PET 三种模态的 22K+ 图像， 302K+ 分割标注，涵盖了人体 8 个主要部位中的 497 个分割目标，实现了通过文本提示的放射学影像通用医疗分割模型。 相关研究以「One Model to Rule them All: Towards Universal Segmentation for Medical Images with Text Prompts」发表在学术平台 arXiv.org 上。

This dataset is the largest-scale 3D medical image segmentation dataset to date, developed by a team at Shanghai Jiao Tong University in 2024. It integrates 72 public datasets, comprising over 22,000 images across three modalities (CT, MR, and PET) and more than 302,000 segmentation annotations. Covering 497 segmentation targets across 8 major human anatomical regions, this dataset enables the construction of a universal medical segmentation model for radiological images powered by text prompts. The corresponding research was published on the academic platform arXiv.org under the title "One Model to Rule them All: Towards Universal Segmentation for Medical Images with Text Prompts".

搜集汇总

背景与挑战

背景概述

SAT-DS是由上海交通大学团队于2024年构建的大规模3D医学图像分割数据集，整合了72个公开数据集，包含来自CT、MR和PET模态的22,000多张图像和302,000多个分割标注，覆盖人体8个主要部位的497个分割目标，旨在通过文本提示实现通用医疗分割模型。