General-Medical-AI/UniMed-5M

UniMed-5M是一个统一的多模态医学数据集，用于训练和评估UniMedVL模型，该模型能够在8种医学成像模态中进行医学图像生成。数据集包含多个部分，目前公开的Part 1主要涉及生成和合成任务，包括IXI数据集（MRI超分辨率）、SynthRAD2023（CT-MR合成）、BraTS 2023（MRI模态转换）、DRIVE（视网膜血管分割）和BCI (HE2IHC)（虚拟染色）等子数据集。数据集以Parquet格式提供元数据文件，原始图像需要从官方来源下载。数据集的组织结构、包含的子数据集及其详细信息、数据准备步骤、使用方法和许可信息均在README中有详细描述。

UniMed-5M is a unified medical multimodal dataset used for training and evaluating UniMedVL, a unified medical multimodal model for medical image generation across 8 medical imaging modalities. The dataset is organized into parts, with Part 1 currently available, focusing on generation and synthesis tasks, including sub-datasets such as IXI Dataset (MRI super-resolution), SynthRAD2023 (CT-MR synthesis), BraTS 2023 (MRI modality translation), DRIVE (retinal vessel segmentation), and BCI (HE2IHC) (virtual staining). The dataset provides metadata files in Parquet format, and original images must be downloaded from official sources. The README details the datasets structure, included sub-datasets, data preparation steps, usage, and licensing information.