3D-MIR

3D-MIR数据集是由微软健康与生命科学团队创建的，专注于3D医学图像检索的基准数据集。该数据集包含281个来自CT扫描的3D图像，涵盖肝脏、肺、胰腺和结肠四个器官系统。数据集的构建过程涉及使用TotalSegmentator模型进行器官分割，并通过OpenAI的GPT模型生成器官病变的详细描述。3D-MIR数据集旨在支持临床决策，通过提供精确的3D图像检索，帮助医生快速找到与特定病例相关的历史病例，从而提高诊断的准确性和效率。

The 3D-MIR dataset was developed by the Microsoft Health & Life Sciences team as a benchmark dataset dedicated to 3D medical image retrieval. It contains 281 3D images sourced from CT scans, covering four organ systems: the liver, lung, pancreas, and colon. The construction of this dataset involved two key processes: first, performing organ segmentation using the TotalSegmentator model, and second, generating detailed descriptions of organ lesions via OpenAI's GPT models. The 3D-MIR dataset aims to support clinical decision-making: by providing precise 3D medical image retrieval, it helps clinicians quickly locate historical cases relevant to specific current cases, thereby improving the accuracy and efficiency of diagnosis.