Multi-Modal Dataset for Federated Learning with DICOM Structured Reports
收藏资源简介:
该数据集由德国心血管研究中心及其合作伙伴创建,旨在通过DICOM结构化报告整合多种医疗数据类型,支持联邦学习。数据集内容包括CT图像、心电图扫描、患者元数据等,通过统一的DICOM表示和过滤选项实现数据集的标准化和一致性。创建过程中,各机构从其临床信息系统中导出相关数据,并转换为DICOM格式以实现数据集成和匹配。该数据集主要应用于预测微创心脏瓣膜置换后的结果,通过多模态数据的整合和分析,提高深度学习模型在临床实践中的预测性能。
This dataset was created by the German Centre for Cardiovascular Research and its partners, aiming to integrate multiple types of medical data via DICOM structured reporting and support federated learning. The dataset includes CT images, electrocardiogram (ECG) scans, patient metadata, and other relevant data, and achieves standardization and consistency through unified DICOM representation and filtering options. During its development, each participating institution exported relevant data from its own clinical information system and converted the data into DICOM format to enable data integration and matching. This dataset is primarily used to predict post-operative outcomes following minimally invasive cardiac valve replacement, and it improves the predictive performance of deep learning models in clinical practice by integrating and analyzing multimodal data.




