Bone X-ray and French Reports Dataset

该数据集包含了219,675项研究，每项研究对应一次放射学检查，包括一份法语报告和一份或多份X光图像，总计789,397张独立的X光图像。此外，还包括一个由4,096项研究组成的验证集，这些研究在训练过程中被排除在外。数据集的创建采用了来自DEDUCE框架的伪匿名化方法。该数据集是从75,600名患者中精心筛选出来的，旨在支持视觉-语言预训练及包括骨关节炎量化、骨龄估计和骨折检测在内的下游任务。

This dataset contains 219,675 studies, each corresponding to a radiological examination, including a French-language report and one or more X-ray images, with a total of 789,397 independent X-ray images. In addition, it includes a validation set consisting of 4,096 studies that were excluded during the training process. The dataset was created using the de-identification method from the DEDUCE framework. This dataset was carefully screened from 75,600 patients, and is intended to support vision-language pre-training and downstream tasks including osteoarthritis quantification, bone age estimation, and fracture detection.