Masks for semantic segmentation of hands from scanned X-Rays in the RSNA Bone Age dataset (released for the RSNA Pediatric Bone Age Challenge in 2017). The masks were obtained manually using thresholding and edge detection and all masks were quality checked and, if needed, corrected. Based on this two models (Tensormask and Efficient-UNet) were trained to obtain the masks on the full RSNA Bone Age dataset. If you use this dataset for your work, please cite the paper this dataset is part of: Rassmann, S., Keller, A., Skaf, K. et al. Deeplasia: deep learning for bone age assessment validated on skeletal dysplasias. Pediatr Radiol 54, 82–95 (2024). https://doi.org/10.1007/s00247-023-05789-1