"BTXRD-2024 Augmented Bone Tumor Segmentation and Triage Dataset (Balanced Malignant Class)"

"The dataset is based on the publicly available BTXRD-2024 collection of musculoskeletal radiographs, but we extended it to deal with the fact that malignant tumor cases were originally rare. It now contains three categories\u2014Normal, Benign, and Malignant\u2014and each image has both lesion masks and a case-level triage label. To make the dataset more balanced and useful for training, we carried out several preprocessing steps, including fixing irregular polygon masks and adding augmented samples of malignant cases. With these changes, the dataset provides a fairer distribution across classes and can be used for studies on segmentation, classification, and triage. The goal is to give researchers a version of BTXRD-2024 that better reflects real clinical needs while still being consistent enough for benchmarking AI models in medical imaging."