BraTioUS - Brain Tumor Intraoperative Ultrasound Dataset

The BraTioUS (Brain Tumor Intraoperative Ultrasound) dataset is a large-scale, multicenter, and publicly available collection of intraoperative ultrasound (iUS) images acquired during glioma surgeries. Created through an international collaboration among six hospitals across five countries, BraTioUS includes 1,669 B-mode 2D iUS images from 142 glioma patients, collected between 2018 and 2023 using various ultrasound systems and acquisition protocols. All images have been anonymized using a deep learning-based nnU-Net segmentation model, and corresponding tumor segmentations are provided. 

B-mode intraoperative ultrasound images were acquired during glioma resections performed in: Rio Hortega University Hospital, Valladolid, Spain (RHUH); Tata Memorial Center, Mumbai, India (TMC); Le Centre Hospitalier Régional Universitaire de Tours, France (CHRUT); Carlo Besta Neurological Institute, Milan, Italy (INCC); University of Palermo, Italy (UPALER); and Massachusetts General Hospital, Boston, USA (MGH).

Ultrasound images were acquired using sterilizable neuro-cranial probes on cart-based ultrasound systems from various manufacturers and models (Hitachi©, Esaote©, BK©, Supersonic© and Sonowand©). Probes used to obtain iUS images were linear or curve according to manufacturer. Image acquisition planes were oriented as parallel as possible to standard anatomical axes, subject to craniotomy constraints. Images with suboptimal quality or artifacts impeding interpretation were excluded. All data were anonymized and standardized using an automatic nnU-Net–based segmentation model trained on manually annotated cases, removing acquisition-related and patient-identifying elements. BraTioUS dataset also includes segmentation mask of tumor tissue that appears in 1,669 images. Final datasets comprise NIfTI-format 2D ultrasound images and corresponding tumor segmentations and a .xlsx file that includes metadata of dataset, all of them previously converted in ZIP format.