The Burdenko's Glioblastoma Progression Dataset (B-GBM-PD) is a systematic data collection of pre-radiotherapy MRI images of 180 patients with primary glioblastoma treated at the Burdenko National Medical Research Center of Neurosurgery between 2014 and 2020. We provide four MRI sequences for all patients - T1, T1c, T2, FLAIR; topometric CT scan; and annotations of the GTV area made for radiotherapy planning. Additionally, the data collection contains follow-up studies (from 1 to 7-time points per patient). Each time-point includes 2-4 MRI sequences (with a minimal set of T1c, FLAIR per patient) with contours of the irradiated volumes. The RTDOSE, RTPLAN files; additional genetic information (IDH1/2, MGMT mutations); and treatment response status (tumor progression, tumor pseudoprogression, treatment response) are available for a subset of subjects. MRI studies were performed on various scanners from different vendors. CT studies were conducted in Burdenko National Medical Research Center of Neurosurgery on a single scanner.

Dataset was collected in the radiation therapy department. The benchmark on the current dataset might be closer to clinical practice owing to data minimal preprocessing, heterogeneity in scanning modalities, sufficient size of the dataset, and longitudinal patient imaging data. It is a novel dataset compared to other TCIA collections. It contains four MRI imaging sequences for each patient (and a CT scan) and manual annotations of the glioblastoma gross tumor volumes (GTV) for radiation treatment planning, thus, complementing BraTS and TCIA-GBM datasets for brain tumor segmentation. The task of delineating GTV currently lacks large publicly available data. Availability of follow-up images allows for assessment of treatment response and development of disease progression models. We provide biological and clinical information along with imaging data: sex, age, idh1/2, mgmt mutations, and treatment response.