Matched MRI, Segmentations, and Histopathologic Images of Brain Metastases from Primary Lung Cancer

Brain metastases occur in approximately 30-40% of lung cancer patients, and outcomes are heterogeneous, dependent on a combination of clinical, radiologic, pathologic, and molecular characteristics. Though artificial intelligence-based prognostic models show promise, their use is limited by a lack of high-quality training data. Thus, there is a critical need for a large, heterogeneous, annotated, open-access brain metastasis dataset with matched radiologic and histopathologic imaging. We present such a dataset composed of 111 cases of magnetic resonance (MR) and histopathologic imaging from patients with brain metastasis from primary lung cancer. We provide pre-operative T1-weighted contrast-enhanced (T1CE) and fluid-attenuated inversion recovery (FLAIR) MR images with matched whole slide images of formalin-fixed, paraffin-embedded (FFPE) brain metastasis biopsies. The dataset also includes segmentations of contrast enhancement and FLAIR hyperintensity, radiomic features, and clinical information. A Kaplan-Meier analysis validated that the dataset's patients are consistent with prior large retrospective studies of patients with brain metastasis from lung cancer. The provided dataset will facilitate the development of multimodal models for improved management of lung cancer patients.