Background Three-dimensional T1 magnetization prepared rapid acquisition gradient echo (3D-T1-MPRAGE) is preferred in detecting brain metastases (BM) among MRI. We developed an automatic deep learning–based detection and segmentation method for BM (named BMDS net) on 3D-T1-MPRAGE images and evaluated its performance. Methods The BMDS net is a cascaded 3D fully convolution network (FCN) to automatically detect and segment BM. In total, 1652 patients with 3D-T1-MPRAGE images from 3 hospitals (n = 1201, 231, and 220, respectively) were retrospectively included. Manual segmentations were obtained by a neuroradiologist and a radiation oncologist in a consensus reading in 3D-T1-MPRAGE images. Sensitivity, specificity, and dice ratio of the segmentation were evaluated. Specificity and sensitivity measure the fractions of relevant segmented voxels. Dice ratio was used to quantitatively measure the overlap between automatic and manual segmentation results. Paired samples t-tests and analysis of variance were employed for statistical analysis. Results The BMDS net can detect all BM, providing a detection result with an accuracy of 100%. Automatic segmentations correlated strongly with manual segmentations through 4-fold cross-validation of the dataset with 1201 patients: the sensitivity was 0.96 ± 0.03 (range, 0.84–0.99), the specificity was 0.99 ± 0.0002 (range, 0.99–1.00), and the dice ratio was 0.85 ± 0.08 (range, 0.62–0.95) for total tumor volume. Similar performances on the other 2 datasets also demonstrate the robustness of BMDS net in correctly detecting and segmenting BM in various settings. Conclusions The BMDS net yields accurate detection and segmentation of BM automatically and could assist stereotactic radiotherapy management for diagnosis, therapy planning, and follow-up.

The cytological origin of central nervous system hemangioblastoma (HB) remains unclear and controversial, largely owing to a lack of in-depth characterization of tumorigenic cells and their progeny tracking. We have now detected a cell subpopulation by stage-specific embryonic antigen-1 expression, which were defined as tumor-initiating cells (TICs) in both sporadic and familial HBs. These TICs subpopulations had universal neural stem cell characteristics. Nevertheless, the freshly sorted TICs endowed with potential of multi-progeny derivatives, including HB components and non-HB ingredients, depended on environmental induction in vitro. Importantly, the freshly harvested TICs formed malignant tumors by injection into conventional mice model, while did redevelop the characteristic HB-like structures within a special mice model with HB-microenvironment, indicating HB niche dependency for the TICs derivative specification. Taken together, the data of the present study suggested that HBs might derive from neoplastic transformation of neural stem cells/progenitors in the specific niche.

Our study showed that GKS is a useful and safe therapeutic method for CaSHs as both a primary and adjuvant treatment. The new classification of CaSHs may help predict their clinical course during tumor development and treatment response after GKS. Further studies with long-term follow-up and larger numbers of cases are necessary to optimize the treatment conditions and verify the benefit of this treatment.