A Web-based Brain Metastases Segmentation and Labeling Platform for Stereotactic Radiosurgery.

PURPOSE: Stereotactic radiosurgery (SRS) serves as a standard of care of brain metastases (BMs), however, the BMs delineation in the SRS workflow can be time-consuming. The manual contouring can be a pronounced bottleneck in multiple BMs, but there is a lack of tools for automatic delineation and quantitative evaluation. In this study, based on our previous developed deep learning-based segmentation algorithms, we developed a web-based automated BMs segmentation and labeling platform to assist the SRS clinical workflow. METHOD: This platform is developed based on the Django framework, including a web client and a back-end server. The web client enables interactions as database access, data import, image viewing. The server performs the segmentation and labeling tasks including: (1) skull stripping; (2) deep learning-based BMs segmentation; (3) affine registration-based BMs labeling. Afterwards the client can display BMs contours with corresponding atlas labels, and allows further post-processing tasks including: (1) change window level; (2) display/hide specific contours; (3) remove false-positive contours; (4) export contours as DICOM RTStruct files; etc. RESULTS: We evaluate this platform on 10 clinical cases with BMs number varied from 12-81. The overall operation takes about 4-5 minutes per patient. The segmentation accuracy is evaluated between the manual contour and automatic segmentation with averaged center of mass shift as 1.55+/-0.36 mm, Hausdorff distance as 2.98+/-0.63 mm, the mean of surface-to-surface distance (SSD) as 1.06+/-0.31 mm and the standard deviation of SSD as 0.80+/-0.16 mm, and the initial averaged false-positive over union (FPoU) and false-negative rate (FNR) as 0.43+/-0.19 and 0.15+/-0.10, respectively. After case-specific post-processing, the averaged FPoU and FNR are 0.19+/-0.10 and 0.15+/-0.10, respectively. CONCLUSION: A web-based BMs segmentation and labeling platform is developed and evaluated. Compared to manual segmentation/labeling, it can substantially improve the clinical efficiency. This platform can be a useful tool for assisting SRS treatment planning and treatment follow-up.