Segmentation

The most common central nervous system tumor types can be segmented pre-/post-operatively: glioblastomas, diffuse lower-grade gliomas, meningiomas, and metastases. Common structures are available for segmentation such as enhancing tissue, non-enhancing tumor core, tumor core, FLAIR hyperintensity, and resection cavity

All models have been trained locally using patient data from our partnered hospitals, and using the Attention U-Net architecture (more details available in the Publications section).

Standardized reporting (RADS)

Features computation, the standardized reports include the following tumor characteristics, all computed in the MNI reference space: volumes, diameters, multifocality, laterality, location w.r.t. cortical structures (MNI, Harvard-Oxford, and Schaefer atlases), and location w.r.t. subcortical structures (BCB atlas). The computational pipeline is illustrated to the left.

Surgical reporting: brain and tumor volumes pre-/post-operatively are reported, including extent of resection and surgical grading following the RANO2.0 guidelines.

Hospital partners

Northwest Clinics (Alkmaar), Amsterdam University Medical Centers, University Medical Center Groningen, Medical Center Haaglanden (The Hague), Humanitas Research Hospital (Milano), Hôpital Lariboisière (Paris), University of California San Francisco Medical Center, Medical Center Slotervaart (Amsterdam), St Elisabeth Hospital (Tilburg), University Medical Center Utrecht, Medical University Vienna, Isala hospital (Zwolle), St. Olavs hospital (Trondheim), Sahlgrenska University Hospital (Gothenburg), Brigham and Women's Hospital (Boston), and Oslo University Hospital.