Brain-arteries segmentation on CTA, MR-TOF and CE-T1
Per-artery segmentation of the cerebrovascular tree on CTA, MR-TOF and CE-T1 to train a vascular segmentation model for a mechanical thrombectomy micro-robot.
A medtech company developing a micro-robot for automated mechanical thrombectomy needed a large database of segmented brain arteries on both CT and MR images to train a vascular segmentation model used for procedure planning and execution.
Correct a model-based pre-segmentation and produce accurate per-artery labels for up to 25 arteries (aorta, supra-aortic trunks, anterior and posterior circulation), following artery-specific rules — including calcifications kept as part of the artery, stents excluded but lumen kept, and aneurysms segmented under the affected artery.
Radiologists worked on RedBrick with a pre-segmentation loaded per case, focused on the intracranial arteries (siphon to carotid T, ophthalmic, ACA / MCA with M1–M2 segments, ACoA, PCA, PComA). Pen and Edge Selection tools were used at high zoom for clean contours, and cases were routed through a technical review by engineers followed by an independent medical review before ground-truth promotion.
The client obtained validated per-artery segmentations across CTA, MR-TOF and CE-T1, with consistent left/right labelling and artery-specific edge cases handled — feeding directly into the vascular model used by the robotic thrombectomy platform.