Multi-vessel segmentation on CT angiography
Per-vessel labeled segmentation of supra-aortic trunks, aorta and lower-limb arteries on CTA studies.
A radiology platform provider was building an AI vessel segmentation module and needed a large, balanced dataset of contrast-enhanced CT studies covering supra-aortic trunks, aorta and lower-limb vasculature, with representative distributions of manufacturers, demographics and pathologies.
Produce per-vessel labeled segmentations across the mandatory anatomy, with a dedicated class for thrombus, respecting the client's numeric label schema and DICOM-SEG output format.
Experienced radiologists (5+ years) were assigned exclusively to the project, trained on the client's label schema and continuity rules at bifurcations. Optional vessels were added incrementally. All segmentations were validated against the requirements list before being packaged as DICOM-SEG files.
The client received a standards-compliant, per-vessel dataset ready for algorithm training and evaluation, with balanced representation across acquisition systems and pathologies as specified in the protocol.