Aortic lumen segmentation on CT angiography

A medical AI team building a segmentation model for the thoracic aorta needed a high-quality training set annotated on CT angiography studies. The workflow required consistent labels across several sequential steps, from lumen localization to true / false lumen differentiation, with rigorous adherence to a multi-stage protocol.

Produce clean, protocol-compliant segmentations of the aortic lumen, aortic centerline, multiplanar reformations and true / false lumen classes, exported in a format directly consumable by the client's training pipeline.

A dedicated pool of radiologists was trained on the client's staged protocol on a widely used browser-based annotation platform. Each case followed the same ordered pipeline, with color-coded labels and cross-slice continuity checks. A project manager tracked throughput and ran systematic quality reviews before weekly deliveries.

The client received a consistent, ready-to-train dataset with predictable weekly volumes and full traceability. Learning-curve gains reduced the average per-case time while preserving the strict protocol required for vascular imaging.