Aortic dissection lumen segmentation on CTA
Slice-by-slice segmentation of true lumen, false lumen and partial thrombosis / hematoma on aortic dissection CTA studies.
An AI cardiology team building an early-detection tool for aortic dissection needed radiologists to segment every slice of the aorta between the arch and the aorto-iliac bifurcation, distinguishing true lumen, false lumen and partial thrombosis or hematoma, with careful handling of branch vessel origins.
Label true lumen, false lumen and thrombus / hematoma on every axial slice from ~3 cm distal to the aortic arch down to ~2 cm below the aorto-iliac bifurcation, including ostia of coronary, celiac, SMA and renal arteries when visualized.
Radiologists worked in axial view on the client's RedBrick workflow (label → review → ground truth) using a suggested window width / level for aortic contrast, with sagittal and coronal views used only for verification. A dedicated review step checked both anatomical accuracy and mask quality before promotion to ground truth.
The client received a densely labeled, per-slice dataset covering the full anatomy of interest with consistent handling of branch vessels, feeding directly into their dissection detection and classification model.