← News

ResearchSources: arXiv · 16 Nov 2026

TotalSegmentator team releases open-source tool measuring aortic diameters at 11 CT landmarks

A retrospective single-center study reports an automated, open-source pipeline that measures the aorta from annulus to bifurcation, with a mean absolute error of about 1 mm against corrected radiologist measurements.

In plain words

Researchers built a free, open-source program that reads CT scans and automatically measures the aorta's width at 11 fixed points, following guideline recommendations. In their tests it came within about 1 mm of a radiologist's corrected measurements. Accuracy was weaker near the aortic root and in scans showing mural thrombus, a blood clot along the vessel wall.

What was announced

The tool, described in a retrospective single-center study, works from TotalSegmentator segmentations of aortic branches and vertebrae to locate 11 standardized landmarks spanning the aortic annulus to the bifurcation, then measures the maximum diameter at each. The evaluation covered 239 arterial phase CT angiography examinations of 230 patients performed between April 2025 and October 2026, most of them ECG-gated. A senior radiologist reviewed and, where needed, corrected the automated measurements to create the reference standard, so the reported errors are relative to that corrected baseline rather than to fully independent readings.

Across all landmarks the mean absolute error was 1.06 mm, and 86.6% of differences were 2 mm or less, with a slight tendency to measure larger diameters (bias 0.45 mm). The sinuses of Valsalva were the hardest site, with an MAE of 2.17 mm, and automatic measurement failed for 1.2% of landmarks. The authors report that examination-level error did not differ significantly by pathology, atherosclerosis grade, or ECG gating, but examinations with mural thrombus more often exceeded a 2 mm error (50% versus 20%, P=.03). They conclude the error sits below published interreader variability. The code is freely available on GitHub.

What it does not show

  • The evidence is the arXiv abstract only; full methods, figures, and peer review status are not visible in the supplied excerpt.
  • Agreement was measured against automated readings corrected by one senior radiologist at a single center, not against fully independent multi-reader or multi-center references.
  • The study population covers arterial phase CT angiography examinations from one center between April 2025 and October 2026, so performance on other scan types or populations is not established in this evidence.

For security and IT teams

Access
The tool is open source and freely available at the TotalSegmentator GitHub repository, per the study's code availability statement.

Sources

Source-support review completed 17 Nov 2026. This story was drafted with AI assistance and checked against retained source passages. Automated checking is not independent verification.

Coverage: arXiv: abstract.

Updates and corrections

No corrections so far.

First published 17 Nov 2026

New to the subject? AI from the Ground Up starts from the beginning.