A Scalable, Annotation-Free Pipeline for Automated CT Pelvimetry: A Validation Study with Landmark Uncertainty Analysis and Clinical Correlation.

Authors

Shih-Feng Huang, Hsin-Ping Tseng, Yu-Hsun Chen, Yung-Lin Tan, Jen-Ping Lee, Chih-Chien Wu, Chao-Wen Hsu

Published in

Journal of imaging informatics in medicine. Jul 29, 2026. Epub Jul 29, 2026.

Abstract

This study presents the validation of a fully automated, annotation-free CT pelvimetry pipeline that is built on a publicly available, interchangeable segmentation backend (here, TotalSegmentator) without task-specific retraining, derives seven pelvimetric parameters through deterministic rule-based landmark extraction, and is released as an open-source package. Technical validation was performed in 60 patients by comparison with repeated manual measurements from two independent measurement sessions using intraclass correlation coefficients (ICC), Bland-Altman analysis, and three-dimensional landmark displacement quantification. Exploratory clinical evaluation was conducted in a single-surgeon cohort (n = 106). The automated pipeline successfully processed 199 of 200 consecutive CT datasets (99.5%). Automated measurements demonstrated moderate-to-excellent agreement with manual pelvimetry (ICC 0.66-0.97). Metrics defined by well-corticated landmarks showed the highest agreement (interspinous-distance ICC, 0.97), whereas measurements dependent on the coccygeal apex exhibited greater variability (mean 3D displacement 5.4 mm intra-rater, 8.9 mm automated-versus-manual). Coordinate-level analysis traced measurement variability predominantly to specific anatomically ambiguous landmarks, providing a diagnostic capability absent from conventional aggregate validation approaches. In exploratory clinical evaluation, wider transverse pelvic dimensions were associated with lower intraoperative blood loss. The open-source implementation enables scalable, annotation-free pelvimetric assessment with coordinate-level uncertainty characterization potentially applicable to other domains where geometric parameters are derived from segmentation masks.

PMID:
42527790
Bibliographic data and abstract were imported from PubMed on 30 Jul 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.