Abstract
Aseptic loosening after total knee arthroplasty can be quantified non-invasively, calculating implant displacement by using load-induced CT scanning and a 3D image analysis workflow that includes segmentation of tibia bone and tibial component. The currently applied semi-automatic region-growing segmentation approach is sensitive to metal artifacts and requires substantial user interaction to compensate for the segmentation mistakes caused by them. SAM2, a promptable segmentation foundation model, was investigated as a semi-automatic deep learning-based segmentation alternative, using prompt configurations based on a single input point. While prompting enhancements with prediction-based auto-prompting and multi-plane aggregation of predictions improved segmentation performances, the under-segmentation of the tibial implant plateau or stem, and the over-segmentation of the cortical bone does not allow for a reliable displacement analysis. The results indicate that SAM2 does not yet meet the requirements for direct integration into the clinical workflow to quantify the load-induced tibial implant displacement, and that further adaptations would be needed before such adoption could be considered.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/AMAI_010.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/profile?id=%7ECaroline_Magg1
BibTex
@InProceedings{MagCar_Assessing_MICCAISAT2026,
author = { Magg, Caroline AND Dobbe, Johannes G.G. AND Streekstra, Geert J. AND Blankevoort, Leendert AND Sánchez, Clara I. AND Kervadec, Hoel},
title = { { Assessing the clinical readiness of SAM2 for CT-based tibial implant displacement quantification } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17273},
month = {pending},
page = {pending}
}