Abstract
Digital twins of the human spine require accurate, subjectspecific anatomical models to enable personalized biomedical simulation and clinical decision support. A critical prerequisite is the precise extraction of 3D anatomical landmarks, which define spinal alignment, vertebral orientation, and attachment points for ligaments and muscles. This work presents a two-step approach for predicting 3D landmarks from clinical CT and vertebra segmentation masks. It employs a DenseNet as the backbone for feature extraction and initial predictions, which are further refined by a transformer. This transformer receives local patch features, initial predictions, global features from the DenseNet, and the vertebra label to predict offset vectors, utilizing both global and local context and explicitly learning the relationship between landmarks. Our results show that prior knowledge from surface segmentation annotations can be successfully exploited to improve landmark accuracy, achieving average L1 errors of close to1 mm. This significantly outperforms the baselines H3DE-Net and PRUnet3D. This work contributes to the digital twin ecosystem by providing a robust, automated pipeline that enables mechanical spine simulation tailored to the subject’s individual anatomy.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/DT4H_006.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/forum?id=OdPPwzglvQ
BibTex
@InProceedings{MölHen_VERPEX_MICCAISAT2026,
author = { Möller, Hendrik AND Wang, Alissa Yuxuan AND Graf, Robert AND Nispel, Kati AND Atad, Matan AND Menze, Bjoern AND Rueckert, Daniel AND Kirschke, Jan S. AND Lerchl, Tanja},
title = { { VERPEX: Anatomical Landmark Extraction on 3D Vertebrae exploiting Segmentation Masks } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17275},
month = {pending},
page = {pending}
}