Zero-Shot Vertebral Instance Segmentation on DICOM Spine Radiographs Using Promptable Segment Anything Models
Journal of Clinical Medicine · 07.03.2026 · 15 (5) · 2042
DOI: 10.3390/jcm15052042 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Clinical Medicine, online erschienen am 07.03.2026. Verfasst von Alexander Sieradzki und 4 weiteren Autor:innen. Beteiligte Einrichtung: Department of Artificial Intelligence, Institute of Information Technology, Warsaw University of Life Sciences, ul. Nowoursynowska 159, 02-776 Warsaw…. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 86 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Background
Accurate vertebral instance segmentation on full-spine radiographs is essential for spinal parameter assessment, but supervised methods require costly instance-level annotations and may be sensitive to domain shift.
Methods
We investigated whether promptable segmentation foundation models can generalize zero-shot to raw DICOM spine radiographs without task-specific training. We evaluated SAM-ViT-Huge, SAM2-Hiera-Large, and MedSAM-ViT-Base on 144 full-spine radiographs with 1309 annotated vertebral masks using a standardized pipeline for DICOM decoding, intensity normalization, automatic prompt generation, and instance-level evaluation. For each prompt, models produced three candidate masks. Performance was reported under an oracle protocol selecting the candidate with the highest IoU against ground truth and a model-score protocol selecting the candidate with the highest predicted IoU. Metrics included IoU, Dice, precision, recall, ASSD, and HD95.
Results
The best configuration was SAM-ViT-Huge with rectangle prompting, reaching a mean IoU/Dice of 0.782/0.870 under oracle selection and 0.737/0.837 under model-score selection. SAM2-Hiera-Large with rectangle prompting followed (0.744/0.848 oracle; 0.699/0.815 model-score), ahead of MedSAM-ViT-Base (0.599/0.737 oracle; 0.387/0.499 model-score). Point prompting yielded consistently low overlap (IoU 0.224–0.319; Dice 0.276–0.414) despite high recall, indicating systematic over-segmentation and large boundary errors.
Conclusions
Zero-shot vertebral instance segmentation on raw DICOM spine radiographs is feasible with promptable foundation models when prompts sufficiently constrain target extent. Rectangle prompting is clearly more effective than point prompting in this setting.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Artificial Intelligence, Institute of Information Technology, Warsaw University of Life Sciences, ul. Nowoursynowska 159, 02-776 Warsaw, Poland.
- Department of Musculoskeletal Disorders, National Institute of Geriatrics, Rheumatology and Rehabilitation, ul Spartańska 1, 02-637 Warsaw, Poland.
- Faculty of Medicine, Medical University of Lodz, Kościuszki 4, 90-419 Łódź, Poland.
ORCID-Profile: Alexander Sieradzki, Kamil Koszela, Szymon Koszykowski, Jakub Bednarek, Jarosław Kurek
Zitieren
APA
Sieradzki, A., Koszela, K., Koszykowski, S., Bednarek, J., & Kurek, J. (2026). Zero-Shot Vertebral Instance Segmentation on DICOM Spine Radiographs Using Promptable Segment Anything Models. Journal of Clinical Medicine, 15(5), 2042. https://doi.org/10.3390/jcm15052042
Vancouver
Sieradzki A, Koszela K, Koszykowski S, Bednarek J, Kurek J. Zero-Shot Vertebral Instance Segmentation on DICOM Spine Radiographs Using Promptable Segment Anything Models. Journal of Clinical Medicine. 2026;15(5):2042. doi:10.3390/jcm15052042
BibTeX
@article{Sieradzki2026,
author = {Sieradzki, Alexander and Koszela, Kamil and Koszykowski, Szymon and Bednarek, Jakub and Kurek, Jarosław},
title = {Zero-Shot Vertebral Instance Segmentation on DICOM Spine Radiographs Using Promptable Segment Anything Models},
journal = {Journal of Clinical Medicine},
year = {2026},
volume = {15},
number = {5},
pages = {2042},
doi = {10.3390/jcm15052042}
}
EUVIMED