Zero-shot segmentation of spinal vertebrae with metastatic lesions: an analysis of Meta’s Segment Anything Model 2 and factors affecting learning free segmentation
These results demonstrate that general-purpose segmentation models like SAM 2 can provide reasonable vertebral segmentation accuracy with no pretraining, with efficacy comparable to previously published trained models. Future research should include optimizations of spine segmentation models for vertebral location and patient body habitus, as well as for variations in imaging quality approaches.
Authors
- Nader S. Dahdaleh (ORCID: https://orcid.org/0000-0002-3105-1404)
- Rushmin Khazanchi
- Christopher S. Ahuja (ORCID: https://orcid.org/0000-0002-9910-148X)
- Najib E. El Tecle (ORCID: https://orcid.org/0000-0002-5242-1646)
- Rebecca Du (ORCID: https://orcid.org/0000-0003-0865-874X)
- Rishi Jain (ORCID: https://orcid.org/0009-0003-0739-9966)
- Sachin Govind
Institutions
- Northwestern University (US)
- Neurological Surgery (US)
Publication Details
- Journal
- Neurosurgical FOCUS
- Published
- 2025-07-01
- DOI
- https://doi.org/10.3171/2025.4.focus25234
- Citations
- 2
- Primary Topic
- Medical Imaging and Analysis
- Type
- article
- Field-Weighted Citation Impact
- 1.35