EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Plug‐and‐play segment anything model improves nnUNet performance

Yunxiang Li, Bowen Jing, Zihan Li, Jing Wang, You Zhang

Medical Physics · 2025-02 · 52 (2) · 899-912

DOI: 10.1002/mp.17481 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Medical Physics, online erschienen am 28.10.2024. Verfasst von Yunxiang Li und 4 weiteren Autor:innen. Beteiligte Einrichtung: Department of Radiation Oncology, UT Southwestern Medical Center, Dallas, Texas, USA. Hauptthemen laut MEDLINE: Image Processing, Computer-Assisted. Die Arbeit wurde bisher 48-mal zitiert (Crossref) und nennt 30 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: NIH HHS, NCI NIH HHS, NIBIB NIH HHS.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

The automatic segmentation of medical images has widespread applications in modern clinical workflows. The Segment Anything Model (SAM), a recent development of foundational models in computer vision, has become a universal tool for image segmentation without the need for specific domain training. However, SAM's reliance on prompts necessitates human‐computer interaction during the inference process. Its performance on specific domains can also be limited without additional adaptation. In contrast, traditional models like nnUNet are designed to perform segmentation tasks automatically during inference and can work well for each specific domain, but they require extensive training on domain‐specific datasets.

Purpose

To leverage the advantages of both foundational and domain‐specific models and achieve fully automated segmentation with limited training samples, we propose nnSAM, which combines the robust feature extraction capabilities of SAM with the automatic configuration abilities of nnUNet to enhance the accuracy and robustness of medical image segmentation on small datasets.

Methods

We propose the nnSAM model for small sample medical image segmentation. We made optimizations for this goal via two main approaches: first, we integrated the feature extraction capabilities of SAM with the automatic configuration advantages of nnUNet, which enables robust feature extraction and domain‐specific adaptation on small datasets. Second, during the training process, we designed a boundary shape supervision loss based on level set functions and curvature calculations, enabling the model to learn anatomical shape priors from limited annotation data.

Results

We conducted quantitative and qualitative assessments on the performance of our proposed method on four segmentation tasks: brain white matter, liver, lung, and heart segmentation. Our method achieved the best performance across all tasks. Specifically, in brain white matter segmentation using 20 training samples, nnSAM achieved the highest DICE score of 82.77 ( 10.12) % and the lowest average surface distance (ASD) of 1.14 ( 1.03) mm, compared to nnUNet, which had a DICE score of 79.25 ( 17.24) % and an ASD of 1.36 ( 1.63) mm. A sample size study shows that the advantage of nnSAM becomes more prominent under fewer training samples.

Conclusions

A comprehensive evaluation of multiple small‐sample segmentation tasks demonstrates significant improvements in segmentation performance by nnSAM, highlighting the vast potential of small‐sample learning.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Image Processing, Computer-AssistedHumans

Stichwörter der Autor:innen

few‐shot learningfoundation modelmedical image segmentation

Autor:innen und Institutionen

  • Department of Radiation Oncology, UT Southwestern Medical Center, Dallas, Texas, USA.
  • Department of Bioengineering, University of Washington, Seattle, Washington, USA.

Förderung

  • NIH HHS
  • NCI NIH HHS
  • NIBIB NIH HHS
  • National Institutes of Health

Zitieren

APA

Li, Y., Jing, B., Li, Z., Wang, J., & Zhang, Y. (2025). Plug‐and‐play segment anything model improves nnUNet performance. Medical Physics, 52(2), 899-912. https://doi.org/10.1002/mp.17481

Vancouver

Li Y, Jing B, Li Z, Wang J, Zhang Y. Plug‐and‐play segment anything model improves nnUNet performance. Medical Physics. 2025;52(2):899-912. doi:10.1002/mp.17481

BibTeX
@article{Li2025,
  author = {Li, Yunxiang and Jing, Bowen and Li, Zihan and Wang, Jing and Zhang, You},
  title = {Plug‐and‐play segment anything model improves nnUNet performance},
  journal = {Medical Physics},
  year = {2025},
  volume = {52},
  number = {2},
  pages = {899--912},
  doi = {10.1002/mp.17481}
}
Kostenlos starten →