EUVIMEDEUVIMED entdecken
Crossref

Optimizing Artificial Intelligence - Segmentation of Sinonasal Masses in Endoscopic Images Through Augmentation of Meta's Segment Anything Model

Kenan Ye, Lirit Levi, Mahdokht Manavi, Maxime Fieux, Axel Renteria, Jayakar Nayak, Zara M. Patel, Noel F. Ayoub, Peter H. Hwang, Michael T. Chang

Journal of Neurological Surgery Part B: Skull Base · 01.03.2026 · 87 (S 01 S 01) · S1-S548

DOI: 10.1055/s-0046-1818752 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Neurological Surgery Part B: Skull Base, erschienen am 01.03.2026. Verfasst von Kenan Ye und 9 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Objective

Accurate segmentation of sinonasal lesions is essential for developing AI-based applications in endoscopic procedures. Current segmentation models are developed primarily using convolutional neural network (CNN) architecture, and plateau at an accuracy of ~0.75. Our objective is to develop an augmented Meta’s Segment Anything Model (SAM) model optimized for segmentation of sinonasal masses in endoscopy images.

Methods

We integrated SAM with a ResNet-50 CNN using cross-attention layers and trained the model on 1,242 endoscopy images from patients evaluated at an otolaryngology center ([ Fig. 1 ]). Images included tumors, polyps, or normal findings. Images were annotated and independently validated by 3 otolaryngologists. We compared performance of augmented SAM to nnUNet and zero-shot SAM using Dice similarity coefficient and Intersection-over-Union (IoU) metrics, applying an 8:1:1 train-validation-test split. For external validation, we tested augmented SAM on a publicly available dataset from another institution.

Results

Augmented SAM achieved a Dice of 0.90 and an IoU of 0.82 for polyp segmentation ([ Fig. 2 ]), outperforming nnUNet (Dice 0.74, IoU 0.60) and zero-shot SAM (DICE 0.58, IoU 0.43). For tumors ([ Fig. 3 ]), augmented SAM reached Dice 0.87 and IoU of 0.77, compared with nnUNet (Dice 0.73, IoU 0.58) and zero-shot SAM (Dice 0.61, IoU 0.45). On the external dataset, augmented SAM achieved a high accuracy for tumors (Dice 0.86) and polyps (Dice 0.90).

Conclusion

Augmented SAM significantly improves segmentation accuracy for sinonasal masses, outperforming existing benchmarks. This lays the potential groundwork for development of real-time, lesion-aware AI tools in rhinologic procedures Publication History Article published online: 27 February 2026 © 2026. Thieme. All rights reserved. Georg Thieme Verlag KG Oswald-Hesse-Straße 50, 70469 Stuttgart, Germany

Zitieren

APA

Ye, K., Levi, L., Manavi, M., Fieux, M., Renteria, A., Nayak, J., Patel, Z. M., Ayoub, N. F., Hwang, P. H., & Chang, M. T. (2026). Optimizing Artificial Intelligence - Segmentation of Sinonasal Masses in Endoscopic Images Through Augmentation of Meta's Segment Anything Model. Journal of Neurological Surgery Part B: Skull Base, 87(S 01 S 01), S1-S548. https://doi.org/10.1055/s-0046-1818752

Vancouver

Ye K, Levi L, Manavi M, Fieux M, Renteria A, Nayak J, et al. Optimizing Artificial Intelligence - Segmentation of Sinonasal Masses in Endoscopic Images Through Augmentation of Meta's Segment Anything Model. Journal of Neurological Surgery Part B: Skull Base. 2026;87(S 01 S 01):S1-S548. doi:10.1055/s-0046-1818752

BibTeX
@article{Ye2026,
  author = {Ye, Kenan and Levi, Lirit and Manavi, Mahdokht and Fieux, Maxime and Renteria, Axel and Nayak, Jayakar and Patel, Zara M. and Ayoub, Noel F. and Hwang, Peter H. and Chang, Michael T.},
  title = {Optimizing Artificial Intelligence - Segmentation of Sinonasal Masses in Endoscopic Images Through Augmentation of Meta's Segment Anything Model},
  journal = {Journal of Neurological Surgery Part B: Skull Base},
  year = {2026},
  volume = {87},
  number = {S 01 S 01},
  pages = {S1--S548},
  doi = {10.1055/s-0046-1818752}
}
Kostenlos starten →