EUVIMEDEUVIMED entdecken
VergleichsstudieValidierungsstudieOpen Access

Comparative Evaluation of Deep Learning Models for 3D Segmentation and Volumetry of Vestibular Schwannomas Using Large Heterogeneous Data Sets with External Validation

Parv M Mehta, Sahika Betul Yayli, Pranjal Rai, Milan Sonka, Daniel J Blezek, Honghai Zhang iD, Victoria M Silvera, John C Benson iD, Matthew L Carlson iD, Bradley J Erickson iD, Girish Bathla iD

American Journal of Neuroradiology · 2026-05 · 47 (5) · 1266-1272

DOI: 10.3174/ajnr.a9112 ↗

Kurz gefasst

Vergleichsstudie in der Zeitschrift American Journal of Neuroradiology, online erschienen am 19.11.2025. Verfasst von Parv M Mehta und 10 weiteren Autor:innen. Beteiligte Einrichtung: From the Department of Radiology (P.M.M., S.B.Y., D.J.B., V.M.S., J.C.B., B.J.E., G.B.), Mayo Clinic, Rochester, Minnesota. Hauptthemen laut MEDLINE: Deep Learning, Neuroma, Acoustic, Imaging, Three-Dimensional, Magnetic Resonance Imaging. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 24 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

BACKGROUND AND PURPOSE: 3D segmentation and volumetry of vestibular schwannomas (VSs) is a more accurate method to determine tumor growth on serial imaging, but manual annotation is time-consuming to implement in routine clinical practice. We evaluated and compared 5 deep learning-based segmentation models (nnUNet [Base, ResEncL], U-Mamba, UNETR, and MedSAM) for 3D VS segmentation and volumetry, and we examined the robustness to acquisition heterogeneity and generalization on an external cohort. MATERIALS AND METHODS: Our refined internal data set consisted of T1-contrast-enhanced images, including 2692 scans (n =383 patients) for training and 277 scans (n = 97 patients) for testing. Post-model training and validation, performance was evaluated on both internal and a publicly available external test set (n = 241) using the Dice similarity coefficient, maximum distance between the predicted and ground truth boundaries (Hausdorff distance), surface-to-surface (S2S) distance, and relative volume error (RVE). A subanalysis of the model performance was also performed to evaluate the impact of tumor volumes and data set heterogeneity. RESULTS: The median Dice score on the external test set varied between 0.899 and 0.927 with U-Mamba achieving the highest performance, followed by nnUNet (Base and ResEncL). For these top 3 models, the median Hausdorff distance was 3.59 mm, while the 95th percentile Hausdorff distance was 1.6 mm. The S2S distance was 400 mm3 (median Dice score, 0.925-0.932). CONCLUSIONS: Models based on convolutional neural networks, transformer networks, and foundational models show robust performance for VS segmentation. Given the consistently high performance and self-optimizing frameworks of convolutional neural network-based models (U-Mamba, nnUNet), these may be more suitable for clinical applications.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Deep LearningNeuroma, AcousticImaging, Three-DimensionalMagnetic Resonance ImagingImage Interpretation, Computer-AssistedHumansFemaleMaleMiddle AgedAdultReproducibility of ResultsAgedTumor Burden

Autor:innen und Institutionen

  • From the Department of Radiology (P.M.M., S.B.Y., D.J.B., V.M.S., J.C.B., B.J.E., G.B.), Mayo Clinic, Rochester, Minnesota.
  • Department of Radiology (P.M.M.), The University of Texas at San Antonio, San Antonio, Texas.
  • College of Engineering (P.R., M.S., H.Z.), University of Iowa, Iowa City, Iowa rai.pranjal@mayo.edu.
  • College of Engineering (P.R., M.S., H.Z.), University of Iowa, Iowa City, Iowa.
  • Department of Otology/Neurotology (M.L.C.), Mayo Clinic, Rochester, Minnesota.

ORCID-Profile: Honghai Zhang, John C Benson, Matthew L Carlson, Bradley J Erickson, Girish Bathla

Zitieren

APA

Mehta, P. M., Yayli, S. B., Rai, P., Sonka, M., Blezek, D. J., Zhang, H., Silvera, V. M., Benson, J. C., Carlson, M. L., Erickson, B. J., & Bathla, G. (2026). Comparative Evaluation of Deep Learning Models for 3D Segmentation and Volumetry of Vestibular Schwannomas Using Large Heterogeneous Data Sets with External Validation. American Journal of Neuroradiology, 47(5), 1266-1272. https://doi.org/10.3174/ajnr.a9112

Vancouver

Mehta PM, Yayli SB, Rai P, Sonka M, Blezek DJ, Zhang H, et al. Comparative Evaluation of Deep Learning Models for 3D Segmentation and Volumetry of Vestibular Schwannomas Using Large Heterogeneous Data Sets with External Validation. American Journal of Neuroradiology. 2026;47(5):1266-1272. doi:10.3174/ajnr.a9112

BibTeX
@article{Mehta2026,
  author = {Mehta, Parv M and Yayli, Sahika Betul and Rai, Pranjal and Sonka, Milan and Blezek, Daniel J and Zhang, Honghai and Silvera, Victoria M and Benson, John C and Carlson, Matthew L and Erickson, Bradley J and Bathla, Girish},
  title = {Comparative Evaluation of Deep Learning Models for 3D Segmentation and Volumetry of Vestibular Schwannomas Using Large Heterogeneous Data Sets with External Validation},
  journal = {American Journal of Neuroradiology},
  year = {2026},
  volume = {47},
  number = {5},
  pages = {1266--1272},
  doi = {10.3174/ajnr.a9112}
}
Kostenlos starten →