EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Comparing 3D, 2.5D, and 2D Approaches to Brain Image Auto-Segmentation

Arman Avesta, Sajid Hossain, MingDe Lin iD, Mariam Aboian iD, Harlan M Krumholz, Sanjay Aneja iD

Bioengineering · 01.02.2023 · 10 (2) · 181

DOI: 10.3390/bioengineering10020181 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Bioengineering, online erschienen am 01.02.2023. Verfasst von Arman Avesta und 5 weiteren Autor:innen. Beteiligte Einrichtung: Department of Radiology and Biomedical Imaging, Yale School of Medicine, New Haven, CT 06510, USA. Die Arbeit wurde bisher 116-mal zitiert (Crossref) und nennt 40 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: NCATS NIH HHS, Radiological Society of North America, National Center for Advancing Translational Science.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Deep-learning methods for auto-segmenting brain images either segment one slice of the image (2D), five consecutive slices of the image (2.5D), or an entire volume of the image (3D). Whether one approach is superior for auto-segmenting brain images is not known. We compared these three approaches (3D, 2.5D, and 2D) across three auto-segmentation models (capsule networks, UNets, and nnUNets) to segment brain structures. We used 3430 brain MRIs, acquired in a multi-institutional study, to train and test our models. We used the following performance metrics: segmentation accuracy, performance with limited training data, required computational memory, and computational speed during training and deployment. The 3D, 2.5D, and 2D approaches respectively gave the highest to lowest Dice scores across all models. 3D models maintained higher Dice scores when the training set size was decreased from 3199 MRIs down to 60 MRIs. 3D models converged 20% to 40% faster during training and were 30% to 50% faster during deployment. However, 3D models require 20 times more computational memory compared to 2.5D or 2D models. This study showed that 3D models are more accurate, maintain better performance with limited training data, and are faster to train and deploy. However, 3D models require more computational memory compared to 2.5D or 2D models.

Schlagwörter

Stichwörter der Autor:innen

auto-segmentationdeep learningmagnetic resonance imagingneuroimaging

Autor:innen und Institutionen

  • Department of Radiology and Biomedical Imaging, Yale School of Medicine, New Haven, CT 06510, USA.
  • Department of Therapeutic Radiology, Yale School of Medicine, New Haven, CT 06510, USA.
  • Center for Outcomes Research and Evaluation, Yale School of Medicine, New Haven, CT 06510, USA.
  • Visage Imaging, Inc., San Diego, CA 92130, USA.
  • Division of Cardiovascular Medicine, Yale School of Medicine, New Haven, CT 06510, USA.
  • Department of Biomedical Engineering, Yale University, New Haven, CT 06510, USA.

ORCID-Profile: MingDe Lin, Mariam Aboian, Sanjay Aneja

Förderung

  • NCATS NIH HHS
  • Radiological Society of North America
  • National Center for Advancing Translational Science
  • Radiological Society of North America’s (RSNA)

Zitieren

APA

Avesta, A., Hossain, S., Lin, M., Aboian, M., Krumholz, H. M., & Aneja, S. (2023). Comparing 3D, 2.5D, and 2D Approaches to Brain Image Auto-Segmentation. Bioengineering, 10(2), 181. https://doi.org/10.3390/bioengineering10020181

Vancouver

Avesta A, Hossain S, Lin M, Aboian M, Krumholz HM, Aneja S. Comparing 3D, 2.5D, and 2D Approaches to Brain Image Auto-Segmentation. Bioengineering. 2023;10(2):181. doi:10.3390/bioengineering10020181

BibTeX
@article{Avesta2023,
  author = {Avesta, Arman and Hossain, Sajid and Lin, MingDe and Aboian, Mariam and Krumholz, Harlan M and Aneja, Sanjay},
  title = {Comparing 3D, 2.5D, and 2D Approaches to Brain Image Auto-Segmentation},
  journal = {Bioengineering},
  year = {2023},
  volume = {10},
  number = {2},
  pages = {181},
  doi = {10.3390/bioengineering10020181}
}
Kostenlos starten →