We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

eess

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Electrical Engineering and Systems Science > Image and Video Processing

Title: Comparison of 2D vs. 3D U-Net Organ Segmentation in abdominal 3D CT images

Abstract: A two-step concept for 3D segmentation on 5 abdominal organs inside volumetric CT images is presented. First each relevant organ's volume of interest is extracted as bounding box. The extracted volume acts as input for a second stage, wherein two compared U-Nets with different architectural dimensions re-construct an organ segmentation as label mask. In this work, we focus on comparing 2D U-Nets vs. 3D U-Net counterparts. Our initial results indicate Dice improvements of about 6\% at maximum. In this study to our surprise, liver and kidneys for instance were tackled significantly better using the faster and GPU-memory saving 2D U-Nets. For other abdominal key organs, there were no significant differences, but we observe highly significant advantages for the 2D U-Net in terms of GPU computational efforts for all organs under study.
Comments: 9 pages, 6 figure, 2 tables
Subjects: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
Journal reference: International Conference on Computer Graphics, Visualization and Computer Vision 2021 - WSCG
Report number: H61
Cite as: arXiv:2107.04062 [eess.IV]
  (or arXiv:2107.04062v1 [eess.IV] for this version)

Submission history

From: Andre Mastmeyer [view email]
[v1] Thu, 8 Jul 2021 18:35:15 GMT (2256kb,D)

Link back to: arXiv, form interface, contact.