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

Download:

Current browse context:

eess.IV

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: Unsupervised Domain Adaptation for Small Bowel Segmentation using Disentangled Representation

Abstract: We present a novel unsupervised domain adaptation method for small bowel segmentation based on feature disentanglement. To make the domain adaptation more controllable, we disentangle intensity and non-intensity features within a unique two-stream auto-encoding architecture, and selectively adapt the non-intensity features that are believed to be more transferable across domains. The segmentation prediction is performed by aggregating the disentangled features. We evaluated our method using intravenous contrast-enhanced abdominal CT scans with and without oral contrast, which are used as source and target domains, respectively. The proposed method showed clear improvements in terms of three different metrics compared to other domain adaptation methods that are without the feature disentanglement. The method brings small bowel segmentation closer to clinical application.
Comments: Accepted to MICCAI 2021
Subjects: Image and Video Processing (eess.IV)
Cite as: arXiv:2107.02854 [eess.IV]
  (or arXiv:2107.02854v1 [eess.IV] for this version)

Submission history

From: Seung Yeon Shin [view email]
[v1] Tue, 6 Jul 2021 19:45:57 GMT (3524kb,D)

Link back to: arXiv, form interface, contact.