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: Multi-Depth Boundary-Aware Left Atrial Scar Segmentation Network

Abstract: Automatic segmentation of left atrial (LA) scars from late gadolinium enhanced CMR images is a crucial step for atrial fibrillation (AF) recurrence analysis. However, delineating LA scars is tedious and error-prone due to the variation of scar shapes. In this work, we propose a boundary-aware LA scar segmentation network, which is composed of two branches to segment LA and LA scars, respectively. We explore the inherent spatial relationship between LA and LA scars. By introducing a Sobel fusion module between the two segmentation branches, the spatial information of LA boundaries can be propagated from the LA branch to the scar branch. Thus, LA scar segmentation can be performed condition on the LA boundaries regions. In our experiments, 40 labeled images were used to train the proposed network, and the remaining 20 labeled images were used for evaluation. The network achieved an average Dice score of 0.608 for LA scar segmentation.
Subjects: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
Cite as: arXiv:2208.04940 [eess.IV]
  (or arXiv:2208.04940v1 [eess.IV] for this version)

Submission history

From: Wangbin Ding [view email]
[v1] Mon, 8 Aug 2022 03:32:18 GMT (659kb,D)

Link back to: arXiv, form interface, contact.