Joint Semantic Feature and Optical Flow Learning for Automatic Echocardiography Segmentation

Lyu, J; Meng, J; Zhang, Y; Ling, SH; Sun, L

Joint Semantic Feature and Optical Flow Learning for Automatic Echocardiography Segmentation

Lyu, J Meng, J Zhang, Y Ling, SH Sun, L

Permalink

Publisher:: Springer Nature
Publication Type:: Chapter
Citation:: Advanced Intelligent Computing Technology and Applications, 2024, 14868 LNCS, pp. 160-171
Issue Date:: 2024-01-01

Embargoed

	Filename	Description	Size
	Joint Semantic Feature and Optical Flow Learning for Automatic Echocardiography Segmentation.pdf	Accepted version	360.86 kB	Adobe PDF	View/Open

Copyright Clearance Process

Recently Added
In Progress
Embargoed
Open Access

This item is currently unavailable due to the publisher's embargo.

The embargo period expires on 30 Jul 2026

Full metadata record

Field	Value	Language
dc.contributor.author	Lyu, J
dc.contributor.author	Meng, J
dc.contributor.author	Zhang, Y
dc.contributor.author	Ling, SH
dc.contributor.author	Sun, L
dc.date.accessioned	2024-09-09T03:24:07Z
dc.date.available	2024-09-09T03:24:07Z
dc.date.issued	2024-01-01
dc.identifier.citation	Advanced Intelligent Computing Technology and Applications, 2024, 14868 LNCS, pp. 160-171
dc.identifier.isbn	9789819755998
dc.identifier.uri	http://hdl.handle.net/10453/180751
dc.description.abstract	The left ventricle ejection fraction is an important index for assessing cardiac function and diagnosing cardiac diseases. At present, EchoNet-Dynamic dataset is the unique large-scale resource for studying ejection fraction estimation by echocardiography. Through segmentation of the end-systolic and end-diastolic frames, the ejection fraction can be calculated based on the volumes at these phases. However, existing segmentation methods either mostly focus on single-frame segmentation and rarely consider information across consecutive frames, or they fail to effectively exploit temporal information between consecutive frames, resulting in suboptimal segmentation performance. In our study, we constructed a dual-branch spatial-temporal feature extraction model for achieving echocardiogram video segmentation. One branch was dedicated to extracting semantic features of frames under supervision, while the other branch learned the optical flows between frames in an unsupervised manner. Subsequently, we jointly trained these two branches using a temporal consistency mechanism to acquire spatial-temporal features of the frames. This approach enhances both video segmentation performance and the consistency of transition frame segmentation. Experimental results demonstrate that our proposed model achieves promising segmentation performance compared to existing methods.
dc.language	en
dc.publisher	Springer Nature
dc.relation.ispartof	Advanced Intelligent Computing Technology and Applications
dc.relation.ispartofseries	Lecture Notes in Computer Science
dc.relation.isbasedon	10.1007/978-981-97-5600-1_14
dc.rights	info:eu-repo/semantics/embargoedAccess
dc.subject.classification	Artificial Intelligence & Image Processing
dc.subject.classification	46 Information and computing sciences
dc.title	Joint Semantic Feature and Optical Flow Learning for Automatic Echocardiography Segmentation
dc.type	Chapter
utslib.citation.volume	14868 LNCS
pubs.organisational-group	University of Technology Sydney
pubs.organisational-group	University of Technology Sydney/Faculty of Engineering and Information Technology
pubs.organisational-group	University of Technology Sydney/Strength - CHT - Health Technologies
pubs.organisational-group	University of Technology Sydney/Faculty of Engineering and Information Technology/School of Electrical and Data Engineering
pubs.organisational-group	University of Technology Sydney/All Manual Groups
pubs.organisational-group	University of Technology Sydney/All Manual Groups/Centre for Health Technologies (CHT)
utslib.copyright.status	embargoed	*
utslib.copyright.embargo	2026-07-30T00:00:00+1000Z
dc.date.updated	2024-09-09T03:24:06Z
pubs.publication-status	Published
pubs.volume	14868 LNCS

Abstract:

The left ventricle ejection fraction is an important index for assessing cardiac function and diagnosing cardiac diseases. At present, EchoNet-Dynamic dataset is the unique large-scale resource for studying ejection fraction estimation by echocardiography. Through segmentation of the end-systolic and end-diastolic frames, the ejection fraction can be calculated based on the volumes at these phases. However, existing segmentation methods either mostly focus on single-frame segmentation and rarely consider information across consecutive frames, or they fail to effectively exploit temporal information between consecutive frames, resulting in suboptimal segmentation performance. In our study, we constructed a dual-branch spatial-temporal feature extraction model for achieving echocardiogram video segmentation. One branch was dedicated to extracting semantic features of frames under supervision, while the other branch learned the optical flows between frames in an unsupervised manner. Subsequently, we jointly trained these two branches using a temporal consistency mechanism to acquire spatial-temporal features of the frames. This approach enhances both video segmentation performance and the consistency of transition frame segmentation. Experimental results demonstrate that our proposed model achieves promising segmentation performance compared to existing methods.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/180751