Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Deep Label Fusion : A 3D End-To-End Hybrid Multi-atlas Segmentation and Deep Learning Pipeline

Xie, Long ; Wisse, Laura E.M. LU orcid ; Wang, Jiancong ; Ravikumar, Sadhana ; Glenn, Trevor ; Luther, Anica ; Lim, Sydney ; Wolk, David A. and Yushkevich, Paul A. (2021) 27th International Conference on Information Processing in Medical Imaging, IPMI 2021 In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 12729 LNCS. p.428-439
Abstract

Deep learning (DL) is the state-of-the-art methodology in various medical image segmentation tasks. However, it requires relatively large amounts of manually labeled training data, which may be infeasible to generate in some applications. In addition, DL methods have relatively poor generalizability to out-of-sample data. Multi-atlas segmentation (MAS), on the other hand, has promising performance using limited amounts of training data and good generalizability. A hybrid method that integrates the high accuracy of DL and good generalizability of MAS is highly desired and could play an important role in segmentation problems where manually labeled data is hard to generate. Most of the prior work focuses on improving single components of... (More)

Deep learning (DL) is the state-of-the-art methodology in various medical image segmentation tasks. However, it requires relatively large amounts of manually labeled training data, which may be infeasible to generate in some applications. In addition, DL methods have relatively poor generalizability to out-of-sample data. Multi-atlas segmentation (MAS), on the other hand, has promising performance using limited amounts of training data and good generalizability. A hybrid method that integrates the high accuracy of DL and good generalizability of MAS is highly desired and could play an important role in segmentation problems where manually labeled data is hard to generate. Most of the prior work focuses on improving single components of MAS using DL rather than directly optimizing the final segmentation accuracy via an end-to-end pipeline. Only one study explored this idea in binary segmentation of 2D images, but it remains unknown whether it generalizes well to multi-class 3D segmentation problems. In this study, we propose a 3D end-to-end hybrid pipeline, named deep label fusion (DLF), that takes advantage of the strengths of MAS and DL. Experimental results demonstrate that DLF yields significant improvements over conventional label fusion methods and U-Net, a direct DL approach, in the context of segmenting medial temporal lobe subregions using 3T T1-weighted and T2-weighted MRI. Further, when applied to an unseen similar dataset acquired in 7T, DLF maintains its superior performance, which demonstrates its good generalizability.

(Less)
Please use this url to cite or link to this publication:
author
; ; ; ; ; ; ; and
organization
publishing date
type
Chapter in Book/Report/Conference proceeding
publication status
published
subject
host publication
Information Processing in Medical Imaging - 27th International Conference, IPMI 2021, Proceedings
series title
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
editor
Feragen, Aasa ; Sommer, Stefan ; Schnabel, Julia and Nielsen, Mads
volume
12729 LNCS
pages
12 pages
publisher
Springer Science and Business Media B.V.
conference name
27th International Conference on Information Processing in Medical Imaging, IPMI 2021
conference location
Virtual, Online
conference dates
2021-06-28 - 2021-06-30
external identifiers
  • scopus:85111457466
ISSN
0302-9743
1611-3349
ISBN
9783030781903
DOI
10.1007/978-3-030-78191-0_33
language
English
LU publication?
yes
id
cfc8a0fe-4a4c-4891-990f-6f603216b8c4
date added to LUP
2021-08-31 11:32:29
date last changed
2025-03-09 17:01:16
@inproceedings{cfc8a0fe-4a4c-4891-990f-6f603216b8c4,
  abstract     = {{<p>Deep learning (DL) is the state-of-the-art methodology in various medical image segmentation tasks. However, it requires relatively large amounts of manually labeled training data, which may be infeasible to generate in some applications. In addition, DL methods have relatively poor generalizability to out-of-sample data. Multi-atlas segmentation (MAS), on the other hand, has promising performance using limited amounts of training data and good generalizability. A hybrid method that integrates the high accuracy of DL and good generalizability of MAS is highly desired and could play an important role in segmentation problems where manually labeled data is hard to generate. Most of the prior work focuses on improving single components of MAS using DL rather than directly optimizing the final segmentation accuracy via an end-to-end pipeline. Only one study explored this idea in binary segmentation of 2D images, but it remains unknown whether it generalizes well to multi-class 3D segmentation problems. In this study, we propose a 3D end-to-end hybrid pipeline, named deep label fusion (DLF), that takes advantage of the strengths of MAS and DL. Experimental results demonstrate that DLF yields significant improvements over conventional label fusion methods and U-Net, a direct DL approach, in the context of segmenting medial temporal lobe subregions using 3T T1-weighted and T2-weighted MRI. Further, when applied to an unseen similar dataset acquired in 7T, DLF maintains its superior performance, which demonstrates its good generalizability.</p>}},
  author       = {{Xie, Long and Wisse, Laura E.M. and Wang, Jiancong and Ravikumar, Sadhana and Glenn, Trevor and Luther, Anica and Lim, Sydney and Wolk, David A. and Yushkevich, Paul A.}},
  booktitle    = {{Information Processing in Medical Imaging - 27th International Conference, IPMI 2021, Proceedings}},
  editor       = {{Feragen, Aasa and Sommer, Stefan and Schnabel, Julia and Nielsen, Mads}},
  isbn         = {{9783030781903}},
  issn         = {{0302-9743}},
  language     = {{eng}},
  pages        = {{428--439}},
  publisher    = {{Springer Science and Business Media B.V.}},
  series       = {{Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)}},
  title        = {{Deep Label Fusion : A 3D End-To-End Hybrid Multi-atlas Segmentation and Deep Learning Pipeline}},
  url          = {{http://dx.doi.org/10.1007/978-3-030-78191-0_33}},
  doi          = {{10.1007/978-3-030-78191-0_33}},
  volume       = {{12729 LNCS}},
  year         = {{2021}},
}