Y-Net: A Spatiospectral Dual-Encoder Network for Medical Image Segmentation

Azade Farshad; Yousef Yeganeh; Peter Gehlbach; Nassir Navab

doi:10.1007/978-3-031-16434-7_56

Y-Net: A Spatiospectral Dual-Encoder Network for Medical Image Segmentation

Azade Farshad^*, Yousef Yeganeh, Peter Gehlbach, Nassir Navab

^*Corresponding author for this work

Not published at Aalto University

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

63 Citations (Scopus)

Abstract

Automated segmentation of retinal optical coherence tomography (OCT) images has become an important recent direction in machine learning for medical applications. We hypothesize that the anatomic structure of layers and their high-frequency variation in OCT images make retinal OCT a fitting choice for extracting spectral domain features and combining them with spatial domain features. In this work, we present Y-Net, an architecture that combines the frequency domain features with the image domain to improve the segmentation performance of OCT images. The results of this work demonstrate that the introduction of two branches, one for spectral and one for spatial domain features, brings very significant improvement in fluid segmentation performance and allows outperformance as compared to the well-known U-Net model. Our improvement was 13 % on the fluid segmentation dice score and 1.9 % on the average dice score. Finally, removing selected frequency ranges in the spectral domain demonstrates the impact of these features on the fluid segmentation outperformance. Code: github.com/azadef/ynet

Original language	English
Title of host publication	Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 - 25th International Conference, Proceedings
Editors	Linwei Wang, Qi Dou, P. Thomas Fletcher, Stefanie Speidel, Shuo Li
Publisher	Springer
Pages	582-592
Number of pages	11
ISBN (Print)	9783031164330
DOIs	https://doi.org/10.1007/978-3-031-16434-7_56
Publication status	Published - 2022
MoE publication type	A4 Conference publication
Event	International Conference on Medical Image Computing and Computer-Assisted Intervention - Singapore, Singapore Duration: 18 Sept 2022 → 22 Sept 2022 Conference number: 25

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13432 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	International Conference on Medical Image Computing and Computer-Assisted Intervention
Abbreviated title	MICCAI
Country/Territory	Singapore
City	Singapore
Period	18/09/2022 → 22/09/2022

Keywords

Frequency domain in OCT
OCT segmentation
U-Net

Access to Document

10.1007/978-3-031-16434-7_56

Cite this

Farshad, A., Yeganeh, Y., Gehlbach, P., & Navab, N. (2022). Y-Net: A Spatiospectral Dual-Encoder Network for Medical Image Segmentation. In L. Wang, Q. Dou, P. T. Fletcher, S. Speidel, & S. Li (Eds.), Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 - 25th International Conference, Proceedings (pp. 582-592). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13432 LNCS). Springer. https://doi.org/10.1007/978-3-031-16434-7_56

Farshad, Azade ; Yeganeh, Yousef ; Gehlbach, Peter et al. / Y-Net: A Spatiospectral Dual-Encoder Network for Medical Image Segmentation. Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 - 25th International Conference, Proceedings. editor / Linwei Wang ; Qi Dou ; P. Thomas Fletcher ; Stefanie Speidel ; Shuo Li. Springer, 2022. pp. 582-592 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{1b18b08c84db47edb7679f485a8a095e,

title = "Y-Net: A Spatiospectral Dual-Encoder Network for Medical Image Segmentation",

abstract = "Automated segmentation of retinal optical coherence tomography (OCT) images has become an important recent direction in machine learning for medical applications. We hypothesize that the anatomic structure of layers and their high-frequency variation in OCT images make retinal OCT a fitting choice for extracting spectral domain features and combining them with spatial domain features. In this work, we present Y-Net, an architecture that combines the frequency domain features with the image domain to improve the segmentation performance of OCT images. The results of this work demonstrate that the introduction of two branches, one for spectral and one for spatial domain features, brings very significant improvement in fluid segmentation performance and allows outperformance as compared to the well-known U-Net model. Our improvement was 13 \% on the fluid segmentation dice score and 1.9 \% on the average dice score. Finally, removing selected frequency ranges in the spectral domain demonstrates the impact of these features on the fluid segmentation outperformance. Code: github.com/azadef/ynet",

keywords = "Frequency domain in OCT, OCT segmentation, U-Net",

author = "Azade Farshad and Yousef Yeganeh and Peter Gehlbach and Nassir Navab",

note = "Publisher Copyright: {\textcopyright} 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.; International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI ; Conference date: 18-09-2022 Through 22-09-2022",

year = "2022",

doi = "10.1007/978-3-031-16434-7\_56",

language = "English",

isbn = "9783031164330",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer",

pages = "582--592",

editor = "Linwei Wang and Qi Dou and Fletcher, \{P. Thomas\} and Stefanie Speidel and Shuo Li",

booktitle = "Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 - 25th International Conference, Proceedings",

address = "Germany",

}

Farshad, A, Yeganeh, Y, Gehlbach, P & Navab, N 2022, Y-Net: A Spatiospectral Dual-Encoder Network for Medical Image Segmentation. in L Wang, Q Dou, PT Fletcher, S Speidel & S Li (eds), Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 - 25th International Conference, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13432 LNCS, Springer, pp. 582-592, International Conference on Medical Image Computing and Computer-Assisted Intervention, Singapore, Singapore, 18/09/2022. https://doi.org/10.1007/978-3-031-16434-7_56

Y-Net: A Spatiospectral Dual-Encoder Network for Medical Image Segmentation. / Farshad, Azade; Yeganeh, Yousef; Gehlbach, Peter et al.
Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 - 25th International Conference, Proceedings. ed. / Linwei Wang; Qi Dou; P. Thomas Fletcher; Stefanie Speidel; Shuo Li. Springer, 2022. p. 582-592 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13432 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

TY - GEN

T1 - Y-Net: A Spatiospectral Dual-Encoder Network for Medical Image Segmentation

AU - Farshad, Azade

AU - Yeganeh, Yousef

AU - Gehlbach, Peter

AU - Navab, Nassir

N1 - Conference code: 25

PY - 2022

Y1 - 2022

N2 - Automated segmentation of retinal optical coherence tomography (OCT) images has become an important recent direction in machine learning for medical applications. We hypothesize that the anatomic structure of layers and their high-frequency variation in OCT images make retinal OCT a fitting choice for extracting spectral domain features and combining them with spatial domain features. In this work, we present Y-Net, an architecture that combines the frequency domain features with the image domain to improve the segmentation performance of OCT images. The results of this work demonstrate that the introduction of two branches, one for spectral and one for spatial domain features, brings very significant improvement in fluid segmentation performance and allows outperformance as compared to the well-known U-Net model. Our improvement was 13 % on the fluid segmentation dice score and 1.9 % on the average dice score. Finally, removing selected frequency ranges in the spectral domain demonstrates the impact of these features on the fluid segmentation outperformance. Code: github.com/azadef/ynet

AB - Automated segmentation of retinal optical coherence tomography (OCT) images has become an important recent direction in machine learning for medical applications. We hypothesize that the anatomic structure of layers and their high-frequency variation in OCT images make retinal OCT a fitting choice for extracting spectral domain features and combining them with spatial domain features. In this work, we present Y-Net, an architecture that combines the frequency domain features with the image domain to improve the segmentation performance of OCT images. The results of this work demonstrate that the introduction of two branches, one for spectral and one for spatial domain features, brings very significant improvement in fluid segmentation performance and allows outperformance as compared to the well-known U-Net model. Our improvement was 13 % on the fluid segmentation dice score and 1.9 % on the average dice score. Finally, removing selected frequency ranges in the spectral domain demonstrates the impact of these features on the fluid segmentation outperformance. Code: github.com/azadef/ynet

KW - Frequency domain in OCT

KW - OCT segmentation

KW - U-Net

U2 - 10.1007/978-3-031-16434-7_56

DO - 10.1007/978-3-031-16434-7_56

M3 - Conference article in proceedings

AN - SCOPUS:85139043384

SN - 9783031164330

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 582

EP - 592

BT - Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 - 25th International Conference, Proceedings

A2 - Wang, Linwei

A2 - Dou, Qi

A2 - Fletcher, P. Thomas

A2 - Speidel, Stefanie

A2 - Li, Shuo

PB - Springer

T2 - International Conference on Medical Image Computing and Computer-Assisted Intervention

Y2 - 18 September 2022 through 22 September 2022

ER -

Farshad A, Yeganeh Y, Gehlbach P, Navab N. Y-Net: A Spatiospectral Dual-Encoder Network for Medical Image Segmentation. In Wang L, Dou Q, Fletcher PT, Speidel S, Li S, editors, Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 - 25th International Conference, Proceedings. Springer. 2022. p. 582-592. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-16434-7_56