Swin UNETR for Tumor and Lymph Node Segmentation Using 3D PET/CT Imaging: A Transfer Learning Approach

Hung Chu; Luis Ricardo De la O Arévalo; Wei Tang; Baoqiang Ma; Yan Li; Alessia De Biase; Stefan Both; Johannes Albertus Langendijk; Peter van Ooijen; Nanna Maria Sijtsema; Lisanne V. van Dijk

doi:10.1007/978-3-031-27420-6_12

Swin UNETR for Tumor and Lymph Node Segmentation Using 3D PET/CT Imaging: A Transfer Learning Approach

Hung Chu^*, Luis Ricardo De la O Arévalo, Wei Tang, Baoqiang Ma, Yan Li, Alessia De Biase, Stefan Both, Johannes Albertus Langendijk, Peter van Ooijen, Nanna Maria Sijtsema, Lisanne V. van Dijk

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

Abstract

Delineation of Gross Tumor Volume (GTV) is essential for the treatment of cancer with radiotherapy. GTV contouring is a time-consuming specialized manual task performed by radiation oncologists. Deep Learning (DL) algorithms have shown potential in creating automatic segmentations, reducing delineation time and inter-observer variation. The aim of this work was to create automatic segmentations of primary tumors (GTVp) and pathological lymph nodes (GTVn) in oropharyngeal cancer patients using DL. The organizers of the HECKTOR 2022 challenge provided 3D Computed Tomography (CT) and Positron Emission Tomography (PET) scans with ground-truth GTV segmentations acquired from nine different centers. Bounding box cropping was applied to obtain an anatomic based region of interest. We used the Swin UNETR model in combination with transfer learning. The Swin UNETR encoder weights were initialized by pre-trained weights of a self-supervised Swin UNETR model. An average Dice score of 0.656 was achieved on a test set of 359 patients from the HECKTOR 2022 challenge. Code is available at: https://github.com/HC94/swin_unetr_hecktor_2022.

Original language	English
Title of host publication	Head and Neck Tumor Segmentation and Outcome Prediction - 3rd Challenge, HECKTOR 2022, Held in Conjunction with MICCAI 2022, Proceedings
Editors	Vincent Andrearczyk, Valentin Oreiller, Adrien Depeursinge, Mathieu Hatt
Publisher	Springer
Pages	114-120
Number of pages	7
ISBN (Print)	9783031274190
DOIs	https://doi.org/10.1007/978-3-031-27420-6_12
Publication status	Published - 2023
Externally published	Yes
Event	3rd 3D Head and Neck Tumor Segmentation in PET/CT Challenge, HECKTOR 2022, held in Conjunction with the 25th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2022 - Singapore, Singapore Duration: 22 Sept 2022 → 22 Sept 2022

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13626 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	3rd 3D Head and Neck Tumor Segmentation in PET/CT Challenge, HECKTOR 2022, held in Conjunction with the 25th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2022
Country/Territory	Singapore
City	Singapore
Period	22/09/22 → 22/09/22

Keywords

Auto contouring
Deep learning
Head and neck cancer
HECKTOR 2022
Image processing
Lymph node segmentation
Radiotherapy
Swin UNETR
Tumor segmentation

Access to Document

10.1007/978-3-031-27420-6_12

Cite this

Chu, H., De la O Arévalo, L. R., Tang, W., Ma, B., Li, Y., De Biase, A., Both, S., Langendijk, J. A., van Ooijen, P., Sijtsema, N. M., & van Dijk, L. V. (2023). Swin UNETR for Tumor and Lymph Node Segmentation Using 3D PET/CT Imaging: A Transfer Learning Approach. In V. Andrearczyk, V. Oreiller, A. Depeursinge, & M. Hatt (Eds.), Head and Neck Tumor Segmentation and Outcome Prediction - 3rd Challenge, HECKTOR 2022, Held in Conjunction with MICCAI 2022, Proceedings (pp. 114-120). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13626 LNCS). Springer. https://doi.org/10.1007/978-3-031-27420-6_12

Chu, Hung ; De la O Arévalo, Luis Ricardo ; Tang, Wei et al. / Swin UNETR for Tumor and Lymph Node Segmentation Using 3D PET/CT Imaging : A Transfer Learning Approach. Head and Neck Tumor Segmentation and Outcome Prediction - 3rd Challenge, HECKTOR 2022, Held in Conjunction with MICCAI 2022, Proceedings. editor / Vincent Andrearczyk ; Valentin Oreiller ; Adrien Depeursinge ; Mathieu Hatt. Springer, 2023. pp. 114-120 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{5ec6f1bc322b47c394ac43552fe6f15b,

title = "Swin UNETR for Tumor and Lymph Node Segmentation Using 3D PET/CT Imaging: A Transfer Learning Approach",

abstract = "Delineation of Gross Tumor Volume (GTV) is essential for the treatment of cancer with radiotherapy. GTV contouring is a time-consuming specialized manual task performed by radiation oncologists. Deep Learning (DL) algorithms have shown potential in creating automatic segmentations, reducing delineation time and inter-observer variation. The aim of this work was to create automatic segmentations of primary tumors (GTVp) and pathological lymph nodes (GTVn) in oropharyngeal cancer patients using DL. The organizers of the HECKTOR 2022 challenge provided 3D Computed Tomography (CT) and Positron Emission Tomography (PET) scans with ground-truth GTV segmentations acquired from nine different centers. Bounding box cropping was applied to obtain an anatomic based region of interest. We used the Swin UNETR model in combination with transfer learning. The Swin UNETR encoder weights were initialized by pre-trained weights of a self-supervised Swin UNETR model. An average Dice score of 0.656 was achieved on a test set of 359 patients from the HECKTOR 2022 challenge. Code is available at: https://github.com/HC94/swin\_unetr\_hecktor\_2022.",

keywords = "Auto contouring, Deep learning, Head and neck cancer, HECKTOR 2022, Image processing, Lymph node segmentation, Radiotherapy, Swin UNETR, Tumor segmentation",

author = "Hung Chu and \{De la O Ar{\'e}valo\}, \{Luis Ricardo\} and Wei Tang and Baoqiang Ma and Yan Li and \{De Biase\}, Alessia and Stefan Both and Langendijk, \{Johannes Albertus\} and \{van Ooijen\}, Peter and Sijtsema, \{Nanna Maria\} and \{van Dijk\}, \{Lisanne V.\}",

note = "Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 3rd 3D Head and Neck Tumor Segmentation in PET/CT Challenge, HECKTOR 2022, held in Conjunction with the 25th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2022 ; Conference date: 22-09-2022 Through 22-09-2022",

year = "2023",

doi = "10.1007/978-3-031-27420-6\_12",

language = "English",

isbn = "9783031274190",

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

publisher = "Springer",

pages = "114--120",

editor = "Vincent Andrearczyk and Valentin Oreiller and Adrien Depeursinge and Mathieu Hatt",

booktitle = "Head and Neck Tumor Segmentation and Outcome Prediction - 3rd Challenge, HECKTOR 2022, Held in Conjunction with MICCAI 2022, Proceedings",

address = "Singapore",

}

Chu, H, De la O Arévalo, LR, Tang, W, Ma, B, Li, Y, De Biase, A, Both, S, Langendijk, JA, van Ooijen, P, Sijtsema, NM & van Dijk, LV 2023, Swin UNETR for Tumor and Lymph Node Segmentation Using 3D PET/CT Imaging: A Transfer Learning Approach. in V Andrearczyk, V Oreiller, A Depeursinge & M Hatt (eds), Head and Neck Tumor Segmentation and Outcome Prediction - 3rd Challenge, HECKTOR 2022, Held in Conjunction with MICCAI 2022, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13626 LNCS, Springer, pp. 114-120, 3rd 3D Head and Neck Tumor Segmentation in PET/CT Challenge, HECKTOR 2022, held in Conjunction with the 25th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2022, Singapore, Singapore, 22/09/22. https://doi.org/10.1007/978-3-031-27420-6_12

Swin UNETR for Tumor and Lymph Node Segmentation Using 3D PET/CT Imaging: A Transfer Learning Approach. / Chu, Hung; De la O Arévalo, Luis Ricardo; Tang, Wei et al.
Head and Neck Tumor Segmentation and Outcome Prediction - 3rd Challenge, HECKTOR 2022, Held in Conjunction with MICCAI 2022, Proceedings. ed. / Vincent Andrearczyk; Valentin Oreiller; Adrien Depeursinge; Mathieu Hatt. Springer, 2023. p. 114-120 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13626 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - Swin UNETR for Tumor and Lymph Node Segmentation Using 3D PET/CT Imaging

T2 - 3rd 3D Head and Neck Tumor Segmentation in PET/CT Challenge, HECKTOR 2022, held in Conjunction with the 25th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2022

AU - Chu, Hung

AU - De la O Arévalo, Luis Ricardo

AU - Tang, Wei

AU - Ma, Baoqiang

AU - Li, Yan

AU - De Biase, Alessia

AU - Both, Stefan

AU - Langendijk, Johannes Albertus

AU - van Ooijen, Peter

AU - Sijtsema, Nanna Maria

AU - van Dijk, Lisanne V.

PY - 2023

Y1 - 2023

N2 - Delineation of Gross Tumor Volume (GTV) is essential for the treatment of cancer with radiotherapy. GTV contouring is a time-consuming specialized manual task performed by radiation oncologists. Deep Learning (DL) algorithms have shown potential in creating automatic segmentations, reducing delineation time and inter-observer variation. The aim of this work was to create automatic segmentations of primary tumors (GTVp) and pathological lymph nodes (GTVn) in oropharyngeal cancer patients using DL. The organizers of the HECKTOR 2022 challenge provided 3D Computed Tomography (CT) and Positron Emission Tomography (PET) scans with ground-truth GTV segmentations acquired from nine different centers. Bounding box cropping was applied to obtain an anatomic based region of interest. We used the Swin UNETR model in combination with transfer learning. The Swin UNETR encoder weights were initialized by pre-trained weights of a self-supervised Swin UNETR model. An average Dice score of 0.656 was achieved on a test set of 359 patients from the HECKTOR 2022 challenge. Code is available at: https://github.com/HC94/swin_unetr_hecktor_2022.

AB - Delineation of Gross Tumor Volume (GTV) is essential for the treatment of cancer with radiotherapy. GTV contouring is a time-consuming specialized manual task performed by radiation oncologists. Deep Learning (DL) algorithms have shown potential in creating automatic segmentations, reducing delineation time and inter-observer variation. The aim of this work was to create automatic segmentations of primary tumors (GTVp) and pathological lymph nodes (GTVn) in oropharyngeal cancer patients using DL. The organizers of the HECKTOR 2022 challenge provided 3D Computed Tomography (CT) and Positron Emission Tomography (PET) scans with ground-truth GTV segmentations acquired from nine different centers. Bounding box cropping was applied to obtain an anatomic based region of interest. We used the Swin UNETR model in combination with transfer learning. The Swin UNETR encoder weights were initialized by pre-trained weights of a self-supervised Swin UNETR model. An average Dice score of 0.656 was achieved on a test set of 359 patients from the HECKTOR 2022 challenge. Code is available at: https://github.com/HC94/swin_unetr_hecktor_2022.

KW - Auto contouring

KW - Deep learning

KW - Head and neck cancer

KW - HECKTOR 2022

KW - Image processing

KW - Lymph node segmentation

KW - Radiotherapy

KW - Swin UNETR

KW - Tumor segmentation

UR - https://www.scopus.com/pages/publications/85151064780

U2 - 10.1007/978-3-031-27420-6_12

DO - 10.1007/978-3-031-27420-6_12

M3 - Conference contribution

AN - SCOPUS:85151064780

SN - 9783031274190

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

SP - 114

EP - 120

BT - Head and Neck Tumor Segmentation and Outcome Prediction - 3rd Challenge, HECKTOR 2022, Held in Conjunction with MICCAI 2022, Proceedings

A2 - Andrearczyk, Vincent

A2 - Oreiller, Valentin

A2 - Depeursinge, Adrien

A2 - Hatt, Mathieu

PB - Springer

Y2 - 22 September 2022 through 22 September 2022

ER -

Chu H, De la O Arévalo LR, Tang W, Ma B, Li Y, De Biase A et al. Swin UNETR for Tumor and Lymph Node Segmentation Using 3D PET/CT Imaging: A Transfer Learning Approach. In Andrearczyk V, Oreiller V, Depeursinge A, Hatt M, editors, Head and Neck Tumor Segmentation and Outcome Prediction - 3rd Challenge, HECKTOR 2022, Held in Conjunction with MICCAI 2022, Proceedings. Springer. 2023. p. 114-120. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-27420-6_12