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 voor dit werk

Onderzoeksoutput › Academic › peer review

8 Citaten (Scopus)

202 Downloads (Pure)

Samenvatting

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.

Originele taal-2	English
Titel	Head and Neck Tumor Segmentation and Outcome Prediction - 3rd Challenge, HECKTOR 2022, Held in Conjunction with MICCAI 2022, Proceedings
Redacteuren	Vincent Andrearczyk, Valentin Oreiller, Adrien Depeursinge, Mathieu Hatt
Uitgeverij	Springer Science and Business Media Deutschland GmbH
Pagina's	114-120
Aantal pagina's	7
ISBN van geprinte versie	9783031274190
DOI's	https://doi.org/10.1007/978-3-031-27420-6_12
Status	Published - 2023
Evenement	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 Duur: 22-sep.-2022 → 22-sep.-2022

Publicatie series

Naam	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13626 LNCS
ISSN van geprinte versie	0302-9743
ISSN van elektronische versie	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
Land/Regio	Singapore
Stad	Singapore
Periode	22/09/2022 → 22/09/2022

Toegang tot document

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

Swin UNETR for Tumor and Lymph Node Segmentation Using 3D PET/CT ImagingFinal publisher's version, 894 KBLicentie: Taverne

Handle.net

Permanente koppeling

Citeer dit

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 (editors), Head and Neck Tumor Segmentation and Outcome Prediction - 3rd Challenge, HECKTOR 2022, Held in Conjunction with MICCAI 2022, Proceedings (blz. 114-120). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13626 LNCS). Springer Science and Business Media Deutschland GmbH. 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. redacteur / Vincent Andrearczyk ; Valentin Oreiller ; Adrien Depeursinge ; Mathieu Hatt. Springer Science and Business Media Deutschland GmbH, 2023. blz. 114-120 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{f63beabad23d446fb01a7edcaec839e7,

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 = "Funding Information: We thank the Center for Information Technology of the University of Groningen for their support and for providing access to the Peregrine high performance computing cluster. 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 Science and Business Media Deutschland GmbH",

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",

}

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 (uitgave), 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 Science and Business Media Deutschland GmbH, blz. 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, 22/09/2022. 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. uitgave / Vincent Andrearczyk; Valentin Oreiller; Adrien Depeursinge; Mathieu Hatt. Springer Science and Business Media Deutschland GmbH, 2023. blz. 114-120 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13626 LNCS).

Onderzoeksoutput › 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.

N1 - Funding Information: We thank the Center for Information Technology of the University of Groningen for their support and for providing access to the Peregrine high performance computing cluster. Publisher Copyright: © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

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

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 Science and Business Media Deutschland GmbH

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, redacteurs, Head and Neck Tumor Segmentation and Outcome Prediction - 3rd Challenge, HECKTOR 2022, Held in Conjunction with MICCAI 2022, Proceedings. Springer Science and Business Media Deutschland GmbH. 2023. blz. 114-120. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). Epub 2023 mrt. 18. doi: 10.1007/978-3-031-27420-6_12