Deep learning auto-segmentation of cervical skeletal muscle for sarcopenia analysis in patients with head and neck cancer.
auto-segmentation
deep learning
head and neck cancer
sarcopenia
skeletal muscle index
Journal
Frontiers in oncology
ISSN: 2234-943X
Titre abrégé: Front Oncol
Pays: Switzerland
ID NLM: 101568867
Informations de publication
Date de publication:
2022
2022
Historique:
received:
27
04
2022
accepted:
29
06
2022
entrez:
15
8
2022
pubmed:
16
8
2022
medline:
16
8
2022
Statut:
epublish
Résumé
Sarcopenia is a prognostic factor in patients with head and neck cancer (HNC). Sarcopenia can be determined using the skeletal muscle index (SMI) calculated from cervical neck skeletal muscle (SM) segmentations. However, SM segmentation requires manual input, which is time-consuming and variable. Therefore, we developed a fully-automated approach to segment cervical vertebra SM. 390 HNC patients with contrast-enhanced CT scans were utilized (300-training, 90-testing). Ground-truth single-slice SM segmentations at the C3 vertebra were manually generated. A multi-stage deep learning pipeline was developed, where a 3D ResUNet auto-segmented the C3 section (33 mm window), the middle slice of the section was auto-selected, and a 2D ResUNet auto-segmented the auto-selected slice. Both the 3D and 2D approaches trained five sub-models (5-fold cross-validation) and combined sub-model predictions on the test set using majority vote ensembling. Model performance was primarily determined using the Dice similarity coefficient (DSC). Predicted SMI was calculated using the auto-segmented SM cross-sectional area. Finally, using established SMI cutoffs, we performed a Kaplan-Meier analysis to determine associations with overall survival. Mean test set DSC of the 3D and 2D models were 0.96 and 0.95, respectively. Predicted SMI had high correlation to the ground-truth SMI in males and females (r>0.96). Predicted SMI stratified patients for overall survival in males (log-rank p = 0.01) but not females (log-rank p = 0.07), consistent with ground-truth SMI. We developed a high-performance, multi-stage, fully-automated approach to segment cervical vertebra SM. Our study is an essential step towards fully-automated sarcopenia-related decision-making in patients with HNC.
Sections du résumé
Background/Purpose
UNASSIGNED
Sarcopenia is a prognostic factor in patients with head and neck cancer (HNC). Sarcopenia can be determined using the skeletal muscle index (SMI) calculated from cervical neck skeletal muscle (SM) segmentations. However, SM segmentation requires manual input, which is time-consuming and variable. Therefore, we developed a fully-automated approach to segment cervical vertebra SM.
Materials/Methods
UNASSIGNED
390 HNC patients with contrast-enhanced CT scans were utilized (300-training, 90-testing). Ground-truth single-slice SM segmentations at the C3 vertebra were manually generated. A multi-stage deep learning pipeline was developed, where a 3D ResUNet auto-segmented the C3 section (33 mm window), the middle slice of the section was auto-selected, and a 2D ResUNet auto-segmented the auto-selected slice. Both the 3D and 2D approaches trained five sub-models (5-fold cross-validation) and combined sub-model predictions on the test set using majority vote ensembling. Model performance was primarily determined using the Dice similarity coefficient (DSC). Predicted SMI was calculated using the auto-segmented SM cross-sectional area. Finally, using established SMI cutoffs, we performed a Kaplan-Meier analysis to determine associations with overall survival.
Results
UNASSIGNED
Mean test set DSC of the 3D and 2D models were 0.96 and 0.95, respectively. Predicted SMI had high correlation to the ground-truth SMI in males and females (r>0.96). Predicted SMI stratified patients for overall survival in males (log-rank p = 0.01) but not females (log-rank p = 0.07), consistent with ground-truth SMI.
Conclusion
UNASSIGNED
We developed a high-performance, multi-stage, fully-automated approach to segment cervical vertebra SM. Our study is an essential step towards fully-automated sarcopenia-related decision-making in patients with HNC.
Identifiants
pubmed: 35965493
doi: 10.3389/fonc.2022.930432
pmc: PMC9366009
doi:
Types de publication
Journal Article
Langues
eng
Pagination
930432Subventions
Organisme : NIDCR NIH HHS
ID : F31 DE031502
Pays : United States
Organisme : NIDCR NIH HHS
ID : R01 DE028290
Pays : United States
Informations de copyright
Copyright © 2022 Naser, Wahid, Grossberg, Olson, Jain, El-Habashy, Dede, Salama, Abobakr, Mohamed, He, Jaskari, Sahlsten, Kaski and Fuller.
Déclaration de conflit d'intérêts
CF has received direct industry grant support, speaking honoraria, and travel funding from Elekta AB. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Références
Sci Data. 2022 Aug 2;9(1):470
pubmed: 35918336
Eur Radiol. 2021 Jun;31(6):4053-4062
pubmed: 33219847
Acad Radiol. 2020 Mar;27(3):311-320
pubmed: 31126808
Radiother Oncol. 2020 Jun;147:103-110
pubmed: 32251949
Tech Innov Patient Support Radiat Oncol. 2020 Nov 09;16:50-57
pubmed: 33385074
Pract Radiat Oncol. 2021 May-Jun;11(3):226-229
pubmed: 33607331
Clin Nutr. 2020 Oct;39(10):3049-3055
pubmed: 32007318
Sci Data. 2017 Jul 04;4:170077
pubmed: 28675381
Front Oncol. 2020 Feb 05;9:1561
pubmed: 32117787
Comput Biol Med. 2020 Jun;121:103758
pubmed: 32568668
AJR Am J Roentgenol. 2022 Jan;218(1):124-131
pubmed: 34406056
Cancer Nurs. 2015 Nov-Dec;38(6):E37-45
pubmed: 25730590
J Digit Imaging. 2013 Dec;26(6):1045-57
pubmed: 23884657
J Clin Med. 2021 Feb 03;10(4):
pubmed: 33546506
JAMA Oncol. 2016 Jun 01;2(6):782-9
pubmed: 26891703
Radiat Oncol. 2015 Jan 17;10:21
pubmed: 25679310
Front Oncol. 2018 Oct 22;8:457
pubmed: 30460194
Clin Transl Radiat Oncol. 2021 Oct 16;32:6-14
pubmed: 34765748
Ther Adv Med Oncol. 2021 May 13;13:17588359211008844
pubmed: 34035838
Br J Radiol. 2019 Aug;92(1100):20190327
pubmed: 31199670
BMC Med Imaging. 2015 Aug 12;15:29
pubmed: 26263899
Oral Oncol. 2016 Nov;62:28-33
pubmed: 27865369
Clin Nutr. 2021 Apr;40(4):1711-1718
pubmed: 32994071
Head Neck. 2019 Mar;41(3):707-714
pubmed: 30582237
Age Ageing. 2010 Jul;39(4):412-23
pubmed: 20392703
J Clin Oncol. 2013 Apr 20;31(12):1539-47
pubmed: 23530101
Oral Oncol. 2016 Oct;61:62-9
pubmed: 27688106
JAMA Otolaryngol Head Neck Surg. 2019 Jul 1;145(7):647-654
pubmed: 31169874
Head Neck Tumor Segm Chall (2021). 2022;13209:121-132
pubmed: 35399869
Front Oncol. 2022 Feb 14;12:812159
pubmed: 35237517
Front Oncol. 2021 May 07;11:580806
pubmed: 34026597
J Clin Med. 2018 Apr 08;7(4):
pubmed: 29642478
Med Image Anal. 2020 Jul;63:101693
pubmed: 32289663
Support Care Cancer. 2020 Oct;28(10):4613-4625
pubmed: 32533436
Lancet Oncol. 2008 Jul;9(7):629-35
pubmed: 18539529
Clin Radiol. 2022 May;77(5):e363-e371
pubmed: 35260232