Preselection of robust radiomic features does not improve outcome modelling in non-small cell lung cancer based on clinical routine FDG-PET imaging.
Lung cancer
Multi-centre
PET
Radiomics
Robust
Journal
EJNMMI research
ISSN: 2191-219X
Titre abrégé: EJNMMI Res
Pays: Germany
ID NLM: 101560946
Informations de publication
Date de publication:
21 Aug 2021
21 Aug 2021
Historique:
received:
19
03
2021
accepted:
08
07
2021
entrez:
21
8
2021
pubmed:
22
8
2021
medline:
22
8
2021
Statut:
epublish
Résumé
Radiomics is a promising tool for identifying imaging-based biomarkers. Radiomics-based models are often trained on single-institution datasets; however, multi-centre imaging datasets are preferred for external generalizability owing to the influence of inter-institutional scanning differences and acquisition settings. The study aim was to determine the value of preselection of robust radiomic features in routine clinical positron emission tomography (PET) images to predict clinical outcomes in locally advanced non-small cell lung cancer (NSCLC). A total of 1404 primary tumour radiomic features were extracted from pre-treatment [ Robustness studies identified 179 stable features (13%), with 25% stable features for 3D versus 4D acquisition, 31% for attenuation correction and 78% for delineation. Univariable analysis found no significant robust features predicting 12-/24-month EFS and 12-month OS (p value > 0.076). Prognostic models without robust preselection performed well for 12-month EFS in training (AUC = 0.73) and validation (AUC = 0.74). Patient stratification into two risk groups based on 12-month EFS was significant for training (p value = 0.02) and validation cohorts (p value = 0.03). A PET-based radiomics model using a standardized, multi-centre dataset to predict EFS in locally advanced NSCLC was successfully established and validated with good performance. Prediction models with robust feature preselection were unsuccessful, indicating the need for a standardized imaging protocol.
Sections du résumé
BACKGROUND
BACKGROUND
Radiomics is a promising tool for identifying imaging-based biomarkers. Radiomics-based models are often trained on single-institution datasets; however, multi-centre imaging datasets are preferred for external generalizability owing to the influence of inter-institutional scanning differences and acquisition settings. The study aim was to determine the value of preselection of robust radiomic features in routine clinical positron emission tomography (PET) images to predict clinical outcomes in locally advanced non-small cell lung cancer (NSCLC).
METHODS
METHODS
A total of 1404 primary tumour radiomic features were extracted from pre-treatment [
RESULTS
RESULTS
Robustness studies identified 179 stable features (13%), with 25% stable features for 3D versus 4D acquisition, 31% for attenuation correction and 78% for delineation. Univariable analysis found no significant robust features predicting 12-/24-month EFS and 12-month OS (p value > 0.076). Prognostic models without robust preselection performed well for 12-month EFS in training (AUC = 0.73) and validation (AUC = 0.74). Patient stratification into two risk groups based on 12-month EFS was significant for training (p value = 0.02) and validation cohorts (p value = 0.03).
CONCLUSIONS
CONCLUSIONS
A PET-based radiomics model using a standardized, multi-centre dataset to predict EFS in locally advanced NSCLC was successfully established and validated with good performance. Prediction models with robust feature preselection were unsuccessful, indicating the need for a standardized imaging protocol.
Identifiants
pubmed: 34417899
doi: 10.1186/s13550-021-00809-3
pii: 10.1186/s13550-021-00809-3
pmc: PMC8380219
doi:
Types de publication
Journal Article
Langues
eng
Pagination
79Subventions
Organisme : Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung
ID : 310030_173303
Informations de copyright
© 2021. The Author(s).
Références
Radiology. 2016 Feb;278(2):563-77
pubmed: 26579733
J Med Imaging (Bellingham). 2015 Oct;2(4):041002
pubmed: 26251842
J Nucl Med. 2016 Jun;57(6):842-8
pubmed: 26912429
Insights Imaging. 2015 Apr;6(2):141-55
pubmed: 25763994
Radiology. 2020 May;295(2):328-338
pubmed: 32154773
Eur J Nucl Med Mol Imaging. 2017 Jan;44(1):151-165
pubmed: 27271051
J Nucl Med. 2011 Dec;52 Suppl 2:93S-100S
pubmed: 22144561
Int J Radiat Oncol Biol Phys. 2018 Nov 15;102(4):1074-1082
pubmed: 30170101
Nucl Med Mol Imaging. 2018 Apr;52(2):99-108
pubmed: 29662558
BMC Bioinformatics. 2011 Mar 17;12:77
pubmed: 21414208
Acta Oncol. 2013 Oct;52(7):1398-404
pubmed: 24047338
Eur J Nucl Med Mol Imaging. 2019 Feb;46(2):455-466
pubmed: 30173391
Curr Opin HIV AIDS. 2010 Nov;5(6):463-6
pubmed: 20978388
Eur J Nucl Med Mol Imaging. 2008 Dec;35(12):2320-33
pubmed: 18704407
J Radiat Res. 2019 Jan 1;60(1):150-157
pubmed: 30247662
BJOG. 2015 Feb;122(3):434-43
pubmed: 25623578
Radiology. 2016 Dec;281(3):947-957
pubmed: 27347764
Med Phys. 2019 Apr;46(4):1677-1685
pubmed: 30714158
EJNMMI Res. 2016 Dec;6(1):39
pubmed: 27118538
J Stat Softw. 2010;33(1):1-22
pubmed: 20808728
Sci Rep. 2017 Apr 18;7:46349
pubmed: 28418006
J Nucl Med. 2013 Jan;54(1):19-26
pubmed: 23204495
Med Phys. 2020 Sep;47(9):4045-4053
pubmed: 32395833
PLoS One. 2018 Mar 01;13(3):e0192859
pubmed: 29494598
Lancet. 2015 Sep 12;386(9998):1049-56
pubmed: 26275735
Radiother Oncol. 2018 Nov;129(2):209-217
pubmed: 30279049
PLoS One. 2019 Sep 5;14(9):e0221877
pubmed: 31487307
Int J Radiat Oncol Biol Phys. 2018 Nov 15;102(4):1083-1089
pubmed: 29395627
Clin Radiol. 2019 Jun;74(6):467-473
pubmed: 30898382
Tomography. 2018 Sep;4(3):148-158
pubmed: 30320214
Int J Radiat Oncol Biol Phys. 2017 Nov 15;99(4):921-928
pubmed: 28807534
Sci Rep. 2017 Jun 14;7(1):3519
pubmed: 28615677
Acta Oncol. 2013 Oct;52(7):1391-7
pubmed: 24047337
J Appl Clin Med Phys. 2017 Nov;18(6):32-48
pubmed: 28891217
Psychol Bull. 1979 Mar;86(2):420-8
pubmed: 18839484
Nat Commun. 2014 Jun 03;5:4006
pubmed: 24892406
Radiother Oncol. 2019 Jan;130:2-9
pubmed: 30416044
Radiology. 2016 Jan;278(1):214-22
pubmed: 26176655
Acta Oncol. 2017 Nov;56(11):1531-1536
pubmed: 28820287
J Radiat Res. 2017 Nov 1;58(6):862-869
pubmed: 29036692
Q J Nucl Med Mol Imaging. 2019 Dec;63(4):355-370
pubmed: 31527578
Oncotarget. 2016 Nov 1;7(44):71440-71446
pubmed: 27669756
Contrast Media Mol Imaging. 2018 Sep 10;2018:5324517
pubmed: 30275800
Eur Radiol. 2017 Nov;27(11):4498-4509
pubmed: 28567548
Bioinformatics. 2011 Nov 15;27(22):3206-8
pubmed: 21903630
Jpn J Radiol. 2018 Nov;36(11):686-690
pubmed: 30251115
Oncotarget. 2017 Jun 27;8(26):43169-43179
pubmed: 28574816
Radiother Oncol. 2020 Mar;144:72-78
pubmed: 31733491
Transl Oncol. 2015 Dec;8(6):524-34
pubmed: 26692535
Anticancer Res. 2018 Feb;38(2):685-690
pubmed: 29374691
Lung Cancer. 2018 Jan;115:34-41
pubmed: 29290259
Nucl Med Mol Imaging. 2014 Mar;48(1):16-25
pubmed: 24900134
Nat Rev Clin Oncol. 2017 Dec;14(12):749-762
pubmed: 28975929