Interpretability of radiomics models is improved when using feature group selection strategies for predicting molecular and clinical targets in clear-cell renal cell carcinoma: insights from the TRACERx Renal study.
Feature selection
Group selection
Histology
Interpretable
Machine learning
Molecular subtyping
Nested validation
Radiogenomics
Radiomics
Renal cancer
Journal
Cancer imaging : the official publication of the International Cancer Imaging Society
ISSN: 1470-7330
Titre abrégé: Cancer Imaging
Pays: England
ID NLM: 101172931
Informations de publication
Date de publication:
14 Aug 2023
14 Aug 2023
Historique:
received:
10
01
2023
accepted:
12
07
2023
medline:
16
8
2023
pubmed:
15
8
2023
entrez:
14
8
2023
Statut:
epublish
Résumé
The aim of this work is to evaluate the performance of radiomics predictions for a range of molecular, genomic and clinical targets in patients with clear cell renal cell carcinoma (ccRCC) and demonstrate the impact of novel feature selection strategies and sub-segmentations on model interpretability. Contrast-enhanced CT scans from the first 101 patients recruited to the TRACERx Renal Cancer study (NCT03226886) were used to derive radiomics classification models to predict 20 molecular, histopathology and clinical target variables. Manual 3D segmentation was used in conjunction with automatic sub-segmentation to generate radiomics features from the core, rim, high and low enhancing sub-regions, and the whole tumour. Comparisons were made between two classification model pipelines: a Conventional pipeline reflecting common radiomics practice, and a Proposed pipeline including two novel feature selection steps designed to improve model interpretability. For both pipelines nested cross-validation was used to estimate prediction performance and tune model hyper-parameters, and permutation testing was used to evaluate the statistical significance of the estimated performance measures. Further model robustness assessments were conducted by evaluating model variability across the cross-validation folds. Classification performance was significant (p < 0.05, H Use of the Proposed pipeline, which includes the novel feature selection methods, leads to more interpretable models without compromising prediction performance. NCT03226886 (TRACERx Renal).
Sections du résumé
BACKGROUND
BACKGROUND
The aim of this work is to evaluate the performance of radiomics predictions for a range of molecular, genomic and clinical targets in patients with clear cell renal cell carcinoma (ccRCC) and demonstrate the impact of novel feature selection strategies and sub-segmentations on model interpretability.
METHODS
METHODS
Contrast-enhanced CT scans from the first 101 patients recruited to the TRACERx Renal Cancer study (NCT03226886) were used to derive radiomics classification models to predict 20 molecular, histopathology and clinical target variables. Manual 3D segmentation was used in conjunction with automatic sub-segmentation to generate radiomics features from the core, rim, high and low enhancing sub-regions, and the whole tumour. Comparisons were made between two classification model pipelines: a Conventional pipeline reflecting common radiomics practice, and a Proposed pipeline including two novel feature selection steps designed to improve model interpretability. For both pipelines nested cross-validation was used to estimate prediction performance and tune model hyper-parameters, and permutation testing was used to evaluate the statistical significance of the estimated performance measures. Further model robustness assessments were conducted by evaluating model variability across the cross-validation folds.
RESULTS
RESULTS
Classification performance was significant (p < 0.05, H
CONCLUSIONS
CONCLUSIONS
Use of the Proposed pipeline, which includes the novel feature selection methods, leads to more interpretable models without compromising prediction performance.
TRIAL REGISTRATION
BACKGROUND
NCT03226886 (TRACERx Renal).
Identifiants
pubmed: 37580840
doi: 10.1186/s40644-023-00594-3
pii: 10.1186/s40644-023-00594-3
pmc: PMC10424427
doi:
Banques de données
ClinicalTrials.gov
['NCT03226886']
Types de publication
Clinical Trial
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Pagination
76Subventions
Organisme : MRF
ID : FC10988
Pays : United Kingdom
Organisme : Cancer Research UK
ID : A29911
Pays : United Kingdom
Organisme : NCI NIH HHS
ID : U01 CA247439
Pays : United States
Organisme : Wellcome Trust
Pays : United Kingdom
Organisme : Melanoma Research Alliance
ID : 686061
Pays : United States
Organisme : Cancer Research UK
ID : FC10988
Pays : United Kingdom
Organisme : Wellcome Trust
ID : FC10988
Pays : United Kingdom
Informations de copyright
© 2023. International Cancer Imaging Society (ICIS).
Références
Turajlic S, Swanton C, T.R. Consortium. TRACERx Renal: tracking renal cancer evolution through therapy. Nat Rev Urol. 2017;14(10):575–6.
doi: 10.1038/nrurol.2017.112
Turajlic S, et al. Tracking cancer evolution reveals constrained routes to metastases: TRACERx renal. Cell. 2018;173(3):581-594.e12.
doi: 10.1016/j.cell.2018.03.057
pubmed: 29656895
pmcid: 5938365
Turajlic S, et al. Deterministic evolutionary trajectories influence primary tumor growth: TRACERx renal. Cell. 2018;173(3):595-610.e11.
doi: 10.1016/j.cell.2018.03.043
pubmed: 29656894
pmcid: 5938372
Lambin P, et al. Radiomics: the bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol. 2017;14(12):749–62.
doi: 10.1038/nrclinonc.2017.141
pubmed: 28975929
van Timmeren JE, et al. Radiomics in medical imaging-“how-to” guide and critical reflection. Insights Imaging. 2020;11(1):91.
doi: 10.1186/s13244-020-00887-2
pubmed: 32785796
pmcid: 7423816
Papanikolaou N, Matos C, Koh DM. How to develop a meaningful radiomic signature for clinical use in oncologic patients. Cancer Imaging. 2020;20(1):33.
doi: 10.1186/s40644-020-00311-4
pubmed: 32357923
pmcid: 7195800
Tomaszewski MR, Gillies RJ. The biological meaning of radiomic features. Radiology. 2021;298(3):505–16.
doi: 10.1148/radiol.2021202553
pubmed: 33399513
Sugai Y, et al. Impact of feature selection methods and subgroup factors on prognostic analysis with CT-based radiomics in non-small cell lung cancer patients. Radiat Oncol. 2021;16(1):80.
doi: 10.1186/s13014-021-01810-9
pubmed: 33931085
pmcid: 8086112
Laajili R, Said M, Tagina M. Application of radiomics features selection and classification algorithms for medical imaging decision: MRI radiomics breast cancer cases study. Inform Med Unlocked. 2021;27:100801.
doi: 10.1016/j.imu.2021.100801
Demircioğlu A. Benchmarking feature selection methods in radiomics. Invest Radiol. 2022;57(7):433–43.
doi: 10.1097/RLI.0000000000000855
pubmed: 35045555
Shur JD, et al. Radiomics in oncology: a practical guide. Radiographics. 2021;41(6):1717–32.
doi: 10.1148/rg.2021210037
pubmed: 34597235
Marcus DS, et al. The Extensible Neuroimaging Archive Toolkit: an informatics platform for managing, exploring, and sharing neuroimaging data. Neuroinformatics. 2007;5(1):11–34.
doi: 10.1385/NI:5:1:11
pubmed: 17426351
Zwanenburg A, et al. The image biomarker standardization initiative: standardized quantitative radiomics for high-throughput image-based phenotyping. Radiology. 2020;295(2):328–38.
doi: 10.1148/radiol.2020191145
pubmed: 32154773
van Griethuysen JJM, et al. Computational radiomics system to decode the radiographic phenotype. Cancer Res. 2017;77(21):e104–7.
doi: 10.1158/0008-5472.CAN-17-0339
pubmed: 29092951
pmcid: 5672828
Tibshirani R. Regression shrinkage and selection via the Lasso. J R Stat Soc Ser B Methodol. 1996;58(1):267–88.
Breiman L. Random forests. Mach Learn. 2001;45(1):5–32.
doi: 10.1023/A:1010933404324
Fisher A, Rudin C, Dominici F. All models are wrong, but many are useful: learning a variable’s importance by studying an entire class of prediction models simultaneously. J Mach Learn Res. 2019;20:177.
pubmed: 34335110
pmcid: 8323609
Ojala M, Garriga GC. Permutation tests for studying classifier performance. J Mach Learn Res. 2010;11(6):1833–63.
Fu X, et al. Spatial patterns of tumour growth impact clonal diversification in a computational model and the TRACERx Renal study. Nat Ecol Evol. 2022;6(1):88–102.
doi: 10.1038/s41559-021-01586-x
pubmed: 34949820
Chen X, et al. Reliable gene mutation prediction in clear cell renal cell carcinoma through multi-classifier multi-objective radiogenomics model. Phys Med Biol. 2018;63(21):215008.
doi: 10.1088/1361-6560/aae5cd
pubmed: 30277889
pmcid: 6240911
Feng Z, et al. Identifying BAP1 mutations in clear-cell renal cell carcinoma by CT radiomics: preliminary findings. Front Oncol. 2020;10:279.
doi: 10.3389/fonc.2020.00279
pubmed: 32185138
pmcid: 7058626
Kocak B, et al. Radiogenomics in clear cell renal cell carcinoma: machine learning-based high-dimensional quantitative CT texture analysis in predicting PBRM1 mutation status. AJR Am J Roentgenol. 2019;212(3):W55–63.
doi: 10.2214/AJR.18.20443
pubmed: 30601030
Zhao Y, et al. Selection of metastasis competent subclones in the tumour interior. Nat Ecol Evol. 2021;5(7):1033–45.
doi: 10.1038/s41559-021-01456-6
pubmed: 34002049
pmcid: 7611703
Klau S, et al. Priority-Lasso: a simple hierarchical approach to the prediction of clinical outcome using multi-omics data. BMC Bioinformatics. 2018;19(1):322.
doi: 10.1186/s12859-018-2344-6
pubmed: 30208855
pmcid: 6134797
Carrasquinha E, et al. Regularization techniques in radiomics: a case study on the prediction of pCR in breast tumours and the axilla. In: Computational intelligence methods for bioinformatics and biostatistics. Cham: Springer International Publishing; 2020.
Yuan M, Lin Y. Model selection and estimation in regression with grouped variables. J R Stat Soc Ser B Stat Methodol. 2006;68(1):49–67.
doi: 10.1111/j.1467-9868.2005.00532.x