Highlights
- •Intramodular Gabor and Law features are highly discriminating–stable radiomic features.
- •Shape features are highly stable but less accurate for distinguishing malignant from benign nodules.
- •Perinodular Gabor features are highly discriminating–stable radiomic features in CTs with slice thickness <3 mm.
- •In slice CTs with thickness >3 mm, most radiomic features tend to become unstable.
Abstract
Objective
To identify stable and discriminating radiomic features on non-contrast CT scans to
develop more generalisable radiomic classifiers for distinguishing granulomas from
adenocarcinomas.
Methods
In total, 412 patients with adenocarcinomas and granulomas from three institutions
were retrospectively included. Segmentations of the lung nodules were performed manually
by an expert radiologist in a 2D axial view. Radiomic features were extracted from
intra- and perinodular regions. A total of 145 patients were used as part of the training
set (, whereas 205 patients were used as part of test set I () and 62 patients were used as part of independent test set II (). To mitigate the variation of CT acquisition parameters, we defined ‘stable’ radiomic
features as those for which the feature expression remains relatively unchanged between
different sites, as assessed using a Wilcoxon rank-sum test. These stable features
were used to develop more generalisable radiomic classifiers that were more resilient
to variations in lung CT scans. Features were ranked based on two criteria, firstly
based on discriminability (i.e. maximising AUC) alone and subsequently based on maximising
both feature stability and discriminability. Different machine-learning classifiers
(Linear discriminant analysis, Quadratic discriminant analysis, Support vector machines
and random forest) were trained with features selected using the two different criteria
and then compared on the two independent test sets for distinguishing granulomas from
adenocarcinomas, in terms of area under the receiver operating characteristic curve.
Results
In the test sets, classifiers constructed using the criteria involving maximising
feature stability and discriminability simultaneously achieved higher AUC compared
with the discriminating alone criteria ( [n = 205]: maximum AUCs of 0.85versus . 0.80; p-value = 0.047 and [n = 62]: maximum AUCs of 0.87 versus. 0.79; p-value = 0.021). These differences
held for features extracted from scans with <3 mm slice thickness (AUC = 0.88 versus.
0.80; p-value = 0.039, n = 100) and for the ≥3 mm cases (AUC = 0.81 versus. 0.76;
p-value = 0.034, n = 105). In both experiments, shape and peritumoural texture features
had a higher stability compared with intratumoural texture features.
Conclusions
Our study suggests that explicitly accounting for both stability and discriminability
results in more generalisable radiomic classifiers to distinguish adenocarcinomas
from granulomas on non-contrast CT scans. Our results also showed that peritumoural
texture and shape features were less affected by the scanner parameters compared with
intratumoural texture features; however, they were also less discriminating compared
with intratumoural features.
Keywords
To read this article in full you will need to make a payment
Purchase one-time access:
Academic & Personal: 24 hour online accessCorporate R&D Professionals: 24 hour online accessOne-time access price info
- For academic or personal research use, select 'Academic and Personal'
- For corporate R&D use, select 'Corporate R&D Professionals'
Subscribe:
Subscribe to European Journal of CancerAlready a print subscriber? Claim online access
Already an online subscriber? Sign in
Register: Create an account
Institutional Access: Sign in to ScienceDirect
References
- Pulmonary nodules: CT evaluation of enhancement with iodinated contrast material.. 1995 Feb; 194: 393-398https://doi.org/10.1148/radiology.194.2.7824716
- Offering lung cancer screening to high-risk medicare beneficiaries saves lives and is cost-effective: an actuarial analysis.American Health & Drug Benefits. 2014 Aug; 7: 272-282
- CT-based radiomic signature predicts distant metastasis in lung adenocarcinoma.Radiother Oncol: journal of the European Society for Therapeutic Radiology and Oncology. 2015; 114: 345-350https://doi.org/10.1016/j.radonc.2015.02.015
- Prognostic value and reproducibility of pretreatment CT texture features in stage III non-small cell lung cancer.Int J Radiat Oncol Biol Phys. 2014; 90: 834-842https://doi.org/10.1016/j.ijrobp.2014.07.020
- Are pretreatment 18F-FDG PET tumor textural features in non-small cell lung cancer associated with response and survival after chemoradiotherapy?.J Nucl Med. 2013 Jan; 54: 19-26https://doi.org/10.2967/jnumed.112.107375
- Breast cancer MRI radiomics: an overview of algorithmic features and impact of inter-reader variability in annotating tumors.Med Phys. 2018 Jul; 45: 3076-3085https://doi.org/10.1002/mp.12925
- Glioblastoma multiforme: exploratory radiogenomic analysis by using quantitative image features.Radiology. 2014 Oct; 273: 168-174https://doi.org/10.1148/radiol.14131731
- Integrated genomic analysis identifies clinically relevant subtypes of glioblastoma characterized by abnormalities in PDGFRA, IDH1, EGFR, and NF1.Canc Cell. 2010 Jan 19; 17: 98https://doi.org/10.1016/j.ccr.2009.12.020
- Perinodular and intranodular radiomic features on lung CT images distinguish adenocarcinomas from granulomas.. 2019 Mar; 290: 783-792https://doi.org/10.1148/radiol.2018180910
- Combination of computer extracted shape and texture features enables discrimination of granulomas from adenocarcinoma on chest computed tomography.J Med Imaging. 2018 Apr; 5024501https://doi.org/10.1117/1.JMI.5.2.024501
- Combination of peri- and intratumoral radiomic features on baseline CT scans predicts response to chemotherapy in lung adenocarcinoma.Radiol Artif Intell. 2019 Mar 20; 1e180012https://doi.org/10.1148/ryai.2019180012
- Predicting pathologic response to neoadjuvant chemoradiation in resectable stage III non-small cell lung cancer patients using computed tomography radiomic features.Lung Canc. 2019 Sep; 135: 1-9https://doi.org/10.1016/j.lungcan.2019.06.020
- Changes in CT radiomic features associated with lymphocytic patterns predict overall survival and response to immunotherapy in non-small cell lung cancer.Cancer Immunol Res. 2020 Jan; 8: 108-119https://doi.org/10.1158/2326-6066.CIR-19-0476
- Novel, non-invasive imaging approach to identify patients with advanced non-small cell lung cancer at risk of hyperprogressive disease with immune checkpoint blockade.J Immunother Cancer. 2020 Oct; 8e001343https://doi.org/10.1136/jitc-2020-001343
- Comparison of the sensitivities of 5 different computed tomography scanners for the assessment of the progression of pulmonary emphysema: a phantom study.Invest Radiol. 2004 Jan; 39: 1-7https://doi.org/10.1097/01.rli.0000091842.82062.a3
- Pulmonary nodule volumetric measurement variability as a function of CT slice thickness and nodule morphology.AJR Am J Roentgenol. 2007 Feb; 188: 306-312https://doi.org/10.2214/AJR.05.1063
- Effects of contrast-enhancement, reconstruction slice thickness and convolution kernel on the diagnostic performance of radiomics signature in solitary pulmonary nodule.Sci Rep. 2016 Oct 10; 6: 34921https://doi.org/10.1038/srep34921
- Radiomics of CT features may Be nonreproducible and redundant: influence of CT acquisition parameters.Radiology. 2018 Aug; 288: 407-415https://doi.org/10.1148/radiol.2018172361
- An integrated segmentation and shape-based classification scheme for distinguishing adenocarcinomas from granulomas on lung CT.Med Phys. 2017; 44: 3556-3569https://doi.org/10.1002/mp.12208
- Defining a radiomic response phenotype: a pilot study using targeted therapy in NSCLC.Sci Rep. 2016 Sep 20; 6: 33860https://doi.org/10.1038/srep33860
- Measuring computed tomography scanner variability of radiomics features.Invest Radiol. 2015 Nov; 50: 757-765https://doi.org/10.1097/RLI.0000000000000180
- Hypoxia and defective apoptosis drive genomic instability and tumorigenesis.Genes Dev. 2004 Sep 1; 18: 2095-2107https://doi.org/10.1101/gad.1204904
- Stable and discriminating radiomic predictor of recurrence in early stage non-small cell lung cancer: multi-site study.Lung Canc. 2020 Apr; 142: 90-97https://doi.org/10.1016/j.lungcan.2020.02.018
- Stable and discriminating features are predictive of cancer presence and Gleason grade in radical prostatectomy specimens: a multi-site study.Sci Rep. 2018 Oct 8; 8: 14918https://doi.org/10.1038/s41598-018-33026-5
- Repeatability of radiomics and machine learning for DWI: short-term repeatability study of 112 patients with prostate cancer.Magn Reson Med. 2020 Jun; 83: 2293-2309https://doi.org/10.1002/mrm.28058
- Multisite evaluation of radiomic feature reproducibility and discriminability for identifying peripheral zone prostate tumors on MRI.J Med Imaging. 2019 Apr; 6024502https://doi.org/10.1117/1.JMI.6.2.024502
- Stability of FDG-PET Radiomics features: an integrated analysis of test-retest and inter-observer variability.Acta Oncol. 2013 Oct; 52: 1391-1397https://doi.org/10.3109/0284186X.2013.812798
- Intrinsic dependencies of CT radiomic features on voxel size and number of gray levels.Med Phys. 2017 Mar; 44: 1050-1062https://doi.org/10.1002/mp.12123
- Invasion of blood vessels as significant prognostic factor in radically resected T1-3N0M0 non-small-cell lung cancer.Eur J Cardio Thorac Surg. 2004 Mar; 25: 439-442https://doi.org/10.1016/j.ejcts.2003.11.033
- Lymphangiogenesis in development and human disease.Nature. 2005 Dec 15; 438: 946-953https://doi.org/10.1038/nature04480
- Peritumoral lymphatic microvessel density is related to poor prognosis in lung adenocarcinoma: a retrospective study of 65 cases.Exp Ther Med. 2012 Apr; 3: 636-640https://doi.org/10.3892/etm.2012.470
- Shashua. Feature selection for unsupervised and supervised inference: the emergence of sparsity in a weight-based approach.J Mach Learn Res. 2005; 6: 1855-1887
- Regression modeling strategies with applications to linear models, logistic and ordinal regression, and survival analysis.Springer-Verlag, 2015https://doi.org/10.1007/978-3-319-19425-7
McInnes L, Healy J, Melville J. Umap: uniform manifold approximation and projection for dimension reduction. arXiv:1802.03426v2. last revised 18 Sep. 2020.
- Preoperative prediction of lymphovascular invasion in invasive breast cancer with dynamic contrast-enhanced-MRI-based radiomics.J Magn Reson Imag. 2019 Feb 17; https://doi.org/10.1002/jmri.26688
- Radiomic MRI signature reveals three distinct subtypes of glioblastoma with different clinical and molecular characteristics, offering prognostic value beyond IDH1.Sci Rep. 2018 Mar 23; 8: 5087https://doi.org/10.1038/s41598-018-22739-2
- Core samples for radiomics features that are insensitive to tumor segmentation: method and pilot study using CT images of hepatocellular carcinoma.J Med Imaging. 2015 Oct; 2041011https://doi.org/10.1117/1.JMI.2.4.041011
- Accounting for reconstruction kernel-induced variability in CT radiomic features using noise power spectra.J Med Imaging. 2018 Jan; 5011013https://doi.org/10.1117/1.JMI.5.1.011013
- Reproducibility and prognosis of quantitative features extracted from CT images.Transl Oncol. 2014 Feb 1; 7: 72-87https://doi.org/10.1593/tlo.13844
- High quality machine-robust image features: identification in nonsmall cell lung cancer computed tomography images.Med Phys. 2013 Dec; 40: 121916https://doi.org/10.1118/1.4829514
- Quantitative assessment of variation in CT parameters on texture features: pilot study using a nonanatomic phantom.AJNR Am J Neuroradiol. 2017 May; 38: 981-985https://doi.org/10.3174/ajnr.A5139
- Can we trust the calculation of texture indices of CT images? A phantom study.Med Phys. 2018 Apr; 45: 1529-1536https://doi.org/10.1002/mp.12809
- Towards guidelines to harmonize textural features in PET: Haralick textural features vary with image noise, but exposure-invariant domains enable comparable PET radiomics.PloS One. 2020 Mar 16; 15e0229560https://doi.org/10.1371/journal.pone.0229560
- Exploring variability in CT characterization of tumors: a preliminary phantom study.Transl Oncol. 2014 Feb 1; 7: 88-93https://doi.org/10.1593/tlo.13865
- Assessing the effect of CT slice interval on unidimensional, bidimensional and volumetric measurements of solid tumours.Canc Imag. 2012 Oct 31; 12: 497-505https://doi.org/10.1102/1470-7330.2012.0046
- Artificial intelligence solution to classify pulmonary nodules on CT.Diagn Interv Imaging. 2020 Dec; 101: 803-810https://doi.org/10.1016/j.diii.2020.10.004
- Multi-model ensemble learning architecture based on 3D CNN for lung nodule malignancy suspiciousness classification.J Digit Imag. 2020 Oct; 33: 1242-1256https://doi.org/10.1007/s10278-020-00372-8
- Quantitative vessel tortuosity: a potential CT imaging biomarker for distinguishing lung granulomas from adenocarcinomas.Sci Rep. 2018 Oct 16; 8: 15290https://doi.org/10.1038/s41598-018-33473-0
- Stability of feature selection algorithms: a study on high-dimensional spaces.Knowl Inf Syst. 2007; 12: 95-116https://doi.org/10.1007/s10115-006-0040-8
- Stable feature selection via dense feature groups.in: Proceedings of the 14th ACM SIGKDD international conference on knowledge discovery and data mining. KDD, New York, NY, USA2008: 803-811https://doi.org/10.1145/1401890.1401986 (ACM (2008))
Article info
Publication history
Published online: March 17, 2021
Accepted:
February 2,
2021
Received in revised form:
January 28,
2021
Received:
August 3,
2020
Identification
Copyright
© 2021 Elsevier Ltd. All rights reserved.
