Advertisement

Radiomic Features Are Associated With EGFR Mutation Status in Lung Adenocarcinomas

  • Ying Liu
    Affiliations
    Department of Radiology, Tianjin Medical University Cancer Institute and Hospital, National Clinical Research Center of Cancer, Key Laboratory of Cancer Prevention and Therapy, Tianjin, PR China

    Department of Cancer Imaging and Metabolism, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL
    Search for articles by this author
  • Jongphil Kim
    Affiliations
    Department of Biostatistics and Bioinformatics, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL
    Search for articles by this author
  • Yoganand Balagurunathan
    Affiliations
    Department of Cancer Imaging and Metabolism, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL
    Search for articles by this author
  • Qian Li
    Affiliations
    Department of Radiology, Tianjin Medical University Cancer Institute and Hospital, National Clinical Research Center of Cancer, Key Laboratory of Cancer Prevention and Therapy, Tianjin, PR China

    Department of Cancer Imaging and Metabolism, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL
    Search for articles by this author
  • Alberto L. Garcia
    Affiliations
    Department of Cancer Imaging and Metabolism, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL
    Search for articles by this author
  • Olya Stringfield
    Affiliations
    Department of Cancer Imaging and Metabolism, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL
    Search for articles by this author
  • Zhaoxiang Ye
    Correspondence
    Zhaoxiang Ye, MD, Tianjin Medical University Cancer Institute and Hospital, Huan-Hu-Xi Rd, Ti-Yuan-Bei, He Xi District, Tianjin 300060, PR China. Fax: +86-22-23537796
    Affiliations
    Department of Radiology, Tianjin Medical University Cancer Institute and Hospital, National Clinical Research Center of Cancer, Key Laboratory of Cancer Prevention and Therapy, Tianjin, PR China
    Search for articles by this author
  • Robert J. Gillies
    Correspondence
    Addresses for correspondence: Robert J. Gillies, PhD, H. Lee Moffitt Cancer Center and Research Institute, 12902 Magnolia Dr, Tampa, FL 33612. Fax: 813-745-7265
    Affiliations
    Department of Cancer Imaging and Metabolism, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL

    Department of Radiology, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL
    Search for articles by this author
Open AccessPublished:February 15, 2016DOI:https://doi.org/10.1016/j.cllc.2016.02.001

      Abstract

      Background

      In this study we retrospectively evaluated the capability of computed tomography (CT)-based radiomic features to predict epidermal growth factor receptor (EGFR) mutation status in surgically-resected peripheral lung adenocarcinomas in an Asian cohort of patients.

      Patients and Methods

      Two hundred ninety-eight patients with surgically resected peripheral lung adenocarcinomas were investigated in this institutional review board-approved retrospective study with requirement waived to obtain informed consent. Two hundred nineteen quantitative 3-D features were extracted from segmented volumes of each tumor, and 59 of these, which were considered independent features, were included in the analysis. Clinical and pathological information was obtained from the institutional database.

      Results

      Mutant EGFR was significantly associated with female sex (P = .0005); never smoker status (P < .0001), lepidic predominant adenocarcinomas (P = .017), and low or intermediate pathologic grade (P = .0002). Statistically significant differences were found in 11 radiomic features between EGFR mutant and wild type groups in univariate analysis. Mutant EGFR status could be predicted by a set of 5 radiomic features that fell into 3 broad groups: CT attenuation energy, tumor main direction, and texture defined according to wavelets and Laws (area under the curve [AUC], 0.647). A multiple logistic regression model showed that adding radiomic features to a clinical model resulted in a significant improvement of predicting power, because the AUC increased from 0.667 to 0.709 (P < .0001).

      Conclusion

      Computed tomography-based radiomic features of peripheral lung adenocarcinomas can capture useful information regarding tumor phenotype, and the model we built can be useful to predict the presence of EGFR mutations in peripheral lung adenocarcinoma in Asian patients when mutational profiling is not available or possible.

      Keywords

      Introduction

      Lung cancer accounts for 13% of new cancers diagnosed worldwide and is the leading cause of cancer-related deaths.
      • Halpenny D.F.
      • Riely G.J.
      • Hayes S.
      • et al.
      Are there imaging characteristics associated with lung adenocarcinomas harboring ALK rearrangements?.
      Of all lung cancers 85% are non–small-cell lung carcinoma (NSCLC), with adenocarcinoma being the most common histologic subtype.
      • Ganeshan B.
      • Panayiotou E.
      • Burnand K.
      • Dizdarevic S.
      • Miles K.
      Tumour heterogeneity in non–small-cell lung carcinoma assessed by CT texture analysis: a potential marker of survival.
      Over the past decade, molecular translational research advances have heralded major breakthroughs in the understanding, diagnosis, and management of lung cancer, particularly the development of new target-based therapies directed against key signaling pathways involved in lung cancer growth and malignant progression.
      • Mok T.S.
      • Wu Y.L.
      • Thongprasert S.
      • et al.
      Gefitinib or carboplatin–paclitaxel in pulmonary adenocarcinoma.
      • Jackman D.M.
      • Miller V.A.
      • Cioffredi L.A.
      • et al.
      Impact of epidermal growth factor receptor and KRAS mutations on clinical outcomes in previously untreated non–small-cell lung cancer patients: results of an online tumor registry of clinical trials.
      Small-molecule tyrosine kinase inhibitors (TKIs) that target the epidermal growth factor receptor (EGFR) were the first targeted drugs to enter clinical use for the treatment of NSCLC. Patients with EGFR mutations have a higher response rate to EGFR TKIs (60%-80%) than those with EGFR wild type or unknown mutation status (10%-20%).
      • Riely G.J.
      • Pao W.
      • Pham D.
      • et al.
      Clinical course of patients with non–small-cell lung cancer and epidermal growth factor receptor exon 19 and exon 21 mutations treated with gefitinib or erlotinib.
      Randomized trials have clearly shown that treatments with targeted TKIs, such as erlotinib, gefitinib, or afatinib are associated with longer progression-free survival (PFS) and higher objective radiographic response rates than standard first-line chemotherapy in patients with mutated EGFR lung cancer.
      • Yang J.C.
      • Schuler M.H.
      • Yamamoto N.
      • et al.
      LUX-Lung 3: a randomized, open-label, phase III study of afatinib versus pemetrexed and cisplatin as first-line treatment for patients with advanced adenocarcinoma of the lung harboring EGFR-activating mutations.
      • Maemondo M.
      • Inoue A.
      • Kobayashi K.
      • et al.
      Gefitinib or chemotherapy for non–small-cell lung cancer with mutated EGFR.
      • Kobayashi K.
      • Inoue A.
      • Maemondo M.
      • et al.
      First-line gefitinib versus first-line chemotherapy by carboplatin (CBDCA) plus paclitaxel (TXL) in non–small-cell lung cancer (NSCLC) patients (pts) with EGFR mutations: a phase III study (002) by North East Japan Gefitinib Study Group.
      • Mitsudomi T.
      • Morita S.
      • Yatabe Y.
      • et al.
      Gefitinib versus cisplatin plus docetaxel in patients with non–small-cell lung cancer harbouring mutations of the epidermal growth factor receptor (WJTOG3405): an open label, randomised phase 3 trial.
      However, if gefitinib is administered in the case of non-EGFR mutated lung cancer, the patient will experience a shorter PFS compared with platinum-based chemotherapy,
      • Mok T.S.
      • Wu Y.L.
      • Thongprasert S.
      • et al.
      Gefitinib or carboplatin–paclitaxel in pulmonary adenocarcinoma.
      highlighting the importance of identifying this genetically unique subset of patients.
      Cumulative epidemiology studies have identified several clinicopathological factors such as female, nonsmoker, adenocarcinoma histology, and East Asian origin that were associated with a high prevalence of EGFR mutation.
      • Paez J.G.
      • Jänne P.A.
      • Lee J.C.
      • et al.
      EGFR mutations in lung cancer: correlation with clinical response to gefitinib therapy.
      • Kosaka T.
      • Yatabe Y.
      • Endoh H.
      • Kuwano H.
      • Takahashi T.
      • Mitsudomi T.
      Mutations of the epidermal growth factor receptor gene in lung cancer biological and clinical implications.
      Unfortunately, there are no reliable clinical characteristics that allow for accurate prediction of EGFR mutation status. For some patients, biopsy samples might be the only tumor materials available for testing EGFR mutation status and they are often composed of variable ratios of tumor to normal cells.
      • Shahi R.B.
      • De Brakeleer S.
      • De Greve J.
      • Geers C.
      • In't Veld P.
      • Teugels E.
      Detection of EGFR-TK domain-activating mutations in NSCLC with generic PCR-based methods.
      Thus, mutant DNA alleles present at extremely low concentrations become difficult to detect, leading to a false negative result. Further, because of intratumoral heterogeneity, the portion of the tumor tested for EGFR mutation might also result as negative but might be truly positive.
      • Tomonaga N.
      • Nakamura Y.
      • Yamaguchi H.
      • et al.
      Analysis of intratumor heterogeneity of EGFR mutations in mixed type lung adenocarcinoma.
      • Bai H.
      • Wang Z.
      • Wang Y.
      • et al.
      Detection and clinical significance of intratumoral EGFR mutational heterogeneity in Chinese patients with advanced non–small-cell lung cancer.
      • Taniguchi K.
      • Okami J.
      • Kodama K.
      • Higashiyama M.
      • Kato K.
      Intratumor heterogeneity of epidermal growth factor receptor mutations in lung cancer and its correlation to the response to gefitinib.
      In a recent study,
      • Querings S.
      • Altmüller J.
      • Ansén S.
      • et al.
      Benchmarking of mutation diagnostics in clinical lung cancer specimens.
      2 sequencing-based mutation detection approaches (dideoxy and pyrosequencing) were validated against parallel sequencing in a clinical setting, and the results showed that dideoxy sequencing missed 4 responders and pyrosequencing missed 2 responders; meanwhile, precise quantification of mutant alleles revealed a low correlation of histopathological estimates of tumor content and frequency of mutant alleles, indicating that sequencing technologies with inferior sensitivity might fail to detect clinically relevant oncogene mutations in cancer patients. Therefore, when receiving a negative mutation analysis result, one must consider whether the cell sample was truly representative for the EGFR mutation status of the lung tumor.
      There have been several reports regarding the relationship between computed tomography (CT) features and EGFR mutation status in NSCLC
      • Rizzo S.
      • Petrella F.
      • Buscarino V.
      • et al.
      CT radiogenomic characterization of EGFR, K-RAS, and ALK mutations in non–small -cell lung cancer.
      • Zhou J.Y.
      • Zheng J.
      • Yu Z.F.
      • et al.
      Comparative analysis of clinicoradiologic characteristics of lung adenocarcinomas with ALK rearrangements or EGFR mutations.
      • Lee Y.
      • Lee H.J.
      • Kim Y.T.
      • et al.
      Imaging characteristics of stage I non–small-cell lung cancer on CT and FDG-PET: relationship with epidermal growth factor receptor protein expression status and survival.
      • Lee H.J.
      • Kim Y.T.
      • Kang C.H.
      • et al.
      Epidermal growth factor receptor mutation in lung adenocarcinomas: relationship with CT characteristics and histologic subtypes.
      ; however, the findings were not consistent with each other. According to a study conducted by Zhou et al,
      • Zhou J.Y.
      • Zheng J.
      • Yu Z.F.
      • et al.
      Comparative analysis of clinicoradiologic characteristics of lung adenocarcinomas with ALK rearrangements or EGFR mutations.
      there were no differences in morphological CT features between EGFR mutation and wild type tumors. In contrast, Rizzo et al
      • Rizzo S.
      • Petrella F.
      • Buscarino V.
      • et al.
      CT radiogenomic characterization of EGFR, K-RAS, and ALK mutations in non–small -cell lung cancer.
      reported that EGFR mutation was significantly associated with air bronchogram, pleural retraction, small lesion size, and absence of fibrosis. Recent technological advances in medical imaging allow high-throughput extraction of quantitative imaging features. Radiomics is the process of converting images to mineable data through computational approaches. These data can be used to develop decision support systems to accurately estimate patient risk and improve individualized treatment.
      • Kumar V.
      • Gu Y.
      • Basu S.
      • et al.
      Radiomics: the process and the challenges.
      • Parmar C.
      • Velazquez E.R.
      • Leijenaar R.
      • et al.
      Robust radiomics feature quantification using semiautomatic volumetric segmentation.
      Several studies have shown that such features extracted from CT images of lung cancers can be useful to distinguish radiation-induced fibrosis from tumor recurrence,
      • Mattonen S.A.
      • Palma D.A.
      • Haasbeek C.J.
      • Senan S.
      • Ward A.D.
      Early prediction of tumor recurrence based on CT texture changes after stereotactic ablative radiotherapy (SABR) for lung cancer.
      differentiate the presence of Kirsten rat sarcoma viral oncogene (K-ras) mutation from pan wild type NSCLC,
      • Weiss G.J.
      • Ganeshan B.
      • Miles K.A.
      • et al.
      Noninvasive image texture analysis differentiates K-ras mutation from pan-wildtype NSCLC and is prognostic.
      provide an independent predictive indicator of response to first-line chemotherapy,
      • Ravanelli M.
      • Farina D.
      • Morassi M.
      • et al.
      Texture analysis of advanced non–small-cell lung cancer (NSCLC) on contrast-enhanced computed tomography: prediction of the response to the first-line chemotherapy.
      and identify patients with locally advanced lung adenocarcinoma at risk of developing distant metastasis.
      • Coroller T.P.
      • Grossmann P.
      • Hou Y.
      • et al.
      CT-based radiomic signature predicts distant metastasis in lung adenocarcinoma.
      CT imaging is routinely used in lung cancer, and we thus hypothesize that if CT-based radiomic features associated with EGFR mutation status can be determined, they could provide a useful clinical predictor in patients with unresectable lung cancer or those in whom biopsy is unable to be performed. Imaging-based risk models might also provide additional information for clinicians on whether rebiopsy is needed for patients with a negative EGFR mutation result. Therefore, in this retrospective study, we performed a radiomic analysis to identify image biomarkers of harboring the EGFR mutation in peripheral lung adenocarcinomas in a Chinese cohort of patients.

      Patients and Methods

      This retrospective study was approved by the institutional review board. Requirement for informed consent was waived.

      Study Population

      A consecutive search of the surgical database at our institution between December 2012 and March 2014 identified 397 patients with primary lung adenocarcinoma who fulfilled the following inclusion criteria: (1) pathology reports with diagnosis of lung adenocarcinoma; (2) preoperative thin-section CT images at in the Picture Archiving and Communication System (PACS), and the location of the lesion was peripheral (tumor involving subsegmental bronchus or smaller airway); (3) available test results for EGFR mutation status; and (4) available clinical data. Thereafter, 99 patients were excluded because of the following reasons: receiving preoperative treatment, such as radiotherapy or chemotherapy; the duration between CT examination and subsequence surgery exceeded 1 month; and patients with lung cancer for which it is difficult to contour the tumor margin on CT images.
      Clinical and pathological data collected for analysis included sex, age at diagnosis, smoking status, pathologic tumor, node, metastases stage, and histologic lung adenocarcinoma subtypes. Smoking status was categorized into 2 groups; never smokers and smokers, which included former and current smokers. Tumors were staged pathologically according to the seventh edition of the American Joint Committee on Cancer Staging Manual.
      • Edge S.B.
      • Compton C.C.
      The American Joint Committee on Cancer: the 7th edition of the AJCC cancer staging manual and the future of TNM.
      Tumors were diagnosed as adenocarcinoma and then categorized according to the 2011 International Association for the Study of Lung Cancer, American Thoracic Society, and European Respiratory Society classification system.
      • Travis W.D.
      • Brambilla E.
      • Noguchi M.
      • et al.
      International Association for the Study of Lung Cancer/American Thoracic Society/European Respiratory Society: international multidisciplinary classification of lung adenocarcinoma: executive summary.

      Computed Tomography Examination

      Chest CT examinations were conducted using Somatom Sensation 64 (Siemens Medical Solutions, Forchheim, Germany), Light speed 16 (GE Medical Systems, Milwaukee, WI), or Discovery CT750 HD scanner (GE Medical Systems). The parameters used were as follows: 120 kVp with tube current adjusted automatically, pitch was 0.969, reconstruction thickness was 1.5 mm, and reconstruction interval was 1.5 mm for the 64-detector scanner; tube voltage was 120 kVp, tube current was 150 to 200 mA, pitch was 0.969; reconstruction thickness was 1.25 mm, and reconstruction interval was 1.25 mm for the 16-detector scanner and Discovery CT750 HD scanner (GE Medical Systems).

      Detection of EGFR Mutations

      For the gene mutation analysis, tumor specimens were obtained using surgical resection. We performed EGFR mutation analyses of 4 tyrosine kinase domain (exons 18-21), which are frequently mutated in lung cancer, as previously described.
      • Locatelli-Sanchez M.
      • Couraud S.
      • Arpin D.
      • Riou R.
      • Bringuier P.P.
      • Souquet P.J.
      Routine EGFR molecular analysis in non–small-cell lung cancer patients is feasible: exons 18–21 sequencing results of 753 patients and subsequent clinical outcomes.
      EGFR mutations were determined using an amplification refractory mutation system real-time technology using Human EGFR Gene Mutations Detection Kit (Beijing ACCB Biotech Ltd).

      Tumor Segmentation

      We used Definiens Developer XD (Munich, Germany) as the image analysis platform to perform tumor segmentation and feature extraction. A lung tumor analysis tool within Definiens Cognition Network Technology was used. Lesions were volumetrically segmented using a semiautomatic approach by 2 radiologists with more than 6 and 3 years of experience in CT imaging of thoracic malignancies, respectively (Figure 1). The semiautomatic segmentation work flow which contained the following 4 steps named preprocessing, semiautomated correction of the pulmonary boundary, click and grow, and manual refinement and generation of lesion statistics were described in detail in previous studies.
      • Balagurunathan Y.
      • Kumar V.
      • Gu Y.
      • et al.
      Test-retest reproducibility analysis of lung CT image features.
      • Gu Y.
      • Kumar V.
      • Hall L.O.
      • et al.
      Automated delineation of lung tumors from CT images using a single-click ensemble segmentation approach.
      • Velazquez E.R.
      • Aerts H.J.
      • Gu Y.
      • et al.
      A semiautomatic CT-based ensemble segmentation of lung tumors: Comparison with oncologists' delineations and with the surgical specimen.
      Here, the single-click ensemble segmentation (SCES) algorithm,
      • Gu Y.
      • Kumar V.
      • Hall L.O.
      • et al.
      Automated delineation of lung tumors from CT images using a single-click ensemble segmentation approach.
      which is an advanced version of the previous click and grow algorithm and reduces sensitivity toward the location of the initial seed-point, was used. The SCES makes use of the original algorithm by choosing different seed points automatically within a specified area of the lesion and performing region-growing with each generated seed point. Each of the 2 radiologists reviewed the segmented images in consensus, and any discrepancies were resolved by discussion until consensus was reached.
      Figure thumbnail gr1
      Figure 1Representative Computed Tomography Image With Tumor Segmentation Using a Semiautomated Algorithm for Peripheral Adenocarcinoma. (A) A Lobulated Lung Tumor in the Right Middle Lobe Was Chosen for Segmentation. (B) One Radiologist Segmented the Boundary of the Tumor, Which Is Shown in Green Outline. (C) Three-Dimensional View of the Lung and Segmented Tumor

      Radiomic Feature Extraction

      We extracted a total of 219 features from each of the 3-D objects. These features were divided into 8 categories, including tumor size, shape, location, air space, pixel intensity histogram, run length and co-occurrence, Laws texture, and wavelets. A description of all features is provided (Supplemental Table 1 in the online version), and detailed description of texture features can be found in the previous study.
      • Balagurunathan Y.
      • Gu Y.
      • Wang H.
      • et al.
      Reproducibility and prognosis of quantitative features extracted from CT images.
      Our features cover a variety of descriptors from size, location, attachment of the lesion of interest, to CT pixel distribution and texture according to appearance on CT image. We characterized texture in observed CT images using the Laws feature descriptor and in the decomposed domain by using wavelets. Texture features are known to carry information that are not always typically observed by the human eye. Metrics were described by wavelet transformation, and Laws features are considered to describe subtle characteristics in the image and have been shown to be useful in image classifications.
      • Randen T.
      • Husøy J.H.
      Filtering for texture classification: a comparative study.

      Statistical Analysis

      All statistical analyses was performed using SAS software version 9.4 (SAS Institute Inc, Cary NC). Power transformation was considered to apply parametric analytical tools. The correlation between features was investigated to address collinearity issues. The highly correlated features (correlation > 0.9) were regarded as dependent features, which were not considered in this analysis. Thus, 59 of 219 features were considered as mutually independent features for prediction of EGFR mutation. Fisher exact test and the Kruskal–Wallis test were used for categorical and continuous variables between 2 groups, respectively. Multiple logistic regression analysis was performed. The final model was selected using the backward elimination method. Variables with a P value of < .25 in the univariate model were entered in the initial model, and then a variable with a P value of > .15 was eliminated at each step. The elimination procedure was terminated when the P value of all variables in the model was < .15. The accuracy and error of the predictive models (ie, the area under the curve [AUC] and 95% confidence interval [CI]) were estimated using the bootstrap method with 1000 bootstrap samples. Receiver operating characteristic (ROC) curves for each model were constructed and the AUC was calculated with EGFR mutation status determined using an amplification refractory mutation system-polymerase chain reaction (PCR) as outcome. Further, various predictive models were developed using the support vector machine and principal component analysis and were compared with the logistic regression model. In this study, the regression model was selected as the AUC of the model that was higher than any other models (results were omitted). A 2-sided P value of < .05 was regarded as statistically significant.

      Results

      The patient demographic and clinicopathological data are presented in Table 1. All 298 of the enrolled patients were surgically treated: lobectomy in 280 patients, pneumonectomy in 4, wedge resection in 6, and segmentectomy in 8 patients. Overall, there were 126 men and 172 women with a median age of 60 years (range, 30-80 years). The pathologic stage distribution was as follows: IA in 117 patients (39.26%), IB in 63 patients (21.14%), IIA in 22 patients (7.38%), IIB in 6 patients (2.01%), IIIA in 72 patients (24.16%), IIIB in 2 patients (0.67%), and IV in 16 patients (5.37%). Most of the tumors were early stage (stage I or II; 208 [69.80%] of 298). All cases were lung adenocarcinomas and the most common histologic subtype among invasive adenocarcinomas was acinar predominant subtype (128 [42.95%] of 298), followed by lepidic predominant subtype (71 [23.83%] of 298). EGFR mutation results (Supplemental Table 2 in the online version) were satisfactorily shown in all patents, with 137 patients (45.97%) of the full cohort of 298 cases identified as EGFR mutant and 161 (54.03%) as EGFR wild type.
      Table 1Clinicopathological Characteristics of Patients
      VariableOverall Data Set (n = 298)EGFR Mutant (n = 137)EGFR Wild Type (n = 161)P
      Median Age (Range)60 (30-80)59 (37-80)60 (30-79).51
      Sex.001
       Male12643 (31.39)83 (51.55)
       Female17294 (68.61)78 (48.45)
      Smoking Status<.0001
       Smoker13643 (31.39)93 (57.76)
       Never smoker16294 (68.61)68 (42.24)
      Pathologic Grade
      Adenocarcinoma in situ and MIA were classified as low grade; lepidic, acinar, and papillary as intermediate grade; micropapillary, solid, and invasive mucinous adenocarcinoma as high grade. One case with enteric was eliminated in this analysis.
      .0002
       High7420 (14.60)54 (33.75)
       Low/intermediate223117 (85.40)106 (66.25)
      Histologic Subtype
      Histologic subtype was categorized as lepidic predominant adenocarcinomas (adenocarcinoma in situ, MIA, and lepidic predominant invasive adenocarcinoma) and other subtypes of dominant histologic findings (acinar, papillary, micropapillary, and solid predominant as well as variants of invasive adenocarcinoma).
      .022
       Lepidic predominant adenocarcinomas7443 (31.39)31 (19.25)
       Other22494 (68.61)130 (80.75)
      Stage.53
       I or II20893 (67.88)115 (71.43)
       III or IV9044 (32.12)46 (28.57)
      Data are presented as n, or n (%), except where otherwise noted.
      Abbreviation: MIA = minimally invasive adenocarcinoma.
      a Adenocarcinoma in situ and MIA were classified as low grade; lepidic, acinar, and papillary as intermediate grade; micropapillary, solid, and invasive mucinous adenocarcinoma as high grade. One case with enteric was eliminated in this analysis.
      b Histologic subtype was categorized as lepidic predominant adenocarcinomas (adenocarcinoma in situ, MIA, and lepidic predominant invasive adenocarcinoma) and other subtypes of dominant histologic findings (acinar, papillary, micropapillary, and solid predominant as well as variants of invasive adenocarcinoma).
      There were significant differences in sex, smoking status, pathologic grade, and histologic subtype between the EGFR wild type group and the EGFR mutant group (Table 1). Concerning sex, there were significantly more female patients with mutant, compared with wild type EGFR in lung adenocarcinomas (odds ratio [OR], 2.33; 95% CI, 1.45-3.74; P = .001). Smokers with mutant EGFR were significant fewer than smokers with wild type (OR, 2.99; 95% CI, 1.85-4.82; P < .0001). EGFR mutations were also significantly more frequent in patients with low or intermediate pathologic grade (OR, 2.98; 95% CI, 1.67-5.30; P = .0002), and patients with lepidic-predominant adenocarcinomas (OR, 1.92; 95% CI, 1.13-3.27; P = .022). There were no differences in stage distribution or median age between EGFR mutant and wild type groups (P = .53; P = .51). Univariate analyses revealed that 4 clinical factors might be associated with EGFR mutation status in peripheral lung adenocarcinoma: sex, smoking status, histologic subtype, and pathologic grade were all significant predictors of harboring an EGFR mutation (Table 2).
      Table 2Univariate Analysis for Clinicopathological Factors That Predict Epidermal Growth Factor Receptor Mutation Status
      Clinical FeaturesOdds RatioP
      Point95% CI
      LowerUpper
      Stage
       I or IIReference
       III or IV1.180.721.94.51
      Sex
       FemaleReference
       Male0.430.270.69.0005
      Histologic Subtype
       OthersReference
       Lepidic predominant adenocarcinomas1.921.133.27.017
      Pathological Grade
       Low or intermediateReference
       High0.340.190.60.0002
      Smoking Status
       Never smokerReference
       Smoker0.340.210.54<.0001
      Age (per 5-Year Increase)0.970.851.11.68
      We then investigated the association of radiomic features with EGFR mutation status. Because this analysis produced far more features that were considered dependent; only a prioritized subset of features were selected for further analysis to avoid overfitting. Prioritization methods are described in Locatelli-Sanchez et al.
      • Locatelli-Sanchez M.
      • Couraud S.
      • Arpin D.
      • Riou R.
      • Bringuier P.P.
      • Souquet P.J.
      Routine EGFR molecular analysis in non–small-cell lung cancer patients is feasible: exons 18–21 sequencing results of 753 patients and subsequent clinical outcomes.
      Fifty-nine independent features were selected finally (Supplemental Table 3 in the online version). Among these 59 independent features, 11 were independent predictors of harboring the EGFR mutation (Table 3, Supplemental Table 4 in the online version) in univariate analysis, including 1 feature describing tumor shape (F26), 2 features describing tumor location (F17, F27), 1 feature describing air space (F19), 2 pixel intensity histogram-based features (F185, F186), 2 run length and co–occurrence-based features (F47, F51), 2 Laws texture features (F90, F111), and 1 wavelet texture feature (F190).
      Table 3Univariate Analysis for Radiomic Features That Predict Epidermal Growth Factor Receptor Mutation Status
      Radiomic FeaturesOdds RatioP
      Point95% CI
      LowerUpper
      F19
       10a_3D_Relative_Volume_AirSpaces1.161.041.3.007
      F186
       Histogram ENERGY Layer 10.490.290.82.007
      F90
       3-D Laws features L5 L5 L5 Layer 10.820.70.96.015
      F51
       AvgLRE0.110.020.72.021
      F47
       AvgCoocurrence-mean0.980.971.025
      F190
       3-D Wavelet decomposition. P2 L2 C9 Layer 11.261.031.55.025
      F185
       Histogram SD Layer 11.881.073.31.027
      F26
       Elliptic Fit0.110.020.79.028
      F17
       9f-3D-Min-Dist-COG-to-Border0.590.360.96.032
      F111
       3-D Laws features R5 E5 R5 Layer 10.710.510.99.041
      F27
       Main direction1.2111.45.048
      With multiple logistic regression analyses, clinical features of smoking status and pathologic grade proved to be independent predictors of EGFR mutation, and the AUC of ROC was 0.667 (95% CI, 0.604-0.721). The multiple logistic regression model produced from radiomic features alone showed moderate predictive power (AUC, 0.647; 95% CI, 0.576-0.701) for identifying EGFR mutant status. There was a significant difference between AUCs of the logistic regression model incorporating only clinical features, and that incorporating only radiomic features (P < .0001). When clinical and radiomic features were combined, the AUC was increased to 0.709 (95% CI, 0.654-0.766; Table 4). The model generated with combined clinical and radiomic features was superior to the model generated with clinical features alone (P < .0001) and the model created with radiomic features alone (P < .0001; Figure 2).
      Table 4Multiple Logistic Regression Analysis of Clinicopathological Parameters and Radiomic Features Predicting the Presence of Epidermal Growth Factor Receptor Mutation in Peripheral Lung Adenocarcinomas
      ModelFeaturePOdds RatioAUC
      Point95% CIPoint95% CI
      LowerUpperLowerUpper
      Clinical FeaturesPathologic grade (ref = lower or Intermediate)High.0020.390.220.700.6670.6040.721
      Smoking status (ref = never smoker)Yes<.00010.380.230.61
      Radiomic FeatureMain direction (F27).1151.170.961.420.6470.5760.701
      3-D Laws features L5 L5 S5 Layer 1 (F92).0100.520.310.85
      Histogram ENERGY layer 1 (F186).0250.500.270.92
      3-D wavelet decomposition. P2 L2 C9 Layer 1 (F190).0481.291.001.67
      3-D wavelet decomposition. P1 L2 C5 Layer 1 (F216).1371.160.961.40
      Clinical and Radiomic FeaturesPathologic grade (ref = lower or intermediate)High.0160.470.250.870.7090.6540.766
      Smoking status (ref = never smoker)Yes.0010.410.250.67
      Main direction (F27).201.140.931.40
      3-D Laws features L5 L5 S5 layer 1 (F92).0560.600.361.01
      Histogram ENERGY layer 1 (F186).110.600.321.13
      3-D Wavelet decomposition. P2 L2 C9 Layer 1 (F190).131.220.941.59
      3-D wavelet decomposition. P1 L2 C5 Layer 1 (F216).241.130.921.38
      Abbreviation: AUC = area under the curve.
      Figure thumbnail gr2
      Figure 2Receiver Operating Characteristic Curves for the Prediction of Epidermal Growth Factor Receptor Mutation Using a Logistic Regression Model That Included Clinical Factors Alone (Green Line), a Model That Used Radiomic Features Alone (Red Line), and a Model That Combined Clinical Factors and Radiomic Features (Blue Line). The Highest Area Under the Curve (AUC) Was Achieved for the Combination of Clinical Factors and Radiomic Features (AUC = 0.709)

      Discussion

      There are various methods to detect EGFR mutations, such as direct sequencing of PCR-amplified genomic DNA, high-resolution melting analysis, fragment analysis, restriction fragment length polymorphism, and the amplification refractory mutation system
      • da Cunha Santos G.
      • Shepherd F.A.
      • Tsao M.S.
      EGFR mutations and lung cancer.
      ; however, these molecular methods are generally costly, and sometimes in a low percentage of tumor cells, we would have most likely missed the mutations and rebiopsy had to be recommended. In this study, we sought to apply radiomic features to peripheral lung adenocarcinomas to determine if we could noninvasively discriminate EGFR-mutant from EGFR-wild type cases in a routine practice without adding additional cost.
      We found that 137 of the tumors harbored the EGFR mutation, which corresponded to 45.97% of the 298 tumor samples, this is in keeping with previous reports on Asian patients
      • Liu W.S.
      • Zhao L.J.
      • Pang Q.S.
      • Yuan Z.Y.
      • Li B.
      • Wang P.
      Prognostic value of epidermal growth factor receptor mutations in resected lung adenocarcinomas.
      ; in addition, exon 19 and exon 21, the most common mutation types of the EGFR gene, showed nearly the same percentage in these patients (46.72% (64/137), 48.91% (67/137), respectively). Dual mutations were detected in 1 patient (0.73%) in this study, confirming that the existence of EGFR TKI-resistant (exon 20) and -sensitive mutation (exon 21) is rare. Several studies indicated that the EGFR mutation was strongly related with never smokers, female sex, adenocarcinoma, and pathologic stage.
      • Locatelli-Sanchez M.
      • Couraud S.
      • Arpin D.
      • Riou R.
      • Bringuier P.P.
      • Souquet P.J.
      Routine EGFR molecular analysis in non–small-cell lung cancer patients is feasible: exons 18–21 sequencing results of 753 patients and subsequent clinical outcomes.
      • Liu W.S.
      • Zhao L.J.
      • Pang Q.S.
      • Yuan Z.Y.
      • Li B.
      • Wang P.
      Prognostic value of epidermal growth factor receptor mutations in resected lung adenocarcinomas.
      Consistent with most of these findings, the rate of EGFR mutation in female and never-smoker patients with lung adenocarcinomas was considerably higher compared with male and smoker patients in this study. Although the EGFR mutation was detected more frequently in early-stage patients (93 [67.88%] of 298) compared with advanced stage (44 [32.12%] of 137), the difference was not significant. The results of the multiple logistic regression analysis revealed that pathologic grade and smoking status were independent predictors for EGFR mutation and the AUC was 0.667. Besides clinical information, more detailed factors will likely be required to identify those at high probability of harboring EGFR mutations.
      Computed tomography imaging is used widely in oncologic practice for lung tumor characterization. Usually, we interpret the image on the basis of visual assessment; there are features, however, within each image that might not be perceived by the naked eye and require computer-aided techniques. Previous authors have shown the potential of quantitative CT-based texture analysis in differentiation of K-ras mutation from pan-wild type NSCLC.
      • Weiss G.J.
      • Ganeshan B.
      • Miles K.A.
      • et al.
      Noninvasive image texture analysis differentiates K-ras mutation from pan-wildtype NSCLC and is prognostic.
      To the best of our knowledge, only 1 study
      • Ozkan E.
      • West A.
      • Dedelow J.A.
      • et al.
      CT gray-level texture analysis as a quantitative imaging biomarker of epidermal growth factor receptor mutation status in adenocarcinoma of the lung.
      that was published recently has explored the association between CT gray-level texture features and EGFR mutations in a relatively small sample size (25 patients with EGFR mutation and 20 patients with EGFR wild type). In this study, we present comprehensive radiomic analysis using semiautomatic segmentation in 298 peripheral lung adenocarcinomas. Two hundred nineteen radiomic features were extracted to assess the ability to predict EGFR mutation status. Considering that using too many features in the classification algorithm can lead to overfitting, in which noise or irrelevant features might exert undue influence on classification decisions, only features that are not associated with other features were selected for further analysis. With this approach, we found that 11 radiomic features from 7 different feature categories were significantly associated with EGFR mutations. Texture features, wavelet features, Laws features, and along with pixel statistical, have seen a resurgent use in medical images, especially CT and magnetic resonance images.
      • Weiss G.J.
      • Ganeshan B.
      • Miles K.A.
      • et al.
      Noninvasive image texture analysis differentiates K-ras mutation from pan-wildtype NSCLC and is prognostic.
      • Zhao B.
      • James L.P.
      • Moskowitz C.S.
      • et al.
      Evaluating variability in tumor measurements from same-day repeat CT scans of patients with non–small-cell lung cancer.
      In our study we formed sets of 5 one-dimensional Laws filters, each designed to describe different structures in the image (E: edges, S: spots, R: ripple, W: waves, L: low pass). The wavelet transformation was limited to 2 levels (L) of decomposition on each of the 9 faces (C) on the 3-D tumor, with 2 types of metrics (P), namely, energy and entropy. Our statistical model finds Laws features L5L5S5 and wavelet P2L2C9 and P1L2C5 to be 1 of the 5 predictors of EGFR mutation status. Several previous studies
      • Rizzo S.
      • Petrella F.
      • Buscarino V.
      • et al.
      CT radiogenomic characterization of EGFR, K-RAS, and ALK mutations in non–small -cell lung cancer.
      • Hsu J.S.
      • Huang M.S.
      • Chen C.Y.
      • et al.
      Correlation between EGFR mutation status and computed tomography features in patients with advanced pulmonary adenocarcinoma.
      showed that EGFR mutation was associated with small tumor size. In our analysis, size-based features including longest diameter and short axis were not significant predictors for the EGFR mutation. We believe that this difference can be explained in part by the fact that most clinically assessed tumor diameters are manually drawn on the central slice of CT images and limited to 1 dimension of the tumor, and whole tumor volume was taken into account in the size-based radiomic features.
      With logistic regression analysis, we identified that radiomic features could be served as an imaging surrogate for EGFR mutation, although the AUC of radiomic features alone was not as high as the model created with clinical features, they provide complementary information, as the combination results in an improved ROC curve (AUC, 0.709), and we were able to significantly increase the predictive performance of EGFR mutation. These findings suggested that the combination of radiomic data and demographic information in a system model is more effective.
      There are some limitations in this study. One limitation is that this study is retrospective and limited to only Eastern Asian populations; care should be taken before generalizing our findings to other populations. Second, radiomic features were derived from semi-automatic segmentation by radiologists, which can be influenced by observers' subjective trend. However, the results of automatic boundary extraction method were not satisfactory for all the lesions, particularly in the case of tumors with GGO components, as their margins are usually unclear from the adjacent normal lung parenchyma. Furthermore, since atelectasis is common in patients with central lung cancers, and differentiation between them is rather difficult as both appear as solid density on CT, our analysis did not include central lung cancers. A prospective multi-Institutional study with a large patient cohort would be required to confirm our observations.

      Conclusion

      In summary, this study revealed associations between CT based radiomic features and EGFR mutation status in peripheral lung adenocarcinomas, and therefore non-invasive radiomic phenotype analysis has the potential to improve the differentiation of EGFR mutate from wild type when used in addition to clinical predictors.

      Clinical Practice Points

      • The presence of activating EGFR mutations has been shown to be prognostic for a more favorable outcome to TKI therapy in lung adenocarcinomas.
      • Several reports have described the relationship between mutation status of EGFR and traditional radiological features. Radiomic-based approach allows high throughput extraction of quantitative parameters from CT images which beyond what is visually perceived by the human eyes. Thus, we hypothesized that radiomic analysis of routinely performed preoperative CT images could provide imaging biomarkers for EGFR mutations.
      • In this study we demonstrated that eleven CT based radiomic features have significant association with EGFR mutations. Adding radiomic features to clinical model could improve the predicting power of EGFR mutations, thus helping in formulating a better clinical decision without adding additional cost.

      Disclosure

      R.J.G. is a consultant and shareholder in HealthMyne, Inc, and oncology-specific PACS system. The remaining authors have stated that they have no conflicts of interest.

      Acknowledgments

      We thank Zhongli Zhan (Department of Pathology, Tianjin Medical University Cancer Institute and Hospital) for doing histologic evaluation and EGFR mutation analysis.
      This work was supported by the National Cancer Institute (grants U01 CA143062), Tianjin Science and Technology Major Project (No. 12ZCDZSY15500), and Public science and technology research funds projects of NHFPC of the P.R. China (No. 201402013). This work has been supported in part by the Biostatistics Core Facility at the Moffitt Cancer Center & Research Institute, a National Cancer Institute designated Comprehensive Cancer Center (5P30CA076292-16).

      Supplemental Data

      Supplemental Table 1Quantitative Radiomic Features
      Feature CategoryFeature IndexDescription of the Features
      C1. Tumor SizeF1Longest diameter, mm
      F 2ShortAx-LongDia
      F 3Short axis, mm
      F 6Vol-cm3
      F 33Area-Pxl
      F 34Volume-Pxl
      F 35Num-Pxl
      F 36Width-Pxl
      F 37Thickness-Pxl
      F 38Length-Pxl
      F 39Length-by-Thick
      F 40Length-by-Width
      F 41Border-Leng-Pxl
      C2. Tumor Shape (Roundness)F 75a-3D-MacSpic
      F 139b-3D-Circularity
      F 149c-3D-Compactness
      F 23Asymmetry
      F 24Compactness
      F 26Elliptic Fit
      F 28Radius of largest enclosed ellipse
      F 29Radius of smallest enclosed ellipse
      F 30Shape-Index
      F 31Roundness
      F 32Rectangular fit
      C3. Tumor LocationF 88a-3D-Attch-Pleural wall
      F 98b-3D-Border-to-Lung
      F 108c-3D-Border-to-Pleural wall
      F 118d-3D-Ratio-Free-to-Attach
      F 129a-3D-FractionalAnisotropy
      F 159d-3D-AV-Dist-COG-to-Border
      F 169e-3D-SD-Dist-COG-to-Border
      F 179f-3D-Min-Dist-COG-to-Border
      F 189g-3D-Max-Dist-COG-to-Border
      F 27Main Direction
      C4. AirspaceF 1910a_3D_Relative_Volume_AirSpaces
      F 2010b_3D_Number_AirSpaces
      F 2110c_3D_Av_Volume_AirSpaces
      F 2210d_3D_SD_Volume_AirSpaces
      C5. Pixel Intensity HistogramF 4Mean, Hounsfield Units
      F 5StdDev [HU]
      F 184Histogram-Mean-Layer 1
      F 185Histogram-SD-Layer 1
      F 186Histogram-Energy-Layer 1
      F 187Histogram-Entropy-Layer 1
      F 188Histogram-Kurt-Layer 1
      F 189Histogram-Skew-Layer 1
      F 25Density
      C6. Run Length and Co-occurrenceF 42AvgCoocurrence-Homo
      F 43AvgCoocurrence-Mp
      F 44AvgCoocurrence-Constrast
      F 45AvgCoocurrence-Energy
      F 46AvgCoocurrence-Entropy
      F 47AvgCoocurrence-Mean
      F 48AvgGLN
      F 49AvgHGRE
      F 50AvgLGRE
      F 51AvgLRE
      F 52AvgLRHGE
      F 53AvgLRLGE
      F 54AvgRLN
      F 55AvgRP
      F 56AvgSRE
      F 57AvgSRHGE
      F 58AvgSRLGE
      C7. Laws Texture Feature (With Different Convolution Filters)F 593-D Laws features E5 E5 E5 Layer 1
      F 603-D Laws features E5 E5 L5 Layer 1
      F 613-D Laws features E5 E5 R5 Layer 1
      F 623-D Laws features E5 E5 S5 Layer 1
      F 633-D Laws features E5 E5 W5 Layer 1
      F 643-D Laws features E5 L5 E5 Layer 1
      F 653-D Laws features E5 L5 L5 Layer 1
      F 663-D Laws features E5 L5 R5 Layer 1
      F 673-D Laws features E5 L5 S5 Layer 1
      F 683-D Laws features E5 L5 W5 Layer 1
      F 693-D Laws features E5 R5 E5 Layer 1
      F 703-D Laws features E5 R5 L5 Layer 1
      F 713-D Laws features E5 R5 R5 Layer 1
      F 723-D Laws features E5 R5 S5 Layer 1
      F 733-D Laws features E5 R5 W5 Layer 1
      F 743-D Laws features E5 S5 E5 Layer 1
      F 753-D Laws features E5 S5 L5 Layer 1
      F 763-D Laws features E5 S5 R5 Layer 1
      F 773-D Laws features E5 S5 S5 Layer 1
      F 783-D Laws features E5 S5 W5 Layer 1
      F 793-D Laws features E5 W5 E5 Layer 1
      F 803-D Laws features E5 W5 L5 Layer 1
      F 813-D Laws features E5 W5 R5 Layer 1
      F 823-D Laws features E5 W5 S5 Layer 1
      F 833-D Laws features E5 W5 W5 Layer 1
      F 843-D Laws features L5 E5 E5 Layer 1
      F 853-D Laws features L5 E5 L5 Layer 1
      F 863-D Laws features L5 E5 R5 Layer 1
      F 873-D Laws features L5 E5 S5 Layer 1
      F 883-D Laws features L5 E5 W5 Layer 1
      F 893-D Laws features L5 L5 E5 Layer 1
      F 903-D Laws features L5 L5 L5 Layer 1
      F 913-D Laws features L5 L5 R5 Layer 1
      F 923-D Laws features L5 L5 S5 Layer 1
      F 933-D Laws features L5 L5 W5 Layer 1
      F 943-D Laws features L5 R5 E5 Layer 1
      F 953-D Laws features L5 R5 L5 Layer 1
      F 963-D Laws features L5 R5 R5 Layer 1
      F 973-D Laws features L5 R5 S5 Layer 1
      F 983-D Laws features L5 R5 W5 Layer 1
      F 993-D Laws features L5 S5 E5 Layer 1
      F 1003-D Laws features L5 S5 L5 Layer 1
      F 1013-D Laws features L5 S5 R5 Layer 1
      F 1023-D Laws features L5 S5 S5 Layer 1
      F 1033-D Laws features L5 S5 W5 Layer 1
      F 1043-D Laws features L5 W5 E5 Layer 1
      F 1053-D Laws features L5 W5 L5 Layer 1
      F 1063-D Laws features L5 W5 R5 Layer 1
      F 1073-D Laws features L5 W5 S5 Layer 1
      F 1083-D Laws features L5 W5 W5 Layer 1
      F 1093-D Laws features R5 E5 E5 Layer 1
      F 1103-D Laws features R5 E5 L5 Layer 1
      F 1113-D Laws features R5 E5 R5 Layer 1
      F 1123-D Laws features R5 E5 S5 Layer 1
      F 1133-D Laws features R5 E5 W5 Layer 1
      F 1143-D Laws features R5 L5 E5 Layer 1
      F 1153-D Laws features R5 L5 L5 Layer 1
      F 1163-D Laws features R5 L5 R5 Layer 1
      F 1173-D Laws features R5 L5 S5 Layer 1
      F 1183-D Laws features R5 L5 W5 Layer 1
      F 1193-D Laws features R5 R5 E5 Layer 1
      F 1203-D Laws features R5 R5 L5 Layer 1
      F 1213-D Laws features R5 R5 R5 Layer 1
      F 1223-D Laws features R5 R5 S5 Layer 1
      F 1233-D Laws features R5 R5 W5 Layer 1
      F 1243-D Laws features R5 S5 E5 Layer 1
      F 1253-D Laws features R5 S5 L5 Layer 1
      F 1263-D Laws features R5 S5 R5 Layer 1
      F 1273-D Laws features R5 S5 S5 Layer 1
      F 1283-D Laws features R5 S5 W5 Layer 1
      F 1293-D Laws features R5 W5 E5 Layer 1
      F 1303-D Laws features R5 W5 L5 Layer 1
      F 1313-D Laws features R5 W5 R5 Layer 1
      F 1323-D Laws features R5 W5 S5 Layer 1
      F 1333-D Laws features R5 W5 W5 Layer 1
      F 1343-D Laws features S5 E5 E5 Layer 1
      F 1353-D Laws features S5 E5 L5 Layer 1
      F 1363-D Laws features S5 E5 R5 Layer 1
      F 1373-D Laws features S5 E5 S5 Layer 1
      F 1383-D Laws features S5 E5 W5 Layer 1
      F 1393-D Laws features S5 L5 E5 Layer 1
      F 1403-D Laws features S5 L5 L5 Layer 1
      F 1413-D Laws features S5 L5 R5 Layer 1
      F 1423-D Laws features S5 L5 S5 Layer 1
      F 1433-D Laws features S5 L5 W5 Layer 1
      F 1443-D Laws features S5 R5 E5 Layer 1
      F 1453-D Laws features S5 R5 L5 Layer 1
      F 1463-D Laws features S5 R5 R5 Layer 1
      F 1473-D Laws features S5 R5 S5 Layer 1
      F 1483-D Laws features S5 R5 W5 Layer 1
      F 1493-D Laws features S5 S5 E5 Layer 1
      F 1503-D Laws features S5 S5 L5 Layer 1
      F 1513-D Laws features S5 S5 R5 Layer 1
      F 1523-D Laws features S5 S5 S5 Layer 1
      F 1533-D Laws features S5 S5 W5 Layer 1
      F 1543-D Laws features S5 W5 E5 Layer 1
      F 1553-D Laws features S5 W5 L5 Layer 1
      F 1563-D Laws features S5 W5 R5 Layer 1
      F 1573-D Laws features S5 W5 S5 Layer 1
      F 1583-D Laws features S5 W5 W5 Layer 1
      F 1593-D Laws features W5 E5 E5 Layer 1
      F 1603-D Laws features W5 E5 L5 Layer 1
      F 1613-D Laws features W5 E5 R5 Layer 1
      F 1623-D Laws features W5 E5 S5 Layer 1
      F 1633-D Laws features W5 E5 W5 Layer 1
      F 1643-D Laws features W5 L5 E5 Layer 1
      F 1653-D Laws features W5 L5 L5 Layer 1
      F 1663-D Laws features W5 L5 R5 Layer 1
      F 1673-D Laws features W5 L5 S5 Layer 1
      F 1683-D Laws features W5 L5 W5 Layer 1
      F 1693-D Laws features W5 R5 E5 Layer 1
      F 1703-D Laws features W5 R5 L5 Layer 1
      F 1713-D Laws features W5 R5 R5 Layer 1
      F 1723-D Laws features W5 R5 S5 Layer 1
      F 1733-D Laws features W5 S5 E5 Layer 1
      F 1743-D Laws features W5 S5 L5 Layer 1
      F 1753-D Laws features W5 R5 W5 Layer 1
      F 1763-D Laws features W5 S5 R5 Layer 1
      F 1773-D Laws features W5 S5 S5 Layer 1
      F 1783-D Laws features W5 S5 W5 Layer 1
      F 1793-D Laws features W5 W5 E5 Layer 1
      F 1803-D Laws features W5 W5 L5 Layer 1
      F 1813-D Laws features W5 W5 R5 Layer 1
      F 1823-D Laws features W5 W5 S5 Layer 1
      F 1833-D Laws features W5 W5 W5 Layer 1
      C8. Wavelets Texture (Feature At Different Layers)F 1903-D Wavelet decomposition. P2 L2 C9 Layer 1
      F 1913-D Wavelet decomposition. P1 L2 C9 Layer 1
      F 1923-D Wavelet decomposition. P2 L2 C10 Layer 1
      F 1933-D Wavelet decomposition. P2 L2 C11 Layer 1
      F 1943-D Wavelet decomposition. P2 L2 C12 Layer 1
      F 1953-D Wavelet decomposition. P2 L2 C13 Layer 1
      F 1963-D Wavelet decomposition. P2 L2 C14 Layer 1
      F 1973-D Wavelet decomposition. P2 L2 C15 Layer 1
      F 1983-D Wavelet decomposition. P2 L2 C1 Layer 1
      F 1993-D Wavelet decomposition. P2 L2 C2 Layer 1
      F 2003-D Wavelet decomposition. P2 L2 C3 Layer 1
      F 2013-D Wavelet decomposition. P2 L2 C4 Layer 1
      F 2023-D Wavelet decomposition. P2 L2 C5 Layer 1
      F 2033-D Wavelet decomposition. P2 L2 C6 Layer 1
      F 2043-D Wavelet decomposition. P2 L2 C7 Layer 1
      F 2053-D Wavelet decomposition. P2 L2 C8 Layer 1
      F 2063-D Wavelet decomposition. P1 L2 C11 Layer 1
      F 2073-D Wavelet decomposition. P1 L2 C10 Layer 1
      F 2083-D Wavelet decomposition. P1 L2 C12 Layer 1
      F 2093-D Wavelet decomposition. P1 L2 C13 Layer 1
      F 2103-D Wavelet decomposition. P1 L2 C14 Layer 1
      F 2113-D Wavelet decomposition. P1 L2 C15 Layer 1
      F 2123-D Wavelet decomposition. P1 L2 C1 Layer 1
      F 2133-D Wavelet decomposition. P1 L2 C2 Layer 1
      F 2143-D Wavelet decomposition. P1 L2 C3 Layer 1
      F 2153-D Wavelet decomposition. P1 L2 C4 Layer 1
      F 2163-D Wavelet decomposition. P1 L2 C5 Layer 1
      F 2173-D Wavelet decomposition. P1 L2 C6 Layer 1
      F 2183-D Wavelet decomposition. P1 L2 C7 Layer 1
      F 2193-D Wavelet decomposition. P1 L2 C8 Layer 1
      Supplemental Table 2Spectrum of Epidermal Growth Factor Receptor Mutations
      Mutation TypeNumber of Patients (n = 137)Percentage
      Single Mutation
       E1821.46
       E196446.72
       E2032.19
       E216748.91
      Combined Mutation
       E20, E2110.73
      Supplemental Table 3Independent Radiomic Features Included in the Analysis
      Independent Imaging Features
      F1F3F4F5F7F8F9F11F12F13
      F16F17F19F20F21F22F25F26F27F28
      F32F40F42F44F47F49F50F51F61F64
      F65F66F70F88F90F92F97F100F101F111
      F113F117F140F145F146F174F185F186F189F190
      F191F206F208F209F212F216F217F218F219
      Supplemental Table 4Univariate Analysis for Radiomic Features That Predict Epidermal Growth Factor Receptor Mutation Status
      Radiomic FeatureOdds RatioP
      Point95% CI
      LowerUpper
      F3
       Short axis, mm0.560.321.051
      F25
       Density0.110.011.03.053
      F1
       Longest Diameter, mm0.570.321.03.063
      F5
       StdDev [HU]1.570.972.54.064
      F145
       3-D Laws features S5 R5 L5 Layer 10.660.431.03.067
      F9
       8b-3-D-Border-to-Lung1.530.952.46.077
      F146
       3-D Laws features S5 R5 R5 Layer 10.670.441.04.077
      F70
       3-D Laws features E5 R5 L5 Layer 10.740.531.04.082
      F49
       AvgHGRE0.60.341.07.084
      F66
       3-D Laws features E5 L5 R5 Layer 10.740.531.04.086
      F4
       Mean, Hounsfield unit0.880.751.02.091
      F12
       9a-3-D-FractionalAnisotropy1.480.892.46.132
      F13
       9b_3-D_Circularity1.860.814.28.143
      F92
       3-D Laws features L5 L5 S5 Layer 10.730.481.12.150
      F189
       Histogram SKEW Layer 11.160.941.43.154
      F218
       3-D wavelet decomposition. P1 L2 C7 Layer 11.110.961.29.156
      F219
       3-D wavelet decomposition. P1 L2 C8 Layer 11.10.961.26.168
      F117
       3-D Laws features R5 L5 S5 Layer 10.750.491.14.176
      F65
       3-D Laws features E5 L5 L5 Layer 10.80.571.12.186
      F50
       AvgLGRE0.730.451.18.199
      F101
       3-D Laws features L5 S5 R5 Layer 10.770.511.16.206
      F8
       8a-3-D-Attch-Pleural wall0.750.481.19.219
      F216
       3-D wavelet decomposition. P1 L2 C5 Layer 11.120.931.35.219
      F22
       10d_3-D_SD_Volume_AirSpaces1.060.961.18.252
      F217
       3-D wavelet decomposition. P1 L2 C6 Layer 11.130.911.39.268
      F206
       3-D wavelet decomposition. P1 L2 C11 Layer 11.180.861.62.294
      F97
       3-D Laws features L5 R5 S5 Layer 10.790.511.24.308
      F209
       3-D wavelet decomposition. P1 L2 C13 Layer 11.150.881.51.318
      F40
       Length-by-Width0.540.161.86.330
      F21
       10c_3-D_Av_Volume_AirSpaces1.060.951.18.333
      F191
       3-D wavelet decomposition. P1 L2 C9 Layer 11.150.851.55.363
      F28
       Radius of largest enclosed ellipse1.180.791.77.412
      F140
       3-D Laws features S5 L5 L5 Layer 10.850.561.27.419
      F20
       10b_3-D_Number_AirSpaces1.080.91.3.424
      F113
       3-D Laws features R5 E5 W5 Layer 10.850.551.31.460
      F212
       3-D wavelet decomposition. P1 L2 C1 Layer 10.920.711.2.525
      F100
       3-D Laws features L5 S5 L5 Layer 10.90.591.35.605
      F44
       AvgCoocurrence-contrast0.90.561.44.650
      F32
       Rectangular Fit0.590.066.07.655
      F11
       8d-3-D-Ratio-Free-to-Attach1.020.921.14.677
      F42
       AvgCoocurrence-Homo1.470.1712.97.727
      F61
       3-D Laws features E5 E5 R5 Layer 10.940.61.45.763
      F16
       9e-3-D-SD-Dist-COG-to-Border0.940.551.58.803
      F7
       5a-3-D-MacSpic1.180.294.81.817
      F64
       3-D Laws features E5 L5 E5 Layer 10.950.621.48.834
      F208
       3-D wavelet decomposition. P1 L2 C12 Layer 11.030.791.33.848
      F174
       3-D Laws features W5 S5 L5 Layer 11.040.681.6.863
      F88
       3-D Laws features L5 E5 W5 Layer 10.990.651.52.976

      References

        • Halpenny D.F.
        • Riely G.J.
        • Hayes S.
        • et al.
        Are there imaging characteristics associated with lung adenocarcinomas harboring ALK rearrangements?.
        Lung Cancer. 2014; 86: 190-194
        • Ganeshan B.
        • Panayiotou E.
        • Burnand K.
        • Dizdarevic S.
        • Miles K.
        Tumour heterogeneity in non–small-cell lung carcinoma assessed by CT texture analysis: a potential marker of survival.
        Eur Radiol. 2012; 22: 796-802
        • Mok T.S.
        • Wu Y.L.
        • Thongprasert S.
        • et al.
        Gefitinib or carboplatin–paclitaxel in pulmonary adenocarcinoma.
        N Engl J Med. 2009; 361: 947-957
        • Jackman D.M.
        • Miller V.A.
        • Cioffredi L.A.
        • et al.
        Impact of epidermal growth factor receptor and KRAS mutations on clinical outcomes in previously untreated non–small-cell lung cancer patients: results of an online tumor registry of clinical trials.
        Clin Cancer Res. 2009; 15: 5267-5273
        • Riely G.J.
        • Pao W.
        • Pham D.
        • et al.
        Clinical course of patients with non–small-cell lung cancer and epidermal growth factor receptor exon 19 and exon 21 mutations treated with gefitinib or erlotinib.
        Clin Cancer Res. 2006; 12: 839-844
        • Yang J.C.
        • Schuler M.H.
        • Yamamoto N.
        • et al.
        LUX-Lung 3: a randomized, open-label, phase III study of afatinib versus pemetrexed and cisplatin as first-line treatment for patients with advanced adenocarcinoma of the lung harboring EGFR-activating mutations.
        ASCO Annual Meeting. 2012; (Abstract No: LBA7500)
        • Maemondo M.
        • Inoue A.
        • Kobayashi K.
        • et al.
        Gefitinib or chemotherapy for non–small-cell lung cancer with mutated EGFR.
        N Engl J Med. 2010; 362: 2380-2388
        • Kobayashi K.
        • Inoue A.
        • Maemondo M.
        • et al.
        First-line gefitinib versus first-line chemotherapy by carboplatin (CBDCA) plus paclitaxel (TXL) in non–small-cell lung cancer (NSCLC) patients (pts) with EGFR mutations: a phase III study (002) by North East Japan Gefitinib Study Group.
        ASCO Annual Meeting. 2009; (Abstract No: 8016)
        • Mitsudomi T.
        • Morita S.
        • Yatabe Y.
        • et al.
        Gefitinib versus cisplatin plus docetaxel in patients with non–small-cell lung cancer harbouring mutations of the epidermal growth factor receptor (WJTOG3405): an open label, randomised phase 3 trial.
        Lancet Oncol. 2010; 11: 121-128
        • Paez J.G.
        • Jänne P.A.
        • Lee J.C.
        • et al.
        EGFR mutations in lung cancer: correlation with clinical response to gefitinib therapy.
        Science. 2004; 304: 1497-1500
        • Kosaka T.
        • Yatabe Y.
        • Endoh H.
        • Kuwano H.
        • Takahashi T.
        • Mitsudomi T.
        Mutations of the epidermal growth factor receptor gene in lung cancer biological and clinical implications.
        Cancer Res. 2004; 64: 8919-8923
        • Shahi R.B.
        • De Brakeleer S.
        • De Greve J.
        • Geers C.
        • In't Veld P.
        • Teugels E.
        Detection of EGFR-TK domain-activating mutations in NSCLC with generic PCR-based methods.
        Appl Immunohistochem Mol Morphol. 2015; 23: 163-171
        • Tomonaga N.
        • Nakamura Y.
        • Yamaguchi H.
        • et al.
        Analysis of intratumor heterogeneity of EGFR mutations in mixed type lung adenocarcinoma.
        Clin Lung Cancer. 2013; 14: 521-526
        • Bai H.
        • Wang Z.
        • Wang Y.
        • et al.
        Detection and clinical significance of intratumoral EGFR mutational heterogeneity in Chinese patients with advanced non–small-cell lung cancer.
        PLoS One. 2013; 8: e54170
        • Taniguchi K.
        • Okami J.
        • Kodama K.
        • Higashiyama M.
        • Kato K.
        Intratumor heterogeneity of epidermal growth factor receptor mutations in lung cancer and its correlation to the response to gefitinib.
        Cancer Sci. 2008; 99: 929-935
        • Querings S.
        • Altmüller J.
        • Ansén S.
        • et al.
        Benchmarking of mutation diagnostics in clinical lung cancer specimens.
        PLoS One. 2011; 6: e19601
        • Rizzo S.
        • Petrella F.
        • Buscarino V.
        • et al.
        CT radiogenomic characterization of EGFR, K-RAS, and ALK mutations in non–small -cell lung cancer.
        Eur Radiol. 2016; 26: 32-42
        • Zhou J.Y.
        • Zheng J.
        • Yu Z.F.
        • et al.
        Comparative analysis of clinicoradiologic characteristics of lung adenocarcinomas with ALK rearrangements or EGFR mutations.
        Eur Radiol. 2015; 25: 1257-1266
        • Lee Y.
        • Lee H.J.
        • Kim Y.T.
        • et al.
        Imaging characteristics of stage I non–small-cell lung cancer on CT and FDG-PET: relationship with epidermal growth factor receptor protein expression status and survival.
        Korean J Radiol. 2013; 14: 375-383
        • Lee H.J.
        • Kim Y.T.
        • Kang C.H.
        • et al.
        Epidermal growth factor receptor mutation in lung adenocarcinomas: relationship with CT characteristics and histologic subtypes.
        Radiology. 2013; 268: 254-264
        • Kumar V.
        • Gu Y.
        • Basu S.
        • et al.
        Radiomics: the process and the challenges.
        Magn Reson Imaging. 2012; 30: 1234-1248
        • Parmar C.
        • Velazquez E.R.
        • Leijenaar R.
        • et al.
        Robust radiomics feature quantification using semiautomatic volumetric segmentation.
        PLoS One. 2014; 9: e102107
        • Mattonen S.A.
        • Palma D.A.
        • Haasbeek C.J.
        • Senan S.
        • Ward A.D.
        Early prediction of tumor recurrence based on CT texture changes after stereotactic ablative radiotherapy (SABR) for lung cancer.
        Med Phys. 2014; 41: 033502
        • Weiss G.J.
        • Ganeshan B.
        • Miles K.A.
        • et al.
        Noninvasive image texture analysis differentiates K-ras mutation from pan-wildtype NSCLC and is prognostic.
        PLoS One. 2014; 9: e100244
        • Ravanelli M.
        • Farina D.
        • Morassi M.
        • et al.
        Texture analysis of advanced non–small-cell lung cancer (NSCLC) on contrast-enhanced computed tomography: prediction of the response to the first-line chemotherapy.
        Eur Radiol. 2013; 23: 3450-3455
        • Coroller T.P.
        • Grossmann P.
        • Hou Y.
        • et al.
        CT-based radiomic signature predicts distant metastasis in lung adenocarcinoma.
        Radiother Oncol. 2015; 114: 345-350
        • Edge S.B.
        • Compton C.C.
        The American Joint Committee on Cancer: the 7th edition of the AJCC cancer staging manual and the future of TNM.
        Ann Surg Oncol. 2010; 17: 1471-1474
        • Travis W.D.
        • Brambilla E.
        • Noguchi M.
        • et al.
        International Association for the Study of Lung Cancer/American Thoracic Society/European Respiratory Society: international multidisciplinary classification of lung adenocarcinoma: executive summary.
        Proc Am Thorac Soc. 2011; 8: 381-385
        • Locatelli-Sanchez M.
        • Couraud S.
        • Arpin D.
        • Riou R.
        • Bringuier P.P.
        • Souquet P.J.
        Routine EGFR molecular analysis in non–small-cell lung cancer patients is feasible: exons 18–21 sequencing results of 753 patients and subsequent clinical outcomes.
        Lung. 2013; 191: 491-499
        • Balagurunathan Y.
        • Kumar V.
        • Gu Y.
        • et al.
        Test-retest reproducibility analysis of lung CT image features.
        J Digit Imaging. 2014; 27: 805-823
        • Gu Y.
        • Kumar V.
        • Hall L.O.
        • et al.
        Automated delineation of lung tumors from CT images using a single-click ensemble segmentation approach.
        Pattern Recognit. 2013; 46: 692-702
        • Velazquez E.R.
        • Aerts H.J.
        • Gu Y.
        • et al.
        A semiautomatic CT-based ensemble segmentation of lung tumors: Comparison with oncologists' delineations and with the surgical specimen.
        Radiother Oncol. 2012; 105: 167-173
        • Balagurunathan Y.
        • Gu Y.
        • Wang H.
        • et al.
        Reproducibility and prognosis of quantitative features extracted from CT images.
        Transl Oncol. 2014; 7: 72-87
        • Randen T.
        • Husøy J.H.
        Filtering for texture classification: a comparative study.
        IEEE Trans Pattern Anal Mach Intell. 1999; 21: 291-310
        • da Cunha Santos G.
        • Shepherd F.A.
        • Tsao M.S.
        EGFR mutations and lung cancer.
        Annu Rev Pathol. 2011; 6: 49-69
        • Liu W.S.
        • Zhao L.J.
        • Pang Q.S.
        • Yuan Z.Y.
        • Li B.
        • Wang P.
        Prognostic value of epidermal growth factor receptor mutations in resected lung adenocarcinomas.
        Med Oncol. 2014; 31: 771
        • Ozkan E.
        • West A.
        • Dedelow J.A.
        • et al.
        CT gray-level texture analysis as a quantitative imaging biomarker of epidermal growth factor receptor mutation status in adenocarcinoma of the lung.
        AJR Am J Roentgenol. 2015; 205: 1016-1025
        • Zhao B.
        • James L.P.
        • Moskowitz C.S.
        • et al.
        Evaluating variability in tumor measurements from same-day repeat CT scans of patients with non–small-cell lung cancer.
        Radiology. 2009; 252: 263-272
        • Hsu J.S.
        • Huang M.S.
        • Chen C.Y.
        • et al.
        Correlation between EGFR mutation status and computed tomography features in patients with advanced pulmonary adenocarcinoma.
        J Thorac Imaging. 2014; 29: 357-363