Lung cancer remains the leading cause of cancer-related mortality worldwide. The 5-year overall survival rate for all patients diagnosed with lung cancer is relatively low, about 15% to 20% regardless of the tumor stage and treatment received.[1–3] The recognition of specific molecular alterations in certain lung cancer sub-types, has facilitated tailored therapy and ushered in the era of “personalized” oncologic practice in the last decade. Within the family of lung carcinomas, the molecular foundation of lung adenocarcinoma is currently best understood; approximately 60% of all lung adenocarcinomas have an oncogenic driver mutation that, in many cases, predicts treatment response and correlates with certain clinicopathologic features.[5,6]
As for the imaging analysis of lung adenocarcinoma, computed tomography (CT) usually constitutes the first modality in evaluation and staging; other functional modalities such as positron emission tomography-CT or thoracic magnetic resonance imaging (MRI) have emerged to be important supplementary tools. Diffusion-weighted imaging (DWI), one of the widely applied functional MRI techniques, has shown potential for improved cancer detection, prediction of cancer aggressiveness, and evaluation of pathologic sub-types. However, a limitation of DWI is that it works on the assumption that water diffusion is Gaussian in behavior, which is unlikely to be the case in micro-structurally complex tissues. In such tissues, diffusion kurtosis imaging (DKI), a more recently described non-Gaussian technique, potentially better reflects water diffusivity in tissues with ultrahigh b values. The DKI model is sensitive to deviations of tissue diffusion from a Gaussian pattern and has been shown to be robust for parameter quantification, in addition to being more accurate to assess micro-structural complexity in a tissue than conventional DWI. Thus, the application of DKI has been successfully investigated in previous diffusion studies involving various human organs.[11–14]
In the current study, we evaluated the correlation of DKI parameters with the status of micro-structural molecular markers such as epidermal growth factor receptor (EGFR), anaplastic lymphoma kinase (ALK), and Ki-67 protein to investigate the link between non-invasive imaging parameters and clinicopathologic features, with the aim of providing more valuable information to increase the accuracy of detection, staging, and treatment monitoring for patients with advanced lung adenocarcinoma.
This study was approved by the Institutional Review Board of the Cancer Institute & Hospital, Chinese Academy of Medical Sciences (No. NCC2017ZDXM-001) and conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all patients.
Between July 2016 and June 2019, a total of 157 consecutive patients diagnosed with lung adenocarcinoma who underwent MRI examination at the Department of Diagnostic Radiology, Cancer Hospital, Chinese Academy of Medical Sciences were included in this study. MRI examinations were performed to evaluate and preoperatively stage these lesions, and the DKI sequence was used for clinical application along with routine conventional MRI. The following inclusion criteria were applied in this study: (1) patients with tumors that were histopathologically confirmed as primary lung adenocarcinomas by subsequent resection or biopsy; (2) patients did not receive any therapy or surgery prior to MRI examination; and (3) patients who underwent routine MRI, DWI, and DKI in the same scanner. The following exclusion criteria were applied: (1) patients with lesions that showed ground-glass opacity (GGO) on MRI; and (2) patients with lesions smaller than 2 cm. Finally, a total of 96 patients (42 males and 54 females) with histopathologically confirmed primary lung adenocarcinoma were enrolled in this retrospective study for data analysis.
All MRI scans were performed on a 3.0-T whole-body scanner equipped with a 32-channel coil (Discovery MR750; GE Healthcare, Milwaukee, WI, USA). The routine lung imaging protocol included the following sequences: axial propeller T2-weighted imaging with fat suppression (T2WI/FS), axial fast spin echo T1-weighted imaging, DWI, and DKI. DWI was obtained by using respiratory-gated, single-shot, spin-echo, echo-planar technology with b values of 0, 800 s/mm2. DKI was performed in the axial plane with b values of 0, 500, 1000, 1500, and 2000 s/mm2. Details of scanning parameters are shown in Table 1. The scanning range of all sequences was 2 cm above and below the target lesion.
Quantitative image analysis
All images were successfully acquired. Subsequently, parametrical maps of apparent diffusion coefficient (ADC), Dapp, and Kapp were calculated on an offline workstation (GE Advantage Workstation AW4.6; GE Healthcare). The mean ADC was determined on the basis of the assumption of a mono-exponential relationship between signal intensity Sb and b value in the DWI model:
DKI model quantifies the non-mono-exponentiality of the diffusion by means of a second-order Taylor series expansion. To obtain Dapp and Kapp parameters, we applied the signal intensity data of five b values based on Rosenkrantz et al's study, using the following equation:
Three quantitative parameters, based on the above calculation models, could be partitioned into two categories: diffusion coefficient (Dapp, ADC) and kurtosis coefficient (Kapp). Images were analyzed by a radiologist with the experience of interpreting DKI data. The imaging quality was evaluated as adequate or inadequate by the reader before further parameter analysis. Inadequate DKI examinations were mainly caused by (a) excessive motion artifacts; or (b) the high signal barely visible owing to substantial signal loss. The size for each lesion was measured as the average of maximal length and width on the T2WI/FS imaging. Then the reader documented the ADC, Dapp, and Kapp values of 96 patients with lung adenocarcinoma on DWI and DKI, respectively. A circular region of interest covering >70% of the maximal lesion area was placed on the T2WI/FS by using freehand selection method and taking care to avoid any cystic components or cavities as well. The identical region of interest was positioned in the corresponding ADC, Dapp, and Kapp maps next. The mean ADC, Dapp, and Kapp values were automatically measured.
Histologic analysis was performed, according to the 2015 World Health Organization Classification of Tumors of the Lung and Pleura, using tissue samples obtained at the time of either surgical resection or image-guided biopsies. In this study, 29 patients underwent surgical resection of the lesion, 52 patients underwent thoracoscopic surgery, and CT-guided transthoracic core-needle biopsy of lung tumors was performed in 15 patients.
An experienced pathologist evaluated the pathologic sub-types and results of EGFR mutations, ALK rearrangements, and Ki-67 protein expression status in all lesions. Genomic deoxyribonucleic acid was extracted from the tumor specimen, and EGFR tyrosine kinase exons 19, 20, and 21 were amplified by a nested polymerase chain reaction (PCR) using specific primers. ALK rearrangements were detected by means of fluorescence in situ hybridization or reverse transcription-PCR. Samples were deemed to be fluorescence in situ hybridization-positive if more than 15% of scored tumor cells showed split ALK 5′ and 3′ probe signals or had isolated 3′ signals. Reverse transcription-PCR was used to detect some other specific ALK fusions, for instance, echinoderm microtubule-associated-protein-like 4-anaplastic lymphoma kinase (EML4-ALK). A recent study that investigated the Ki-67 proliferative index (PI) in three large, independent non-small-cell lung cancer cohorts found that the statistically optimal cutoff for classification of lung adenocarcinoma as good and poor prognosis was 25%. Hence, the populations in this study were divided into high-expression (PI ≥ 25%) and low-expression (PI < 25%) groups. These results were collected and reviewed in conjunction with abovementioned imaging parameters.
Quantitative variables, tested with the Kolmogorov-Smirnov test for normality analysis and with the Levene test for variance homogeneity analysis, were expressed as mean ± standard deviation. Differences between these sub-groups were compared by using independent-samples t test. Categorical data were presented as counts and percentages, and then sub-groups were analyzed by a Chi-squared test or the Fisher's exact test. Receiver operating characteristic (ROC) curve analysis was performed to calculate the area under the curve (AUC) to determine the accuracy of ADC, Kapp, and Dapp in differentiating lesions with different immunohistochemical expression. Sensitivity, specificity, positive predictive value, and negative predictive value were generated using the optimal cutoff values. The diagnostic accuracy could be assessed according to the AUC: excellent, 0.9 < AUC ≤ 1.0; good, 0.8 < AUC ≤ 0.9; fair, 0.7 < AUC ≤ 0.8; poor, 0.6 < AUC ≤ 0.7; fail, 0.5 < AUC ≤ 0.6.
The Spearman correlation analysis was used to evaluate the association of imaging parameters with molecular markers’ expression. A correlation coefficient (r) of 1.0 was deemed to indicate perfect correlation; 0.8 to 0.9, strong correlation; 0.6 to 0.7, moderate correlation; 0.3 to 0.5, fair correlation; and lower than 0.3, poor or no correlation. All statistical analyses were carried out using SPSS 23.0 for Windows (IBM SPSS Inc, Chicago, IL, USA). Two-sided P values of <0.05 were considered statistically significant.
Ninety-six lung adenocarcinoma lesions from 96 patients were included with a mean size of 4.1 ± 3.2 cm (range 2.1–7.3 cm). The 96 patients included 42 males (43.8%) and 54 females (56.2%) with a median age of 66 years (44–82 years). Of these, 53 (55.2%) showed EGFR-positive mutations, 12 (12.5%) showed ALK rearrangements, and 83 (86.5%) showed high Ki-67 expression. In our study, patients with ALK rearrangements tended to be younger than those without (78.0% vs. 22.0%, χ2 = 4.669, P = 0.028). The two groups displayed no significant difference in the mean size (4.0 ± 2.8 cm vs. 4.1 ± 3.1 cm, t = 1.116, P > 0.05). Two (2.1%) lesions showed both EGFR gene mutations and ALK gene rearrangements. The most histologic sub-types were lepidic and acinar adenocarcinomas (accounting for 25%, respectively), and rather late pathologic stages of the tumors were recorded (stage IIIA-IV).
Summary of kurtosis metrics
Table 2 summarizes the ADC, Kapp, and Dapp metrics in lesions with different immunohistochemical expression.
For the Kapp values, EGFR mutation-positive group was significantly higher than EGFR mutation-negative group (0.81 ± 0.12 vs. 0.66 ± 0.10, t = 6.41, P < 0.001); ALK rearrangement-negative group was significantly higher than ALK rearrangement-positive group (0.76 ± 0.12 vs. 0.60 ± 0.15, t = 4.09, P < 0.001); high Ki-67 expression group was significantly higher than low Ki-67 expression group (0.76 ± 0.12 vs. 0.58 ± 0.13, t = 4.88, P < 0.001).
For the Dapp values, EGFR mutation-negative group was significantly higher than EGFR mutation-positive group (3.59 ± 0.77 μm2/ms vs. 3.11 ± 0.73 μm2/ms, t = 3.12, P = 0.002); ALK rearrangement-positive group was not significantly different from ALK rearrangement-negative group (3.73 ± 1.26 μm2/ms vs. 3.26 ± 0.68 μm2/ms, t = 1.96, P = 0.053); low Ki-67 expression group was significantly higher than high Ki-67 expression group (4.20 ± 0.83 μm2/ms vs. 3.19 ± 0.69 μm2/ms, t = 4.80, P < 0.001).
For the ADC values, EGFR mutation-negative group was significantly higher than EGFR mutation-positive group ([1.50 ± 0.53] × 10−3 mm2/s vs. [1.19 ± 0.37] × 10−3 mm2/s, t = 3.38, P = 0.001); ALK rearrangement-positive group was not significantly different from ALK rearrangement-negative group ([1.34 ± 0.81] × 10−3 mm2/s vs. [1.33 ± 0.41] × 10−3 mm2/s, t = 0.07, P = 0.941); low Ki-67 expression group was significantly higher than high Ki-67 expression group ([1.67 ± 0.77] × 10−3 mm2/s vs. [1.28 ± 0.39] × 10−3 mm2/s, t = 2.88, P = 0.005). Representative parametric maps are shown in Figures 1 and 2.
In patients with advanced lung cancer, according to the Spearman analysis, there was a strong positive correlation between Kapp and EGFR mutations or Ki-67 PI (r = 0.844, P = 0.008; r = 0.882, P = 0.001, respectively), and a strong negative correlation with ALK rearrangements (r = −0.772, P = 0.001). Dapp was moderately negatively correlated with EGFR mutations or Ki-67 PI (r = −0.650, P = 0.024; r = −0.734, P = 0.012), whereas ADC only had moderate negative correlation with Ki-67 PI (r = −0.679, P = 0.033). The correlations between Dapp and ALK rearrangements, ADC and EGFR mutations, ADC and ALK rearrangements, were not statistically significant (P = 0.137, 0.061, 0.612, respectively).
Table 3 summarizes the AUC for identification of lesions with different immunohistochemical findings for each of the metrics and the optimal thresholds for each of the metrics identified in the ROC analysis. With either scheme, Kapp had a mildly higher AUC for prediction of adverse final pathologic findings (AUC, 0.79–0.88) than ADC (AUC, 0.49–0.73) or Dapp (AUC, 0.60–0.86), and differences in performance between the metrics were not significant (P = 0.183, P = 0.734, respectively). The ROC curves for the three classification schemes are depicted in Figure 3.
MRI is an attractive technique that provides an integral assessment of several morphologic and functional techniques to evaluate different tumor characteristics. In the current study, we evaluated DKI-derived parameters to characterize the micro-structural properties of advanced lung adenocarcinomas and correlated them with the corresponding histopathologic findings, to provide a better opportunity for radiologists to potentially gain further insights into the tissue characteristics and improve clinical management triage as compared with the use of standard DWI.
In this study, we found there was a strong positive/negative correlation between kurtosis coefficients (Kapp) with these molecular markers’ status. Besides, the Kapp values in the EGFR mutation-positive group, ALK rearrangement-negative group, and high Ki-67 expression group were significantly higher than those in the control groups. Research shows Kapp represents a unitless parameter, larger Kapp indicates greater deviation from perfectly Gaussian diffusion behavior, and the Kapp parameter is likely associated with micro-structural complexity in vivo. When a genetic mutation arose in EGFR gene, it could have an impact on complicated micro-structures in biologic tissues such as membranes, myelin sheaths, and neural axons, leading to topologic rearrangement and complexity[23,24] reflected by increased kurtosis and hence a higher Kapp value. As for the Ki-67 antigen, one of several cell cycle-regulating proteins, is proved to be associated with ribosomal RNA transcription, and numerous studies have suggested that tumor cells with high Ki-67 expression exhibited higher cellularity with nuclear atypia.[26–28] Conversely, tumor cells with negative or low Ki-67 expression have loose cellularity and are typically associated with liquefactive necrosis and local fibrosis, resulting in few diffusion barriers.[29,30] The Kapp parameter represents excessive diffusion kurtosis in the tissue. Thus, it is possible to use differences in Kapp values observed in our study to reflect the differences in micro-structural irregularity and heterogeneity between these sub-groups.
In addition, a moderate correlation between diffusion coefficient (Dapp and ADC) and Ki-67 PI or EGFR mutations was observed in the study. In the DKI model, Dapp is an adjusted ADC value that accounts for this non-Gaussian diffusion behavior. In our current study, the Dapp values were significantly lower among patients with high Ki-67 PI or positive EGFR mutations (P < 0.001, P = 0.002, respectively). It could be explained that the EGFR mutation-positive group and high Ki-67 expression group may have an impact on the restriction of water diffusion that can be reflected by decreasing Dapp values. Further details about this transition remain to be confirmed. In the subsequent ROC analysis for micro-environment assessment, both Kapp and Dapp were a little superior to ADC; thus, DKI has been shown to reflect the micro-structural characteristics of adenocarcinoma tissue slightly more accurately than conventional DWI. However, the diagnostic accuracy of these three parameters was not good enough (AUC ranges from 0.49 to 0.88). Upon correlating the Dapp and ADC values with ALK rearrangements, no statistically significant differences were observed with respect to the DKI Dapp values; however, there was a non-significant trend toward lower Dapp values among patients with negative ALK rearrangements than those with positive rearrangements. The differences in mean Dapp or ADC values between the groups with or without ALK rearrangements were not statistically significant; this may be because of the low rearrangement rate in our study. These results show that the DKI model affords a metric that reflects excess kurtosis in a tissue and contributes to further analysis of molecular biomarkers in lung adenocarcinoma.
There were some limitations in this study. First, image quality of small lesions (for example, less than 2 cm) is more easily affected by the patient's heart rate, fluctuation of heart rate, and breathing artifact, causing increased anatomic distortions; hence, our current study focuses on patients with relatively large lesions. Second, MRI scanners in common clinical use were still limited to obtain sufficient signal-to-noise ratio of GGO lesions at ultrahigh b value images (>1000 s/mm2), so GGO lesions are not included in this research sample. As a whole, our observations based on a small sampling of MRI examinations are still preliminary, and additional prospective studies are warranted to assess the utility of DKI metrics in predicting clinical outcomes.
We found that the DKI model contains specific information on the non-Gaussian diffusion behavior, provides additional parameters such as Kapp, as an indicator of immunohistochemical findings, has a high value when assessing the patients with advanced lung adenocarcinomas, and shows slightly better diagnostic accuracies than the conventional DWI model. Using the DKI model during lung MRI is technically feasible in clinical routine, as it provides a practical clinical tool to quantify non-Gaussian water diffusion and probe the microscopic structure of biologic tissues.
This study was supported by a grant from Chinese Academy of Medical Sciences Initiative for Innovative Medicine Program (No. 2017-I2M-1-005).
Conflicts of interest
1. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2019. CA Cancer J Clin
2019; 69:7–34. doi: 10.3322/caac.21551.
2. Fukui T, Okasaka T, Kawaguchi K, Fukumoto K, Nakamura S, Hakiri S, et al. Conditional survival after surgical intervention in patients with non-small cell lung cancer. Ann Thorac Surg
2016; 101:1877–1882. doi: 10.1016/j.athoracsur.2015.11.067.
3. Jakobsen E, Rasmussen TR, Green A. Mortality and survival of lung cancer in Denmark: results from the Danish Lung Cancer Group 2000–2012. Acta Oncol
2016; 55: (Suppl 2): 2–9. doi: 10.3109/0284186X.2016.1150608.
4. Kim ES, Herbst RS, Wistuba II, Lee JJ, Blumenschein GR Jr, Tsao A, et al. The BATTLE trial: personalizing therapy for lung cancer. Cancer Discov
2011; 1:44–53. doi: 10.1158/2159-8274.CD-10-0010.
5. Sholl LM. Biomarkers in lung adenocarcinoma
: a decade of progress. Arch Pathol Lab Med
2015; 139:469–480. doi: 10.5858/arpa.2014-0128-RA.
6. Papadimitrakopoulou V, Lee JJ, Wistuba II, Tsao AS, Fossella FV, Kalhor N, et al. The BATTLE-2 study: a biomarker-integrated targeted therapy study in previously treated patients with advanced non-small-cell lung cancer. J Clin Oncol
2016; 34:3638–3647. doi: 10.1200/JCO.2015.66.0084.
7. Ohno Y, Koyama H, Yoshikawa T, Matsumoto S, Sugimura K. Lung cancer assessment using MR imaging: an update. Magn Reson Imaging Clin N Am
2015; 23:231–244. doi: 10.1016/j.mric.2015.01.012.
8. Padhani AR, Liu G, Koh DM, Chenevert TL, Thoeny HC, Takahara T, et al. Diffusion-weighted magnetic resonance imaging
as a cancer biomarker: consensus and recommendations. Neoplasia
2009; 11:102–125. doi: 10.1593/neo.81328.
9. Le Bihan D. Apparent diffusion coefficient and beyond: what diffusion MR imaging can tell us about tissue structure. Radiology
2013; 268:318–322. doi: 10.1148/radiol.13130420.
10. Rosenkrantz AB, Padhani AR, Chenevert TL, Koh DM, De Keyzer F, Taouli B, et al. Body diffusion kurtosis imaging
: basic principles, applications, and considerations for clinical practice. J Magn Reson Imaging
2015; 42:1190–1202. doi: 10.1002/jmri.24985.
11. Pentang G, Lanzman RS, Heusch P, Müller-Lutz A, Blondin D, Antoch G, et al. Diffusion kurtosis imaging
of the human kidney: a feasibility study. Magn Reson Imaging
2014; 32:413–420. doi: 10.1016/j.mri.2014.01.006.
12. Sun K, Chen X, Chai W, Fei X, Fu C, Yan X, et al. Breast cancer: diffusion kurtosis MR imaging-diagnostic accuracy and correlation with clinical-pathologic factors. Radiology
2015; 277:46–55. doi: 10.1148/radiol.15141625.
13. Pavilla A, Gambarota G, Arrigo A, Mejdoubi M, Duvauferrier R, Saint-Jalmes H. Diffusional kurtosis imaging (DKI) incorporation into an intravoxel incoherent motion (IVIM) MR model to measure cerebral hypoperfusion induced by hyperventilation challenge in healthy subjects. MAGMA
2017; 30:545–554. doi: 10.1007/s10334-017-0629-9.
14. Barrett T, McLean M, Priest AN, Lawrence EM, Patterson AJ, Koo BC, et al. Diagnostic evaluation of magnetization transfer and diffusion kurtosis imaging
for prostate cancer detection in a re-biopsy population. Eur Radiol
2018; 28:3141–3150. doi: 10.1007/s00330-017-5169-1.
15. Glenn GR, Tabesh A, Jensen JH. A simple noise correction scheme for diffusional kurtosis imaging. Magn Reson Imaging
2015; 33:124–133. doi: 10.1016/j.mri.2014.08.028.
16. Travis WD, Brambilla E, Nicholson AG, Yatabe Y, Austin J, Beasley MB, et al. The 2015 World Health Organization Classification of lung tumors: impact of genetic, clinical and radiologic advances since the 2004 classification. J Thorac Oncol
2015; 10:1243–1260. doi: 10.1097/JTO.0000000000000630.
17. Warth A, Cortis J, Soltermann A, Meister M, Budczies J, Stenzinger A, et al. Tumour cell proliferation (Ki-67) in non-small cell lung cancer: a critical reappraisal of its prognostic role. Br J Cancer
2014; 111:1222–1229. doi: 10.1038/bjc.2014.402.
18. Xia J, Broadhurst DI, Wilson M, Wishart DS. Translational biomarker discovery in clinical metabolomics: an introductory tutorial. Metabolomics
2013; 9:280–299. doi: 10.1007/s11306-012-0482-9.
19. Akoglu H. User's guide to correlation coefficients. Turk J Emerg Med
2018; 18:91–93. doi: 10.1016/j.tjem.2018.08.001.
20. Broncano J, Luna A, Sánchez-González J, Alvarez-Kindelan A, Bhalla S. Functional MR imaging in chest malignancies. Magn Reson Imaging Clin N Am
2016; 24:135–155. doi: 10.1016/j.mric.2015.08.004.
21. Jensen JH, Helpern JA. MRI quantification of non-Gaussian water diffusion by kurtosis analysis. NMR Biomed
2010; 23:698–710. doi: 10.1002/nbm.1518.
22. Raab P, Hattingen E, Franz K, Zanella FE, Lanfermann H. Cerebral gliomas: diffusional kurtosis imaging analysis of microstructural differences. Radiology
2010; 254:876–881. doi: 10.1148/radiol.09090819.
23. Siegelin MD, Borczuk AC. Epidermal growth factor receptor
mutations in lung adenocarcinoma
. Lab Invest
2014; 94:129–137. doi: 10.1038/labinvest.2013.147.
24. Aguirre A, Dupree JL, Mangin JM, Gallo V. A functional role for EGFR signaling in myelination and remyelination. Nat Neurosci
2007; 10:990–1002. doi: 10.1038/nn1938.
25. Collier Q, Veraart J, Jeurissen B, Vanhevel F, Pullens P, Parizel PM, et al. Diffusion kurtosis imaging
with free water elimination: a Bayesian estimation approach. Magn Reson Med
2018; 80:802–813. doi: 10.1002/mrm.27075.
26. Sobecki M, Mrouj K, Camasses A, Parisis N, Nicolas E, Llères D, et al. The cell proliferation antigen Ki-67 organises heterochromatin. Elife
2016; 5:e13722doi: 10.7554/eLife.13722.
27. Karaman A, Durur-Subasi I, Alper F, Araz O, Subasi M, Demirci E, et al. Correlation of diffusion MRI with the Ki-67 index in non-small cell lung cancer. Radiol Oncol
2015; 49:250–255. doi: 10.1515/raon-2015-0032.
28. Chirieac LR. Ki-67 expression in pulmonary tumors. Transl Lung Cancer Res
2016; 5:547–551. doi: 10.21037/tlcr.2016.10.13.
29. Rosenkrantz AB, Prabhu V, Sigmund EE, Babb JS, Deng FM, Taneja SS. Utility of diffusional kurtosis imaging as a marker of adverse pathologic outcomes among prostate cancer active surveillance candidates undergoing radical prostatectomy. AJR Am J Roentgenol
2013; 201:840–846. doi: 10.2214/AJR.12.10397.
30. Bodey B, Bodey B Jr, Gröger AM, Siegel SE, Kaiser HE. Clinical and prognostic significance of Ki-67 and proliferating cell nuclear antigen expression in childhood primitive neuroectodermal brain tumors. Anticancer Res