The associations among quantitative spectral CT parameters, Ki-67 expression levels and EGFR mutation status in NSCLC

Dual-energy spectral computed tomography (DESCT) is based on fast switching between high and low voltages from view to view to obtain dual-energy imaging data, and it can generate monochromatic image sets, iodine-based material decomposition images and spectral CT curves. Quantitative spectral CT parameters may be valuable for reflecting Ki-67 expression and EGFR mutation status in non-small-cell lung cancer (NSCLC). We investigated the associations among the quantitative parameters generated in DESCT and Ki-67 expression and EGFR mutation in NSCLC. We studied sixty-five NSCLC patients with preoperative DESCT scans, and their specimens underwent Ki-67 and EGFR evaluations. Statistical analyses were performed to identify the spectral CT parameters for the diagnosis of Ki-67 expression and EGFR mutation status. We found that tumour grade and the slope of the spectral CT curve in the venous phase were the independent factors influencing the Ki-67 expression level, and the area under the curve (AUC) of the slope of the spectral CT curve in the venous phase in the receiver operating characteristic analysis for distinguishing different Ki-67 expression levels was 0.901. Smoking status and the normalized iodine concentration in the venous phase were independent factors influencing EGFR mutation, and the AUC of the two-factor combination for predicting the presence of EGFR mutation was 0.807. These results show that spectral CT parameters may be useful for predicting Ki-67 expression and the presence of EGFR mutation in NSCLC.

tissues is always generated from biopsy analysis prior to surgery, which is an invasive procedure, and the results may be uncertain because of the small sample size in an individual case or a nonrepresentative sample selection. The gold standard for EGFR mutation testing also relies on the detection of tumour tissues from the biopsy or surgery 16 . Thus, exploring noninvasive methods to assess Ki-67 expression levels and the presence of EGFR mutation in NSCLC would be beneficial for lung cancer patients.
Recently, dual-energy computed tomography (CT), which generates both monochromatic image sets and iodine-based material decomposition images, has become increasingly used in diagnosing cancer patients 17 . One of the dual-energy CT technologies is spectral CT, which is based on fast switching between high and low voltages from view to view to obtain dual-energy imaging data. It is useful for scanning soft tissues and has a higher contrast-to-noise ratio than conventional multi-slice CT. It enables the collection of the quantitative iodine concentration on iodine-based material decomposition images and the reflection of the structural difference in tumours via attenuation changes as a function of photon energy (spectral CT curve); furthermore, dual-energy CT serves as a useful method for distinguishing between benign and malignant pulmonary lesions 18,19 . The quantitative iodine concentration and spectral CT curves in dual-energy spectral CT may be valuable for reflecting the expression of Ki-67 and EGFR mutation status in NSCLC. However, to date, there have been no specific studies in this area. Thus, the goal of our study was to investigate the potential association between the quantitative parameters generated in dual-energy spectral CT and Ki-67 expression levels and the presence of EGFR mutation in NSCLC.

Results patient information.
Sixty-five patients with NSCLC who underwent dual-energy spectral CT examination and subsequently underwent radical surgery were included in our study. According to the results of Ki-67 staining, patients were divided into the Ki-67 low-level expression group (Ki-67 ≤ 30%) and the high-level expression group (Ki-67 index>30%). There were 40 patients in the low-level group and 25 patients in the high-level group (Table 1). There were significant differences in pathological grade, tumour diameter, and EGFR mutation status between the Ki-67 low-level expression group and the high-level expression group, but there were no significant differences in sex, age, smoking status, histological type, or tumour stage. According to the EGFR gene mutation results, patients were divided into the EGFR wild-type group and the EGFR mutation group; there were 36 patients in the EGFR wild-type group and 29 patients in the EGFR mutation group (Table 2). In the EGFR mutation group, there were 15 patients with Exon19 Del, 3 patients with Exon20 Ins, 9 patients with Exon21 L858R, 1 patient with Exon21 L861Q and 1 patient with Exon18 G719X mutations. There was a significant difference in smoking status and Ki-67 expression levels between the EGFR wild-type group and the EGFR mutant group, while sex, age, pathological grade, histological type, tumour stage, and tumour diameter were not significantly different.
Dual-energy spectral CT findings with Ki-67 expression levels and EGFR mutation status. The spectral CT quantitative parameters were different between the two different Ki-67 expression level groups and the two different EGFR gene statuses ( Figs. 1 and 2).
The 70 keV CT values in the VP of the low-level Ki-67 expression tumours were significantly higher than those in the high-level Ki-67 expression tumours (64.63 ± 18.34 versus 52.40 ± 16.04, P = 0.008). The NIC values of the low-level Ki-67 expression tumours were significantly higher than those of the high-level Ki-67 expression tumours in both the AP (0.18 ± 0.08 versus 0.12 ± 0.04, P < 0.001) and VP (0.47 ± 0.11 versus 0.33 ± 0.07, P < 0.001). However, for the 70 keV CT values in the AP, there was no statistically significant difference between the low-level group and the high-level group (P > 0.05). The levels of λ HU were also significantly higher in the low-level Ki-67 expression tumours than in the high-level Ki-67 expression tumours in both the AP (2.02 ± 0.93 versus 1.34 ± 0.42, P < 0.001) and VP (2.81 ± 0.87 versus 1.76 ± 0.32, P < 0.001) ( Table 3). www.nature.com/scientificreports www.nature.com/scientificreports/ The NIC values of the EGFR mutated tumours were significantly higher than those of the EGFR wild-type tumours in both the AP (0.18 ± 0.08 versus 0.13 ± 0.06, P = 0.032) and VP (0.47 ± 0.14 versus 0.37 ± 0.07, P < 0.001). The λ HU values of the EGFR mutated tumours were significantly higher than those of the EGFR wild-type tumours in the VP (2.75 ± 1.06 versus 2.13 ± 0.56, P = 0.006), but the 70 keV CT values in the AP and VP and the λ HU values in the AP demonstrated no statistically significant difference between the EGFR wild-type and the EGFR mutated tumours (P > 0.05) ( Table 4).
The logistics regression analysis of the association of Ki-67 expression level with the spectral ct parameters and clinic characteristics. Logistic regression analysis was performed on the association between the Ki-67 expression level and the spectral CT parameters and clinical characteristics. In the univariate regression analysis, pathological grade, tumour diameter and EGFR mutation status as well as the spectral CT quantitative parameters of the 70 keV CT value in the VP, the NIC in the AP and VP, and the λ HU value in the AP and VP were significantly correlated with Ki-67 expression levels.
Further multivariate regression analysis was used to correct for the interaction of various factors affecting Ki-67 expression. It was found that tumour pathological grade and λ HU in the VP were independently correlated with Ki-67 expression. Between the Ki-67 high-level expression group and the Ki-67 low-level expression group, the odds ratio (OR) of tumour pathological grade (comparing grade III with grade I + II) for high-level Ki-67 expression was 5.291 with a 95% confidence interval of 1.166-24.010, and the OR of the λ HU in the VP (increased 1 Unit) for high-level Ki-67 expression was 0.020 with a 95% confidence interval of 0.002-0.197 (Table 5).
the logistics regression analysis of the association of eGfR mutation status with the spectral ct parameters and clinic characteristics. Logistic regression analysis was performed on the association between EGFR mutation status and the spectral CT parameters and clinic characteristics. In the univariate regression analysis, smoking status, tumour histological type, Ki-67 expression level and the spectral CT quantitative parameters of the NIC in the AP and VP and λ HU in the VP were significantly correlated with EGFR mutation status. Further multivariate regression analysis was used to correct for the interaction of various factors affecting EGFR mutation status. It was found that smoking status and the NIC in the VP were independently correlated with EGFR mutation status. Between the EGFR mutation group and the EGFR wild-type group, the OR of smoking status (comparing non-smoker with smoker) for EGFR mutation was 5.958 with a 95% confidence interval of 1.119-31.709, and the OR of the NIC in the VP (increased 1 Unit) for EGFR mutation was 1.102 with a 95% confidence interval of 1.040-1.168 (Table 6).

Diagnostic implications.
According to multivariate logistic regression analysis, the spectral CT quantitative parameter λ HU in the VP was an independent factor influencing Ki-67 expression level, and the spectral CT quantitative parameter of the NIC in the VP was an independent factor influencing EGFR gene mutation. According to ROC curve analysis, the AUC of using λ HU in the VP to predict the high Ki-67 expression of NSCLC was 0.901 with a sensitivity and specificity of 92% and 75%, respectively, and the diagnostic threshold was 2.16. The AUC of using the NIC in the VP to predict EGFR gene mutation in NSCLC was 0.772 with a sensitivity and specificity of 72.4% and 80.6%, respectively, and the diagnostic threshold was 0.41. Regarding the low sensitivity and specificity, two independent factors, smoking status and the NIC in the VP, were combined for ROC curve analysis. The AUC of the two-factor combination to predict EGFR gene mutation in NSCLC was 0.807 with a sensitivity and specificity of 72.4% and 85.8%, respectively (Figs. 3 and 4).  Table 2. Clinical data of the groups of EGFR mutation status in NSCLC.

Discussion
Ki-67 expression has long been regarded as a biological marker reflecting the aggressiveness of a tumour that can predict the prognosis of NSCLC patients. EGFR-TKIs have been reported as an effective treatment for NSCLC patients with EGFR mutations. Therefore, NSCLC patients, especially adenocarcinoma patients, are recommended to undergo EGFR mutation testing 20 . The results of the present study reveal that the 70 keV CT value in the VP and the NIC and λ HU in both the AP and VP were significantly different between the high-level Ki-67 expression and low-level Ki-67 expression tumours. In addition, the NIC in the AP and VP and λ HU in the VP had higher values in the EGFR-mutated tumours than in the EGFR wild-type tumours.
Enhanced CT values and iodine quantification in a lesion reflect the vascularity of the lesion. Additionally, the NIC serves as a benefit of spectral CT imaging in cancer characterization, and it was calculated to minimize the dose of contrast medium, the flow rate of injection, and individual differences in circulation in our study. We found a significant negative correlation among the 70 keV CT value in the VP, the NIC value in the AP and VP and Ki-67 expression-high expression of Ki-67 was usually correlated with a low 70 keV CT value in the VP and the NIC in the AP and VP. One possible explanation is that NSCLCs with radical proliferation may have an    www.nature.com/scientificreports www.nature.com/scientificreports/ insufficient blood supply in terms of an increasing trend towards a large volume, which usually leads to intratumoural necrosis. The proliferative activity in a tumour is commonly evaluated based on the Ki-67 value, and we also found that the tumour diameter had a significant positive correlation with the Ki-67 value-the higher the expression of Ki-67 was, the faster the growth and the larger the tumour volume. Spira et al. 21 reported that tumours with progressive growth appeared to have a large extent of tumour necrosis. The hypovascularity of a tumour before necrosis can affect tissue blood volume or vessel physiological function and can be reflected in the enhanced CT values and iodine concentration. In dual-energy spectral CT technology, the slope or change rate of the spectral curve has the ability to distinguish different materials because the different chemical compositions and organizations have different spectral curves 22,23 . Our study results showed that the λ HU values for NSCLC in the AP and VP were higher in the low-level Ki-67 expression tumours than in the high-level Ki-67 expression tumours. These differences may be explained by the fact that a tumour with different Ki-67 expression values may have different compositions and organizations. Increased Ki-67 values are associated with the malignancy of tumours, and this kind of tumour is characterized by a complex internal structure, such as an abnormal nuclear/cytoplasmic ratio with cells, highly abundant macromolecular proteins, and compact extracellular space. Karaman et al. 24 reported a strong negative correlation between the minimum apparent diffusion coefficient (ADC min ) values in MRI and the Ki-67 proliferation index in NSCLC. Since the ADC value is always used to quantify the diffusion of water molecules in tissues, this result may suggest that a change in Ki-67 expression   www.nature.com/scientificreports www.nature.com/scientificreports/ may influence the tumour internal structure. Furthermore, λ HU in the VP was an independent factor influencing Ki-67 expression level; λ HU in the VP reflects the difference in chemical compositions and organizations that are complex in fast proliferating tumours.
Regarding EGFR mutations, the NIC in the AP and VP and λ HU in the VP were higher in tumours with EGFR mutations than in those without EGFR mutations, and the NIC in the VP was an independent factor influencing EGFR mutation. The clinical characteristic of smoking status was also an independent influencing factor. EGFR is a member of the type I tyrosine kinase receptor family, has intrinsic tyrosine-kinase activity and plays an important role in tumour development. EGFR can phosphorylate extracellular-regulated kinases (ERKs) and can impact cell proliferation to some extent 25 . The abnormal proliferation of tumour cells is one reason for tumour heterogeneity, and this situation resulted in different λ HU values in the VP of the lesions between different EGFR mutation statuses. Previous studies have shown that the stimulation or inhibition of EGFR may cause differences in tumour-induced angiogenesis 26 , so the NIC, which represents the blood volume in the tumour, may be different according to the EGFR mutation status, results that are consistent with our study results. In tumour angiogenesis, vascular endothelial growth factor (VEGF) is one of the key stimulators of angiogenesis. Petit et al. 27 has reported that EGFR might contribute to tumour angiogenesis partially through its effects on the expression of VEGF. Meng Li et al. 28 explored the role of the NIC in the identification of EGFR mutation status in a cohort of East Asian patients with pulmonary adenocarcinoma. The NIC was an independent risk factor in their study, which is consistent with our results.
The ROC analysis was performed according to the results of the logistic regression. The ROC analysis in this study showed that the optimal diagnostic threshold of λ HU in the VP for predicting the Ki-67 expression level was 2.16, resulting in a sensitivity of 92% and a specificity of 75%. This might be explained by the intratumoural structural changes correlated with different Ki-67 expression levels, which contribute to the difference in λ HU in the VP. Nguyen et al. 29 found that there was a significant difference between the maximum FDG uptake (SUVmax) and Ki-67 expression levels, and SUXmax was an independent factor influencing survival after tumour treatment. The results of research by Murakami et al. 30 also showed a significant correlation between SUVmax and the Ki-67 expression index and tumour diameter. And Karaman et al. 24 found that the ADCmin may be an useful parameter to predict the tumour Ki-67 expression index. However, their diagnostic efficacies for differentiating Ki-67 expression in NSCLC have not explicitly been described in their articles. The AUC of the NIC in the VP for predicting the presence of EGFR mutation was 0.772 in our study. Although this AUC value seems to be slightly low, it is still significantly higher than that of Meng Li et al. 28 , where they explored the role of the NIC for the identification of EGFR mutation status in a cohort of East Asian patients with pulmonary adenocarcinoma; the NIC, as an independent risk factor, had an AUC of 0.650 for the prediction of EGFR mutation in their research. Further, the combination of smoking status and the NIC in the VP promoted an AUC of 0.807 and a specificity of 85.8%.
There are certain limitations in the present study. First, this study focused on the application value of spectral CT, so all patients included should have previously undergone a spectral CT scan. The results were limited by the small number of cases and the retrospective design of the study, as selection bias could not be completely avoided; as is well known, EGFR mutation is usually more frequent in females than in males, but in the present study, the cohort did not show this association. Thus, further studies with larger sample sizes and a prospective design should be performed to validate the results of the present study. Second, the flow rate of injection was fixed at 3 ml/s and was not adapted for body weight. Although we normalized the IC of the tumour with that of the aorta, the results might still have been influenced by this scanning method, and the weight of the patient should be taken into consideration in the future. Third, we worked diligently to achieve a radiological-pathological match in sample analysis; however, a perfect correspondence between pathological samples and the ROI on the spectral CT images was difficult to achieve in practice. Fourth, only limited sections were measured in our study, and although an average value was calculated, measurement deviations could not be avoided completely. Fifth, we focused on NSCLCs with solid portions, and pure ground glass nodules were not analysed because of their poor expression of Ki-67.

conclusions
Tumour pathological grade and λ HU in the VP were independent factors influencing the Ki-67 expression level, and smoking status and NIC in the VP were independent factors influencing EGFR mutation. Furthermore, λ HU in the VP was useful for predicting the Ki-67 expression level, and the combination of smoking status and NIC in the VP was helpful for predicting EGFR mutation in NSCLC. www.nature.com/scientificreports www.nature.com/scientificreports/ Methods patients. This retrospective study was approved by the Ethics Committee of Shanghai Ren Ji Hospital, the experimental protocols were performed in accordance with the approved guidelines, and the requirement for informed consent was waived because of the retrospective nature of the study. A retrospective analysis of data from patients who underwent dual-phase enhanced DESCT before surgery from March 2015 through October 2018 was performed. The inclusion criteria were (1) patients with solid types of non-small-cell lung cancer confirmed after surgery; (2) patients who underwent surgical resection within 2 weeks after CT examinations; and (3) patients with both Ki-67 and EGFR test results after surgery. The exclusion criteria were as follows: (1) patients who had undergone preoperative chemoradiotherapy, chemotherapy, radiotherapy or targeted therapy; (2) patients with metastatic tumours; and (3) patients with severe imaging artefacts. For patients with multifocal or multicentric lung cancer, only the largest surgical pathology-confirmed tumours were included. Ultimately, 65 lung cancer patients (43 men and 22 women; mean age, 64.29 years; range, 34-83 years) were included in the present study. All patients had dual-phase enhanced dual-energy spectral CT images and detailed pathological records of Ki-67 values and EGFR situations.
Dual-energy spectral ct imaging. All patients were scanned in a supine position with the hands above the head on a 64-row single-source, dual-energy spectral CT Discovery CT750 HD scanner (GE Healthcare, Milwaukee, WI, USA). In our institution, dual-phase contrast-enhanced CT examination for lung cancer is performed routinely, and the radiation dose for each contrast-enhanced CT examination was approximately 8.40 mGy in the volumetric CT dose index (CTDIvol) for non-contrast CT scans and 7.37 mGy in the CTDIvol for each of the dual-phase contrast-enhanced CT scans. The dual-phase contrast-enhanced scans were performed with the dual-energy spectral imaging mode. Non-ionic iodinated contrast material (iopamidol, 370 mg iodine/ ml; Shanghai Bracco Sine Pharmaceutical, Shanghai, China) was injected with a power injector at a dose of 1.35 ml/kg body weight at a rate of 3 ml/s through the hand vein. The dual-phasic enhancement CT for the arterial phase (AP) and venous phase (VP) started 35 and 90 seconds, respectively, after the administration of contrast agents. The dual-energy spectral CT mode scan parameters were as follows: instantaneous switch (<0.5 ms) between high (140 kVp) and low (80 kVp) tube voltages, a tube current of 375 mA, a tube rotation time of 0.6 seconds, a helical pitch of 1.375:1, a scan field of view of large body, and a collimation of 64 × 0.625 mm.
image analysis. All images were reconstructed at a section thickness of 1.25 mm, and measurements were performed on an Advanced Workstation (AW 4.6; GE Healthcare, USA) with Gemstone Spectral Imaging (GSI) Viewer software (GE Healthcare). Two experienced radiologists-both with 10 years of experience in chest imaging and who were blinded to the clinical and pathological information-analysed the images and measured data in consensus. Three consecutive image sections containing the section of the largest axial diameter of the lesion and the section above and below were chosen for measurement. Three regions of interest (ROIs) were drawn on the solid parts of the lesions of the three sections (areas close to 1/2 to 2/3 of the lesion area) while avoiding obvious necrosis, vessels, calcification and cystic portions. The 70 keV CT value was measured in the 70 keV monochromatic images. The iodine concentrations (ICs) of the lesions and aorta were measured from the iodine-based material decomposition images, and the CT values of lesions were measured from the 101 sets of virtual monochromatic images (with energy levels from 40 to 140 keV) by the GSI viewer software package. Then, average values were calculated from the measurements of the three consecutive image slices to minimize the bias of the measurement. The IC measurement itself was also influenced by various factors, such as the total dose of contrast medium, the flow rate of injection, and individual differences in circulation. To minimize the influence of variations in patients, the IC value of lung lesions was normalized to that of the aorta at the same section on the same iodine-based material decomposition image to calculate a normalized iodine concentration (NIC): = IC IC NIC / lesion aorta . The range of 40 to 100 keV in the spectral attenuation was chosen to calculate the slope of the spectral CT curve. This range was chosen because the spectral attenuation curve between 100 and 140 keV was almost flat and because the slope in the 40-100 keV range was larger to improve the sensitivity of describing different slopes of spectral attenuation curves than the whole range. The slope of the spectral curve was calculated according to the formula: Ki-67 expression and EGFR mutation analysis. The expression of Ki-67 in surgical specimens was analysed using a standard avidin-biotin immunohistochemistry (IHC) technique. The primary antibody for Ki-67 (Maixin Biotechnology Development Co., Ltd., Fuzhou, China, Kit-0005) was used according to the manufacturer's instructions. The presence of EGFR mutations was determined by the amplification refractory mutation system (ARMS) method using a human EGFR gene mutation detection kit (AmoyDx, Xiamen, China, ADx-EG01), and the EGFR mutations investigated included exon 19 deletions, L858R, T790M, G719X, S768I, and L861Q. The procedures were performed according to the manufacturer's instructions. The evaluation and analysis were carried out by a trained pathologist who was blinded regarding the case data, and the Ki-67 value was calculated. The Ki-67 value was defined as the percentage of positive cancer cells, which was calculated as the percentage of Ki-67-positive tumour cells divided by the total number of tumour cells within one high-power field. Based on previous findings 10,31 , cut-off values of 20-30% distinguishing high versus low Ki-67 expression were used in many studies, and in our study, the median Ki-67 value was 30%, so we determined that if ≤30% of the nuclei were stained, the section was scored as having low expression of Ki-67; when >30% of the malignant nuclei were positive, the section was deemed to have high expression. The EGFR mutation Scientific RepoRtS | (2020) 10:3436 | https://doi.org/10.1038/s41598-020-60445-0 www.nature.com/scientificreports www.nature.com/scientificreports/ positive values and test results were analysed according to the manufacturer's instructions and the experience of the pathologist. Based on the outcome of the EGFR test, the tumours were divided into the EGFR mutation group and the EGFR wild-type group.
Statistical analysis. All statistical calculations were performed with dedicated statistical software packages (SPSS, version 24.0, SPSS, Chicago, IL, USA; MedCalc, version 12.0, MedCalc Software, Mariakerke, Belgium). Quantitative variables were expressed as the means ± standard deviations (SDs). All reported P values were two-sided, and P < 0.05 was considered statistically significant. The two-sample t-test was used to assess the difference between spectral CT parameters for different Ki-67 expression levels or the EGFR mutation status. The logistics regression analyses were performed to identify independent factors in order to predict the Ki-67 expression levels and EGFR mutation status. Receiver operating characteristic (ROC) curves were generated to help establish the threshold values and the diagnostic capabilities for the significant spectral CT parameters for predicting Ki-67 expression and EGFR mutation. The optimal threshold value was evaluated by the Youden index. The diagnostic capability was determined by calculating the area under the ROC curve (AUC), and the sensitivity and specificity for each spectral CT parameter were obtained.

Data availability
The datasets generated and analysed during the current study are available from the corresponding authors upon reasonable request.