Spectral Computed Tomography Findings of Peripheral Lung Adenocarcinoma and Peripheral Squamous Cell Carcinoma

Background: To investigate the spectral computed tomography (CT) ndings of peripheral adenocarcinoma (P-AC) and peripheral squamous cell carcinoma (P-SCC) in lung. Methods: In this retrospective study, A total of 273 patients (150 patients with P-AC and 123 patients with P-SCC) conrmed by surgery and pathology who underwent chest contrast enhanced CT scan with GSI mode, including arterial phase (AP) and venous phase (VP). During two phases, The CT 40keV , CT 70keV , CT 100keV values, iodine concentration (IC), water concentration (WC), effective atomic number (Zeff) were measured and the slope of the spectral curve (K) was calculated. Differences between two groups were compared using two-sample t-test, Receiver operating characteristic (ROC) curves were plotted, and the area under the ROC curve (AUC) was also calculated to calculate diagnostic ecacies. Results: There was signicant difference in gender between the two groups (P < 0.05), No signicant difference between other clinical features and symptoms (P > 0.05). For AP and VP, the CT 40keV , CT 70keV , K70 keV , IC and Zeff of P-AC were signicantly higher than those of P-SCC (P<0.05), but there was no signicant difference in WC and CT100keV between the two groups. ROC curve analysis showed that the combination of all quantitative parameters in AP and VP had the best diagnostic performance, with the area under the curve, sensitivity and specicity of 92%, 88%, and 84%, respectively. Conclusions: Spectral CT can provide reference for the differentiation of P-AC and P-SCC.


Background
Lung cancer is one of the malignant tumors with the highest morbidity and mortality in the world [1]. With the increase of the aging population, the incidence of lung cancer is also on the rise. Non-small cell lung cancer accounts for about 85% of all lung cancer cases, adenocarcinoma (AC) and squamous cell carcinoma (SCC) are the two most common histological types of non-small cell lung cancer [2]. AC and SCC have different histopathological characteristics, biological characteristics and clinical treatment methods [3,4]. According to the anatomical location of the disease, lung cancer is divided into the central lung cancer and peripheral lung cancer. lung cancer can be divided into central type and peripheral type. SCC mainly occurs in the proximal airway, mainly in the central type, while AC mainly occurs in the distal airway, mainly in the peripheral type [5]. At present, the pathological types of lung cancer are determined by ne-needle aspiration biopsy, bronchoscopy, cytological examination or surgery to obtain tumor tissue [6]. However, these are invasive methods of examination, when some tumors are close to the bone or located deep, it is di cult to obtain tumor tissue [7], and biopsies can increase the risk of tumor metastasis [8].In addition, the early clinical symptoms of peripheral lung cancer are not obvious. Symptoms only appear as the disease progresses or metastasis occurs. At this time, most patients are in the middle and advanced stages of lung cancer and have lost the opportunity of surgical resection [9]. Therefore, it is necessary to identify subtypes of lung cancer in a non-invasive and easy-to-operate way.
Computed tomography (CT) is the preferred imaging method for early screening, diagnosis and stage assessment of lung cancer at present. Some studies have used traditional CT to distinguish peripheral adenocarcinoma (P-AC) from peripheral squamous cell carcinoma (P-SCC) [10][11][12][13]. For example, Yue and colleagues found that deep lobulated margins, thickened vascular bundles and pleural indentation are more common in AC, while smooth margins and bronchiectasis are more common in SCC [10]. Jiang et al. [13] studied the thin-slice CT ndings of peripheral lung cancer below 3 cm, and found that the incidence of air bronchogram sign in P-AC was signi cantly higher than that in P-SCC, and the difference was statistically signi cant (P < 0.05). Kunihiro et al. [12] compared the CT ndings of AC and SCC, and found that the maximum thickness of the cavity wall of AC was smaller than that of SCC and groundglass shadow and bronchiectasis were more common in AC. However, the CT signs selected in these studies are easily affected by the subjective judgment of the observer; in addition, the results of these studies are contradictory. Traditional CT can only provide some information about the morphological characteristics of the lesions and the relationship between the lesions and adjacent tissues, while the information for the identi cation of pathological subtypes of lung cancer is limited [14,15].
In recent years, with the development of energy imaging technology and the increase of clinical demand, the research of energy CT has become a hot spot, especially the spectral CT imaging that appeared in 2009, which provides a broader space for the research and clinical application of energy imaging [16,17].
Dual-energy spectral CT uses instantaneous kVp switching technology, which can provide material density images and single energy images, and perform multi-parameter quantitative analysis on the histological and biological characteristics of the lesions [17]. Compared with traditional CT, it has more advantages in distinguishing benign and malignant lung lesions and pathological grade of lung cancer [18,19]. The purpose of this study is to investigate the spectral CT ndings of P-AC and P-SCC and to provide a reference for the differential diagnosis of them.
Inclusion criteria: (1) not receiving any anti-tumor treatment before spectral CT scan; (2) no history of hypersensitivity or contraindications to iodine contrast agent; (3) the time interval between operation and spectral CT scan is within 2 weeks; (4) the patients are over 18 years old; (5) it was con rmed by histopathology. Exclusion criteria: (1) CT scan showed that the tumor is located in the center; (2) the area of necrosis or calci cation exceeded 50% of the tumor volume; (3) the tumor volume was too small to affect the measurement results; (4) there is heavy breathing artifacts or poor image quality. All patients signed an informed consent form before CT scan.

Ct Imaging Techniques
All patients underwent CT examinations with GSI mode using Discovery CT750 HD (GE Healthcare, Waukesha, WI, USA) scanner. Scanning range: the entrance of thorax to the level of costophrenic angle. Scanning parameters: tube voltage: high and low energy (80kVp, 140 kVp) instantaneous (0.5 ms) switching, tube current 375 mA, tube rotation time 0.7 s, screw pitch, 0.984:1; scanning eld of view, 50 cm; collimator width, 40 mm; scanning layer thickness and spacing are 5 mm. Then a bolus of 80 ~ 100 ml (1.2 mL/kg body weight) of non-ionic contrast media (Ultravist 300, Bayer Pharma, Berlin, Germany) was injected into the median cubital vein at a rate of 3.5 ~ 4 ml/s. The arterial phase (AP) and venous phase (VP) scanning were performed in GSI mode for 30 s and 60 s after contrast medium injection, respectively. The CT images of AP and VP are reconstructed with a layer thickness and spacing of 1.25 mm, Adaptive iterative reconstruction (ASIR) algorithm is used to reduce the noise of the image.

Radiological Diagnoses
Two radiologists with 10 years of experience in thoracic tumor diagnosis blindly used the GSI Viewer analysis software in the GE AW4.6 workstation to independently analyze the images, and reached an agreement after discussion in the case of differences. Select the largest layer of the lesion and its adjacent upper and lower layers to measure the region of interest (ROI) in three consecutive layers. In order to re ect the heterogeneity of the tumor as much as possible, when the density of the lesion is uniform, the area of the ROI should be larger than the cross-section of the lesion 1/2 of the surface. When the density of the lesion is uneven, the area with more solid components and uniform enhancement should be selected as much as possible. At the same time, in order to maintain the accuracy of the measurements, copy and paste functions are used to ensure that the ROI is consistent in size, shape, and location during the arterial and venous phases. An elliptical region of interest was placed in the area where should avoid the areas of blood vessels, calci cation, cavitation, necrotic cystic, atelectasis and so on. Each lesion is measured three times and the average value is calculated, and then the average value measured by two radiologists was calculated again. GSI Viewer software automatically generates the CT value, iodine concentration (IC), water concentration (WC) and effective atomic number (Zeff) under single energy of 40-100 keV (interval 10 keV), and calculates the slope of the spectral curve, According to the equation:

Statistical analysis
Statistical analysis was performed using SPSS version 22.0 (IBM Corporation, Armonk, NY, USA). The quantitative parameters were expressed as mean ± standard deviation and the counting data as percentage. The differences in gender, smoking history and clinical characteristics between the two groups were compared using the χ 2 test. the differences between the two groups in age and quantitative parameters were statistically compared using the two-sample t-test; The ROC curve was drawn for the above variables with statistical differences, and calculated the area under the ROC curve, sensitivity, speci city, maximum Youden index and optimal threshold.

Results
There were 92 males (74.7%) and 31 females (25.3%) with P-SCC and 68 males (45.3%) and 82 females (54.7%) with P-AC. there was statistical signi cance between the two groups in terms of gender (P < 0.001). There was no signi cant difference in age, smoking history and clinical symptoms between the two groups (P > 0.05) ( Table 1). In the AP and VP, the CT 40keV , CT 70keV , K 70keV , IC and Zeff of P-AC were greater than those of S-CC, and the difference was statistically signi cant (P < 0.001), while the WC and CT 100keV between the two groups were not statistically signi cant (  The ROC analysis results of spectral CT quantitative parameters of the AP and VP energy spectrum are presented in Table 3 and Fig. 3. The analysis shows that in AP, the area under the ROC curve of CT 40keV (0.84), K 70keV (0.83), Zeff (0.80), and the combination of quantitative parameters of the AP (0.87) are all ≥ 0.8, and the area under the ROC curve of the combination of quantitative parameters of the AP was the largest (0.87). In the venous phase, only the area under the ROC curve of the combination of quantitative parameters was ≥ 0.75, and the area under the ROC curve of CT 40keV (0.73) and K 70keV (0.71) was ≥ 0.7.
The area under the ROC curve of the combination of all quantitative parameters in AP and VP was 0.92, which was the largest among all the parameters in AP and VP. ROC curve analysis was used to determine the optimal threshold for the sensitivity and speci city of differentiating P-SCC and P-AC. For example, during Ap, when CT 40keV threshold was 153.83, the sensitivity and speci city of differentiating P-SCC and P-AC were 0.82 and 0.80, respectively. At the venous stage, the threshold value of CT 40keV was 124.49, and the sensitivity and speci city were 0.85 and 0.63, respectively. For the selected combination of optimal thresholds, the optimal threshold for distinguishing between P-AC and P-SCC were the combination of all combined quantitative parameters in AP and VP of 0.41, and the sensitivity and speci city were 0.88 and 0.84, respectively.

Discussion
With the development of personalized medicine and molecular targeting therapy, it is particularly important to determine the pathological type of lung cancer before starting treatment, because the e cacy of targeted drugs depends on the pathological subtype of lung cancer, For example, such as bevacizumab are effective in the treatment of AC, but it may lead to neutropenia and massive bleeding in patients with SCC [20,21]. The diagnostic accuracy of peripheral lung squamous cell carcinoma and peripheral lung adenocarcinoma is very limited due to the overlap of imaging signs on conventional CT. The multi-parameter of spectral CT can effectively re ect the tissue composition and biological characteristics of the tumor, and has great potential in identifying tumor subtypes and differentiation degree [17]. Wang et al. [22] studied the spectral CT manifestations of AC and SCC and found that spectral CT can provide both qualitative and quantitative parameters of the lesion, which provides a new method for differentiating them.
In this study, we found that CT 40keV , CT 70keV , K 70keV and IC in patients with P-AC in both AP and VP were higher than those in patients with P-SCC, which was consistent with the results of previous studies [23][24][25], indicating that the blood supply of AC was more abundant than that of SCC. According to the results of previous pathological studies [26], the capillary endothelial cells in normal tissues are tightly connected and the basement membrane is intact, and the contrast agent rarely penetrates into the intercellular space. However, a large number of new capillaries are formed in the tumor tissue, in which the capillary endothelial cells are loosely connected and the basement membrane is incomplete, so the contrast agent can easily penetrate into the intercellular space. Yazdani et al. [27] found that compared with SCC, AC is more likely to form rich and homogeneous cribriform capillaries, with greater microvessel density, and the maturity of neovascularization formed by SCC is not as good as that of AC. Neovascularization is more likely to be broken or blocked due to the rapid growth of tumor tissues. Therefore, the uptake of iodine contrast agent in AC is higher than that in SCC. According to histopathological analysis, AC is mainly composed of glandular structure, which contains rich interstiitium, loose internal structure and the number of cells per volume is large, while SCC is mainly composed of cancer nests, keratinocytes, intercellular bridges and other structures. The internal structure is dense and there are fewer tumor cells per volume [24], so the extent of the penetration of contrast agent will also be different.
The spectral curve re ects the change of the CT value of the lesion under different keV. We can judge the properties, homology and difference of lesions by analyzing the spectral curve of lesions, and it also can re ect the absorption of the contrast agent in the lesion [25]. The results showed that the slope of different energy intervals are different, and the CT value of tissues will decrease with the increase of energy, and the CT value of different energy levels representing the mass absorption coe cient of lesions at different energy levels [28]. In this study, the curve is steep in the range of 40 ~ 70 keV, and the curve is at in the range of 70 ~ 100 keV. This is related to the larger the X absorption coe cient and the more Xray attenuation at low energy. In this study, the slope of spectral curve between 40 and 70 keV was selected as the quantitative analysis index. The results showed that the K 70keV of AC was higher than SCC during both AP and VP, and the difference was statistically signi cant, which was consistent with the research results of other scholars [25]. The slope of spectral curve re ects the intensity of lesion enhancement, so it is considered that AC absorbs more iodine contrast agents than SCC, and contrast agents enhance the difference of mass absorption coe cient between the two groups of lesions.
The effective atomic number (Zeff) can directly re ect the atomic number of the compound inside the lesion. If the X-ray attenuation coe cient of the atomic number of an element is the same as that of the compound, then the atomic number of the element is the atomic number of the substance [26]. According to this feature, the composition and properties of compounds can be identi ed [17], especially those with similar densities and CT values. Some studies have shown that the Zeff can accurately describe the histological characteristics of the lesion and distinguish the material components [24,29,30]. In this study, the Zeff of P-AC was greater than that of P-SCC during both AP and VP, and the difference was statistically signi cant. This may be due to the different pathological tissue types, the material composition and cell metabolic activity of the lesions are also different. And the Zeff with enhanced scan is related to the uptake dose of contrast agents by the lesions, which leads to the difference of Zeff between the P-AC and P-SCC [31].
According to ROC curve analysis, the combination of all parameters in AP and VP showed higher sensitivity (88%) and speci city (84%) Compared with the quantitative parameters alone during AP or VP in distinguishing P-SCC and P-AC, and the diagnostic e ciency of the AP is higher than that of VP. Zhang et al. [7] who found that quantitative parameters in VP had greater signi cance in differentiating SCC and AC than in AP. However, Jia et al. [25] found that the quantitative parameters of the two phases had no signi cant difference in distinguishing SCC from AC. The difference between the above studies may be due to the fact that each quantitative parameter is related to the uptake of contrast agents by the lesions, which may be affected by the dose of contrast agent, the enhanced scanning time of each phase, and the patient's hemodynamic status, resulting in different Research results [31]. It may also be due to the samples included in these studies is small or the lack of distinction between lung cancer types, such as peripheral and central squamous cell carcinomas, which may affect the results between different parameters. At present, our study with the largest sample to distinguish P-AC and P-SCC. Therefore, our results are more accurate and generalized.
There are several limitations in this study. First, this study was a retrospective study, which may lead to sampling bias. Secondly, this study focuses on differentiating P-SCC and P-AC, and other histological subtypes of lung cancer were not included. In future studies, more lung cancer subtypes will be included to draw broader conclusions. Finally, in order to reduce the radiation dose, the patients were not scanned for the delayed phase. Therefore, more studies are needed to con rm whether the quantitative parameters during the delay period can differentiate P-SCC and P-AC.

Conclusions
The quantitative parameters of spectral CT have important value in the differentiating P-SCC and P-AC, and can provide certain imaging basis for the choice of treatment and prognosis of lung cancer patients.

Declarations
Ethics approval and consent to participate: The Ethics Committee of the Second Hospital of Lanzhou University (Lanzhou, China) approved the use of patient materials in this study.
Availability of data and materials: The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.