Ultra-High-Resolution CT Features of Pulmonary Nodules Correlate with Visceral Pleural Invasion in Early Stage Lung Adenocarcinoma


 Backgroundvisceral pleural invasion (VPI) is an important prognostic factor in early stage lung adenocarcinoma, which can affect the TNM Classification of Tumors.PurposeTo investigate whether ultra-high-resolution computed tomography (U-HRCT) features can predict VPI of early stage pulmonary nodules contacting the interlobar pleura.Material and MethodsA total of 126 patients with lung adenocarcinoma (age, 24-77 years) confirmed by surgical pathology were retrospectively enrolled. All patients underwent U-HRCT scan and were divided into two groups according to pulmonary nodular type: pure (pGGN) and mixed (mGGN). Clinical features were recorded, and U-HRCT features were manually measured using PHILIPS EBW V4.5.5. Univariate and multivariate logistic regression were used to determine factors that can significantly predict VPI. ResultsU-HRCT and three-dimensional orthogonal post-processing method could better display the relationship between GGNs and interlobar fissures. Among all patients, fifteen patients (12%) had VPI. None of the patients with pGGN had VPI. In the mGGN group, the solid ratio (odds ratio [OR]=1.275, 95% CI 1.1-1.478; P=0.001) and solid diameter (OR=1.139, 95% CI 1.06-2.346; P=0.046) were independent risk factors for VPI in early stage lung adenocarcinoma. For VPI diagnosis, the area under the curve, sensitivity, and specificity of the solid ratio and solid diameter were 0.803, 80%, and 75% and 0.807, 80%, and 80.36%, respectively.ConclusionU-HRCT can display GGNs and interlobar fissures in detail. VPI was not detected in patients with pGGN. In patients with mGGNs, a solid diameter >6mm and solid ratio >38% can be independent predictors of VPI, which may be helpful in surgical decision-making.


Introduction
Incidence and mortality of lung cancer exhibit an increasing trend every year, and non-small cell lung cancer (NSCLC) is reported to account for about 85% of all lung cancers. Visceral pleural invasion (VPI) is an important adverse prognostic factor of NSCLC and can directly change the assessment of the clinical stage, thereby affecting the treatment strategy (1,2). Oyama (3) reported 5-year survival rates of 86%, 62-70%, and 57% for no pleural invasion, VPI, and parietal pleural invasion in peripheral NSCLC, respectively. The 8th ed of TNM staging for lung cancer, tumor with VPI is classi ed as T2. Some studies have suggested that a T1 sized tumor with invasion of interlobar pleura and into the adjacent lobe, it should be classi ed as T3, not T2, even if the tumor is 3cm or less (4,5). Zhao LL found that tumor size, the relationship between nodules and pleura were important predictors of VPI positivity. VPI was more common in larger (> 2 cm) ground glass nodules (GGN). The radiographic ndings of nodule abutment or a pleural tag did not reliably predict or exclude VPI (6). However, the research to determine whether GGN of T1 lung adenocarcinoma contacting the interlobar ssure exists with VPI is lacking. Glazer (7) proposed in 1985 that computed tomography (CT) could be used to predict pleural invasion.
Different prediction and scan methods are needed based on the relationship between the tumor and pleura. VPI is more likely when the tumor is directly connected to the pleura. Kim H (8) found that conventional CT features for pathologic VPI are not independent prognostic factors in clinical T1 lung adenocarcinomas. Thus, upstaging from clinical T1 to T2 for lung cancers closely contacting pleura or exhibiting pleural tags in CT scans is not an evidence-based practice. With the development of highresolution CT technology and the wide application of low-dose CT in lung cancer screening, the detection rate of lung GGNs is increasing. Kakinuma (9) found that ultra-high-resolution CT (U-HRCT) displayed details comparable to conventional CT but provided better image quality. In this study, we adopted the 1024×1024 matrix of the U-HRCT scanning scheme (10) combined with the three-dimensional orthogonal post-processing method, and then tried to explore the U-HRCT features as VPI prognostic factors in T1 peripheral lung adenocarcinoma in contact with the interlobar ssure, and then to help clinicians evaluate surgical options and prognosis.

Material And Methods
The Institutional Review Board of our hospital approved this retrospective study and waived the requirement for informed patient consent.

Patients
Electronic medical records were retrieved from the radiology picture archiving and communication system (PACS) in our hospital, and a retrospective analysis of clinical data, including CT ndings and pathology records, from January 2014 to October 2018 was performed. Inclusion criteria were as follows: (a) lung adenocarcinoma or its preinvasive lesions equal to or less than 3cm in diameter measured on chest CT imaging without distant metastases; (b) GGNs contacting the interlobar ssure; (c) surgical resection; and (d) availability of U-HRCT data with three-dimensional orthogonal post-processing (10). Excluded patients were: (a) those who had received chemotherapy, radiotherapy, or chemoradiotherapy preoperatively; (b) those with multiple primary lung cancers; (c) those without U-HRCT data. In all, 126 patients were included in the study (44 men, 82 women; mean age, 55 ± 11 years). Patients were divided into the pure GGN (pGGN; n = 55) and mixed GGN (mGGN; n = 71) groups.

CT Scanning Protocol
CT scanning was performed using iCT 256 Or Brilliance 64 (Philips Healthcare). All patients underwent breathing exercises before scanning, and all scans were performed under deep inspiration breath-hold. Considering the radiation dose, Only U-HRCT scan having better image quality and detail display ability than conventional CT was performed (10)(11)(12), the parameters were as follows: FOV 250 mm, matrix 1024×1024, lateral or oblique lateral lying position with the lesion positioned as high as possible in the scanning lung eld (the lesion and its surrounding background were in ated as much as possible by physiological breathing), scanning angle of view image (view angle) 90°, 120kV, 250mAs, reconstruction thickness 1mm / space 1mm, FBP-D reconstruction algorithm lung window (C -500HU W 1500HU); reconstruction FOV 130-250mm, reconstruction thickness 0.67mm, space 0.3mm, iDose 4 4-6, low-pass ltered smooth reconstruction of lung window (C -500HU W 1500HU).

CT Image Processing and Evaluation
All images were transmitted to a workstation (PHILIPS EBW V4.5.5) for measurement and threedimensional orthogonal post-processing. Multiplanar reformation (MPR) technology was carried out, and the bronchial vascular bundle or pleura was used as the reference to complete a three-dimensional orthogonal display (in our study, the correlation between GGN and interlobar pleura was studied, and the interlobar pleura was used as the reference).
MPR displayed the original layer thickness (0.67mm and 2mm), while volume rendering was used to display the spatial three-dimensional morphological features of the lesions. All CT images were evaluated separately by two senior chest imaging diagnostic physicians (10-and 15-years' experience) who were blinded to the pathological results. The major CT features included nodule type, size, shape, density, solid diameter, solid ratio, interface length, interface length ratio, contacting solid, pleural thickening, and pleural indentation (13)(14)(15). Maximum nodule size and maximum solid diameter were measured on a three-dimensional image. Nodule shape was divided into round or oval (partial contact with interlobar ssures), lobular (preferred when lobular characteristics were obvious), and polygonal (16). The solid ratio refers to the maximum solid diameter divided by the maximum nodule diameter (Fig. 1A). The contact length measured as curve between the tumor and the interlobar pleura was recorded as the nodule interface length. The interface length ratio was calculated by dividing the interface length by the nodule diameter (Fig. 1B). If the solid part of the nodule contacted the pleura, the length of the contact was measured as the solid interface length. For the continuous variables (solid diameter, solid ratio, et al.), the mean value measured by two doctors was used. For categorical variable (shape, Pleural indentation, et al.), the differences were resolved by a third physician (18 years' experience).

Pathological Analysis
Resected tumor and contact pleura specimens of all 126 patients were evaluated microscopically under hematoxylin-eosin (HE) staining. When the pleural elastic layer of the affected visceral layer was unclear, speci c elastic ber staining or double staining (immunohistochemical staining + elastic ber staining) was performed. Tumor invasion degree was determined under a microscope according to modi ed Hammar grading criteria (1). In the 8th edition of the TNM staging system for NSCLC, VPI is de ned as an invasion beyond the pleural elastic layer of the visceral layer (17). In our retrospectively analyzed sample, speci c elastic ber staining was performed in 33 cases and double staining was performed in 8 cases.

Statistics
All data were analyzed using SPSS 22.0 software (IBM, Armonk, NY, USA) and Medcalc statistical software (Medcalc 19, https://www.medcalc.org). The independent samples t-test or Mann-Whitney U-test was used to compare the continuous variables (represented as mean ± standard deviation [SD]) and chisquare test or Fisher's exact test (number and percentage (%)) to compare the categorical variables between different groups. Univariable and multivariate logistic regression (stepwise regression) was applied to identify the factors signi cantly correlated with VPI. Area under curve (AUC) of Receiver operating curve (ROC) and interactive dot diagram were used to evaluate the diagnostic performance of the risk factors. A two-sided P-value of < 0.05 was considered statistically signi cant.

Results
Baseline characteristics of the U-HRCT ndings U-HRCT and three-dimensional orthogonal post-processing with interlobar ssure could better display the GGN and its relationship with interlobar ssure than routine CT scan. Radiation dose from U-HRCT was quantitatively characterized by CT dose index per unit volume (CTDIvol) and Dose length product (DLP), the CTDIvol was 19.1 ± 3.55mGy and DLP was 184.79 ± 13.74 mGy.cm in the study.
A total of 126 patients meeting the inclusion criteria were included in the analysis. The baseline characteristics are shown in Table 1. The mean age was 55 ± 11 years, with 44 males (35%) and 82 females (65%), and the average nodule size was 1.4cm. The sample included 55 cases with pGGN and 71 cases with mGGN. Regarding the lesion type, there were 3 cases of atypical adenomatous hyperplasia (AAH), 41 cases of adenocarcinoma in situ (AIS), 26 cases of microinvasive adenocarcinoma (MIA), and 56 cases of invasive adenocarcinoma (IA). All NSCLCs with VPI were IAs, and postoperative pathology showed that none of the patients had lymph node metastasis. Morphologically, most of these NSCLCs were round or oval (70.5%). On CT scans, most of the NSCLCs showed oblique interlobar ssure (84%), and CT values in VPI patients were greater than those without VPI. None of the above-mentioned parameters, except age (P = 0.017), were signi cantly different (P > 0.05) between patients with and without VPI.  (Fig. 2), and 21 (38%) had pleural indentation ( Table 2). Univariate and multivariate logistic regression of mGGN The mGGN group was analyzed further as shown in Table 3.  19.73mm, the solid diameters were 5.79 and 11.73mm, the solid ratio was 0.35 and 0.56, the interface length was 9.46 and 14mm, and the density was − 401.96 and − 262.67, respectively. The nodules size, solid diameter, solid ratio and the interface length in VPI positive group all was larger than the VPI negative group, and the difference was statistically signi cant (P < 0.05). In all patients with VPI, the GGNs had solid contact with the pleura (Fig.   3, 4). Compared to the cases without VPI, the proportion of pleural indentation (93.33% vs. 67.86%) as well as the proportion of pleural thickening (80% vs. 64.29%) was higher in the cases with VPI.
Univariate and multivariate analyses were performed to identify the risk factors for pleural in ltration.
Univariate analysis revealed that density, nodule size, interface length, solid diameter, and solid ratio were independent risk factors. The odds ratio (OR) values of each risk factor are shown in Table 4. After multivariate logistic regression analysis, two risk factors remained signi cant, namely, the solid ratio and solid diameter, with the OR values being 1.275 and 1.139, respectively. The areas under curve (AUC), sensitivities, and speci cities of solid diameter and solid ratio for predicting VPI were 0.803, 80%, and 75% and 0.807, 80%, and 80.36%, respectively ( Fig. 5 and Table 5). There was no statistical difference between the diagnostic e cacies of the two variables by the Delong test.

Discussion
As early as 1958, Brewer recognized that VPI in lung cancer was an important prognostic factor (18), especially in early stage NSCLC. In the mid-1970s, VPI was used as a speci c description in the TNM Classi cation of Malignant Tumors (TMN) set forth by the Union for International Cancer Control, which is now in its 8th edition (17). In the 7th edition, NSCLC with a diameter ≤ 3cm and VPI was upgraded from Tla or Tlb stage to T2a stage, and the degree of VPI was classi ed (1). VPI was de ned as tumor invasion beyond the pleural elastic layer of the visceral layer (PL1) and invasion of the pleural surface of the visceral layer (PL2). The VPI classi cation changes the clinician's assessment of staging and prognosis, thus affecting the treatment strategy.
According to the TNM criteria, local invasion of a tumor in the pleura without one in the visceral layer should be classi ed as T2 (18). It has also been suggested that intrapulmonary pleural invasion in the interlobar region should be classi ed as T3 (19). Most literature reports are of T2 and T3 lung cancer, and there are few studies on T1 cases. No taxonomic study has been conducted on T1 GGN. When the tumor is in contact with the pleura, VPI is more likely to exist. As an important method to predict VPI, measuring and calculating the indicators on the corresponding CT images is highly valuable (20). Considering the radiation dose, conventional CT and U-HRCT were not performed simultaneously, only underwent U-HRCT scan were enrolled.
Based on previous research ndings, compared the accuracies of U-HRCT and conventional CT for detecting morphological features on CT in the evaluation of GGN, The result shown that the detection rate was higher for U-HRCT target scanning than for conventional CT target reconstruction, and this improvement signi cantly enhanced the diagnostic accuracy of early lung adenocarcinoma (21). In U-HRCT scans, a large matrix size maintained the spatial resolution and improved the image quality and assessment of lung diseases, despite an increase in image noise, when compared to a 512 matrix size, which can provide more detailed lung anatomy and pathology information for the evaluation of lung diseases (22,23). Compared with the surgical pathological ndings, U-HRCT had a relative higher diagnostic accuracy of early lung adenocarcinoma than conventional CT. the U-HRCT prototype scanner provides a better image quality of subsolid nodules and contributes signi cantly to reduce the patients' follow-up period (24). So, in this study, we adopted U-HRCT to obtain submillimeter images. Through three-dimensional orthogonal post-processing, the morphology, integrity, and features of the interlobar ssure can be displayed from different perspectives, and this information can be obtained accurately before surgery, which is more advantageous than a conventional CT scan; Total scanning dose of U-HRCT is about 180 mGy.cm, it was lower than that of conventional CT scan. So, we believe that HRCT will be a routine scan for further accurate diagnosis.
In this study, 17 patients (13.5%) presented incomplete interlobar ssures, but no cases of interlobular ssures invading adjacent lobes with VPI were observed, which may be related to the low invasiveness of T1 GGN. none of the 55 pGGN contacting the pleural surface of the interlobar ssure showed VPI. The results were consistent with Hattori et al. (25), who reported that in NSCLCs with pGGN resection of less than 3cm, there was no VPI. Ahn et al. (9) also did not nd VPI in the pGGN group. But some studies (26) show that VPI was often seen in pGGN adenocarcinoma. In our study, 4 cases of in ltrating adenocarcinoma were acini subtype, 3 had pleural thickening, 1 had pleural indentation, and 2 were cases of pathological spread through airway spaces. Pathological HE staining and elastic ber dyeing did not reveal VPI, whereas pleural thickening and in ammatory cell in ltration was revealed in the elastic ber layer, with no tumor cells beyond the elastic layer. All above support the fact that pGGNs are less pathologically invasive (27,28) and cannot penetrate the thick elastic layer in early lung adenocarcinoma.
As for mGGN contacting the pleural surface, this study showed that there were signi cant differences between the cases with and without VPI in terms of density, contacting solid, solid diameter, and solid ratio. Some studies have con rmed the correlation between CT ndings and pathological results (29), with the degree of solid component being an important invasive factor that was related to poor prognosis (30). In this study, all the positive VPI groups had solid contact with the pleura, which suggested that the solid part was an important invasive component of GGN for lung adenocarcinoma, but there was no signi cant difference in the interface length (P > 0.05). Partial solid nodules with a solid ratio greater than 0.5 were more likely to invade the visceral pleura. Ohde et al. (31) reported that in T1 NSCLC, a solid ratio greater than 50% was related to invasion, which was consistent with our study. In our study, solid diameter > 6mm and solid ratio > 38% were revealed to be independent predictors of VPI (P < 0.05). The AUC of the two factors for VPI diagnosis reached 0.8, and the sensitivity for both was 80%, which could provide an important reference value for preoperative diagnosis of visceral pleural in ltration. In contrast to the study by Ahn et al. (13), there was no signi cant difference in nodule size, interface length, interface length ratio, and contacting solid in our study (P > 0.05), which may be related to differences in nodule types and sample size. Our study has several limitations. First, our study is a retrospective study, which limits our analytical validity. Second, our study had a small sample size. Similar and larger multi-center studies are needed to con rm our current conclusions. Third, we did not evaluate inter-and intra-observer reproducibility, which may in uence the CT features evaluation. The relationship between VPI and long-term outcomes such as survival, recurrence, and metastasis were not analyzed in this study. However, this study is still in progress, and relevant data, such as progression-free survival and overall survival, are in further follow-up.
In conclusion, U-HRCT and three-dimensional orthogonal post-processing technology were used to improve the spatial resolution and contrast resolution of CT images and to more accurately evaluate GGN and its relationship with interlobar ssure. This was helpful in predicting VPI. When GGN with a diameter of ≤ 3cm was in contact with the interlobar ssure, the ground glass nodule component was less invasive. There was no VPI in patients with pGGN. Regarding mGGN, a solid diameter > 6mm and a solid ratio > 38% observed in CT evaluation could be used as an independent predictor of VPI in early stage NSCLC.
Declarations 33. Gallagher B, Urbanski SJ. The signi cance of pleural elastica invasion by lung carcinomas. Hum Pathol 1990, 21:512-7. Figure 1 Parameter measurements methods. (A) Maximum diameter of ground glass nodules (GGN) (p) was measured to determine the nodule size as well as the solid diameter of mixed GGN (q) and record solid ratio (q/p). (B) The interface length (r) was recorded, and the interface length ratio (r/p) was calculated. If the solid part of the nodule was in contact with the interlobar pleura, the interface length was measured (s).    Example of a patient with mixed ground glass nodules (mGGN) without VPI. A 55-year-old man with invasive adenocarcinoma without VPI. Ultra-high-resolution computed tomography showed mGGN in the right lower lobe. mGGN was close to the interlobar ssure (▲): semicircular, lobular, contacting solid, pleural thickening, and indentation (↑ A, B), nodule size, 28mm, solid diameter, 6mm, solid ratio, 21%, and length interface ratio, 89%. (C) Pathological elastic ber stain (100X). Dark blue bands in the visceral pleura elastic layer (↑), fold thickening zone multilayered elastic bers, and no tumor cells breaking the elastic layer (▲).

Figure 5
The diagnostic e cacy of solid diameter and solid ratio. A. ROC of solid ratio with AUC and P value, B.
Interactive dot diagram of solid ratio with cutoff value, Sensitivity and Speci city, C. ROC of solid