Endobronchial Ultrasound Features for Evaluation of Mediastinal and Hilar Lymph Nodes

Background. The aim of the study was to evaluate the utility of Endobronchial ultrasound (EBUS) features included elastography and B-mode features for differentiating malignant from benign lymph nodes(LNs). Methods. 84 patients with 151 enlarged mediastinal and hilar LNs underwent endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) were involved in the retrospective study from 1 January 2019 to 31 December 2019. Scores of EBUS elastography, EBUS B-mode features and nal pathological results were recorded. Receiver operating characteristics, univariate and multivariate logistic regression analysis were used to evaluate the diagnostic yield of elastography and B-mode features for malignant LNs. Results. Total 84 patients of 151 LNs were enrolled in the single center retrospective study, which included 108 malignant nodes and 43 benign nodes obtained from 59/25 patients respectively. EBUS elastography score 4-5 differentiated malignant LNs from benign nodes with sensitivity, specicity, positive predictive value, negative predictive value and diagnostic accuracy 85.05%, 77.27%, 90.10%, 71.0% and 82.23% respectively. EBUS B-mode features round shape, heterogenenous echogenicity and absence of CHS showed statistical diagnostic yield by multivariate logistic analysis. ROC analysis suggested the combined AUC for elastography, round shape, absence of CHS and hetergeneous echogenicity was 0.849. Conclusions. EBUS features are effective for differentiating between benign and malignant LNs. This study was approved Ethics Committee medical college Yantai Hospital 2014-111).


Introduction
EBUS-TBNA is a less invasive procedure for diagnosing mediastinal and hilar lymphadenopathy than mediastinoscopy since mid 1990s [1][2][3]. LNs metastasis of lung cancer is the most common cause of intrathoracic lymphadenopathy clinically. Early diagnosis and staging of suspected metastasis mediastinal or hilar LNs before surgical procedure can reduce the rate of post-operative upstaging and is also suggested by National Comprehensive Cancer Network (NCCN) for patients with lung cancer [4][5][6]. Advanced stage lung cancer represents approximately 75% of patients when diagnosed [7], for whom the 5-year survival rate is no more than 5% on stage IIIB/IV [8], further chemotherapy or molecular targeted therapy also needs more specimens for precise genetic testing. The common benign intrathoracic lymphadenophy includes tuberculosis (TB), sarcoidosis and in ammation. Since the diseases above mentioned require completely different therapeutic interventions, precise diagnosis particularly the differentiation between malignant and benign LNs wound be the most important goal to achieve. EBUS-TBNA was demonstrated to be a effective technique for mediastinal node sampling [9][10][11].
During the real-time procedures for EBUS-TBNA, the selection of target LNs is based on the characteristics of sonographic features. Nearly 10 sonographic images were mentioned to differentiating malignant LNs including the size, shape, calci cation status, present or absent of central hilar structure (CHS), necrosis sign, margin, echogenicity, clustered formation and so on [12][13][14]. However, no single feature aboved has satisfactory predictive value [15]. During the examination of EBUS, elastography generates images re ective of tissue stiffness to endosonographer, which was originally used as diagnostic imaging tool in breast cancer [16]. The images display through local compression and vibration of bronchoscope.
Previous studies which utilized qualitative method of elastography including three types based on color distribution (Type 1, predominantly non-blue; Type 2, green, yellow and red; Type 3, predominantly blue) showed diagnostic accuracy rate for mediastinal and hilar LNs range from 88.7-96.7%, and with sensitivity and speci city greater than 80% [17,18]. At the same time, Sun et al [19] introduced 5 scores of EBUS elastography in EBUS (Score 1, scatter soft, mixed green-yellow-red; Score 2, homogeneous soft, predominantly green; Score 3, intermediate, mixed blue-green-yellow-red; Score 4, scattered hard, mixed blue-green; Score 5, homogeneous hard, predominantly blue). By means of this method, the accuracy, sensitivity and speci city to predict metastatic LNs were 83.82%, 85.71% and 83.82% respectively.
However, there were lots of false-negative cases in clinical practice, numerous predicted charateristics may led di culties in clinical practice and was time comsuming during the real time procedure. Few study had focus on a concise model to evaluate LNs will be needed in EBUS-TBNA with short time. The purpose of this study was to investigate the diagnostic yield effect of both elastography and the most common EBUS B-mode features, and to evaluate the value for differentiating malignant from benign LNs. EBUS characteristics of B-mode and elatography,EBUS-TBNA procedure After local anesthesia to the pharynx with 4% lidocaine, the patients were anesthetized consciously with intravenous midazolam. The equipment (BF-UC260FW, Olympus Ltd. Tokyo, Japan) with a linear probe on its tip was used to perform EBUS-TBNA. Images was generated by the linear ultrasonic processor. Water balloon was used to improve coupling to the airway wall, the probe attached to the airway wall with proper pressure. EBUS-TBNA was proceeded in LNs with a short axis of at least 5mm [20]. Before elastography observation, conventional B-mode observation was initially performed. Previous studies showed characteristics of EBUS B-mode predicted probable malignant LNs. LNs featured at least one factor wound be chosen as the target for EBUS-TBNA and LN classi cation was based on the newest international staging system [21]. Among these round shape, a size of short axis more than 10mm, heterogeneous echogenicity, absence of CHS, distinct margin were considered in the present study (Representative ndings of EBUS B-mode features were described in Fig. 1). Images of each LN were recorded and examined by independent sonographer (Z-SC) to con rm whether the images covered maximum diameter of the LN. EBUS elastography images were nally generated by vascular pulsations and respiratory movement, the region of interest (ROI) was selected avoiding vessels. Refer to the sorting means of Sun et al, elstography was classi ed into 5 scores mentioned above. Video of each LN were recorded and the nal agreement on the qualitative analysis of elastography should be reached by three independent reviews (P-LL, F-YQ, L-XP). Only accurate diagnosed LNs were included in the study. The procedure was performed by consant bronchoscopist (Y-PF) (Representative lymph nodes on EBUS elastography scored 1-5, CT images and pathology results were described in Fig. 2

. A-E).
After these non-invasive evaluation, TBNA was carried out using 22-gauge needle (NA-201SX-4022, Olympus Optical CO Ltd.), three to four needles aspirations were suggested on the premise of ensuring the sample quality [22,23]. On-site cytological evaluation was not performed. Histology, cytology specimens were all accepted. The nal diagnosis was based on the pathological examination.

Data collection and statistical analysis
Data collection included general information about patients, EBUS B-mode features and elastography score per LN. Final pathological diagnosis was also reviewed. A follow-up of at least 6 months was suggested for all patients and LNs, the follow-up measures involved common bronchoscopy, CT-guided transthoracic needle aspiration,mediastinoscopy, thoracotomy and clinical results.
The participants' characteristics such like age was presented as means and standard deviations (SDs) while other categorical variables were presented as counts and percentage. Receiver operating characteristic (ROC) analysis was performed in order to nd the best factors for predicting malignant LNs by true and false-positive rates. The in uence of some factors including EBUS B-mode features and elastography on having malignant LNs in the patients were calculated respectively using univariate and multivariate logistic regression analyses and presented as odds ratios (ORs) and corresponding 95% con dence intervals (CIs). All statistical analyses were performed using SAS for windows version 9.4 (SAS Institute, Cary, NC, USA) and R (3.4.3). A two-sided P < 0.05 was considered as statistically signi cant.

Patients and LNs
A total 84 patients of 151 LNs were involved in our single center retrospective study, which included 108 malignant nodes and 43 benign nodes obtained from 59/25 patients respectively. The baseline characteristics of all LNs were summarized in Table 1. There were no severe associated complications of EBUS-TBNA. Con rmed by earlier diagnosis and later follow-up, diagnostic accuracy of EBUS-TBNA on malignant LNs was 86.11% (93 of 108 LNs). The remaining 15 LNs from 7 patients were proven as falsenegative cases. Among 2 adenocarcinoma patients (5 LNs), 1 lymphoma patient (3 LNs) were nally diagnosed through mediastinoscope; 2 squamous carcinomapatients (3 LNs   According to EBUS elastography, LNs with score1 were diagnosed as benign (2 LNs), score 2 were benign in 80.0% and malignant in 20.0%, score 3 were benign 60.6% and malignant in 39.4%, LNs diagnosed malignant with score 4 were 89.7% and score 5 were 90.5% respectively. ROC analysis veri ed EBUS elastography as the best factor for predicting the malignant of LNs with AUC 0.635 than other ve single EBUS B-mode factors. Score 1-3 were suggested to indicate benign nodes and score 4-5 malignance. Sensitivity, speci city, PPV, negative predictive value (NPV) and accuracy of elastography scored 4-5 in distinguishing malignant LNs from benign were 85.05%, 77.27%, 90.10%, 71.0% and 82.23% respectively. Correlation between EBUS elastography sores and these ve EBUS B-mode features was showed in Table  2. Diagnostic yield of elastography and B-mode features to predict malignance were showed in Table 3.
Further univariate logistic analysis showed elastography and three EBUS B-mode features (round shape, heterogeneous echogenicity, absence of CHS) above mentioned are all risk factors for predicting malignant LNs (P < 0.05). Diagnostic yield of each factor with malignant LNs in the combined multicategory variable model was expressed by multivariate logistic analysis, round shape, heterogenenous echogenicity and absence of CHS showed statistical diagnostic yield with OR=2.727, 2.567 and 4.783 respectively. Elastography score 4-5 showed signi cant difference compared with score2(p=0.0111, 0.0017), while score 3 showed no statistical difference compared with score 2 (p=0.5078) for differentiating malignant from benign LNs. Score 1 was excluded from the study because of small sample size. While distinct margin and short axis showed no statistical difference for predicting malignant LNs in the combined model (Table.4

Discussion
EBUS-TBNA is a novel and minimally invasive methods for sampling mediastinal and hilar LNs using real-time guidance. While mediastinoscopy is considered gold standard for diagnosing and staging of mediastinal LNs, it is more invasive and costly than EBUS with a morbidity of 2% [24]. Given high rate of sensitivity and speci city, EBUS-TBNA was suggested as desirable alternative to mediastinoscopy, and the diagnostic yield of EBUS-TBNA for mediastinal LNs was 89.57% in our study, consistent with other recent reported studies [25][26][27]. The high diagnosis rate is based on the appropriate selection of sampling target. EBUS elastography is a novel sonographical technique to react tissue compressibility. For malignant tissue was in ltrated and hardly deformed, elastography has feasibility to distinguish between malignant and benign tissue. Thus, elastography signal displayed in colored image is the feedback of tissue hardness. In the colored images, area of stiffness appears blue and green, area of soft appears others. Based on the color distribution, EBUS elastography was classi ed into 3 or 5 categories by different scholars and reported to predict malignant LNs mentioned above. Our study involved the maximum size sample so far using 5 scores of EBUS elastography. Our data revealed 77.27% benign LNs scored 1-3, 85.05% malignant LNs scored 4-5, which suggested that EBUS elastography score1-3 is mainly characteristic for benign, and elastography score 4-5 for malignant LNs. Veri ed by multivariate logistic regression analysis, elastography score 4 and 5 showed 6.109, 10.905 times than score 2 for predicting malignance. Based on the results score 2 were benign in 80.0% and score 3 were benign 60.6%, it seemed score 2 was better for predicting benign LNs, however, multivariate logistic regression analysis showed on signi cant difference between elastography 2 and 3(P=0.5078). Score 1 was excluded from the study because of the small size sample to avoid error. Thus, EBUS elastography score 4-5 were effective predictors of metastatic LNs. However, there were some false-negative/positive cases in our study, in which many malignant LNs were soft, and benign LNs were hard, for some malignant tissue mainly structured by loose tissue, part of benign LNs existed brotic or epithelioid granulomas changes. Data in our study showed, false-negative cases score 2 contained 75% squamous cancer, 25% small cell lung cancer, score 3 contained 14.3% small cell lung cancer, 76.2% non-small cell lung cancer(NSCLC) and 9.5% lymphoma respectively. False-positive cases of elastography score 4 and 5 consist of in ammation (33.3%, 0%), sarcoidosis (16.7%, 50%) and tuberculosis (50%, 50%) respectively. It was implied that sarcoidosis, tuberculosis and small cell lung cancer more harder than diseases of same nature, however, limited to small sample size, we didn't analyze it. Meanwhile, elastography images obtained by local compression and variation of bronchoscope, which was affected by the relative hardness of ROI surrounding areas, thus improper delineation area of ROI may exaggerate or reduce score and lead errors. Besides the qualitative score of EBUS elastography, quantitative methods including mean gray value, strain ratio were also introduced to characterize target lesion [15,28,29], strain ratio was reported to distinguish malignant from benign mediastinal LNs with high accuracy of 86.25% by Rozman et al. Meanwhile, mean gray value calculated by software showed good sensitivity and speci city. For the tedious process for providing information of targeted LNs, which may led to more di culty during the short time procedures of EBUS-TBNA, we did not use these methods in our research.
In our present study, we also evaluated the diagnostic value of EBUS B-mode features for distinguishing malignant from benign mediastinal LNs. The diagnostic yield of the ve EBUS B-mode features was mentioned above, data analyzed by univariate logistic analysis showed only three features had potential for predicting malignant LNs, which did not fully consistent with previous studies [30]. The probability to be a malignant LN was 2.796, 2.457, 2.97 times for round shape, heterogeneous echogenicity, CHS absence VS presence respectively and p value all less than 0.05. In the combined model analyzed by multivariate logistic analysis, round shape, heterogeneous echogenicity, absence of CHS showed statistically different diagnostic yield for malignance, however, diagnostic yield of short axis and distinct margin showed non effective with low sensitivity and speci city. That is to say, round shape, heterogeneous echogenicity, absence of CHS showed a good diagnostic accuracy in our present study, except for short axis and distinct margin. Our present study suggested elastography score 4-5 combined with these three EBUS B-mode features would be good help to selection of LNs for TBNA.
In conclusion, this retrospective study involved 84 patients of 151 LNs, we evaluated EBUS elastography and ve B-mode features for predicting the malignance of mediastinal and hilar LNs. EBUS qualitiative elastography were effective for differentiating malignant from benign LNs, which wound offer guide information in EBUS sampling. For the pathology is the gold standard for diagnosis, elastography could hardly replace TBNA, but the systemic score of elastography and B-mode features wound help the operators for sampling suspected LNs or suspected area in the LNs with the highest probable yield of malignance.
There were still some limitations in our present study, rstly, the sample size was small relatively, larger sample study wound be needed to verify the predictable yield of elastography for speci c disease in the next study. Secondly, the quantitative elastography score may agreed subjectively, objective and simple score model may be needed for diagnosing malignant LNs. Finally, we expected prospective study to con rm our ndings in the future.

Abbreviations
All abbreviations were list in the manuscript.   Representative morphologic ndings of EBUS images.Round shape,short axis 10mm,heterogeneous echogenicity,absence of CHS,distince margin were described in the rst line.Images with opposite in the second line.

Figure 2
A-E. Representative lymph nodes on EBUS elastography scored 1-5,CT images and pathology results.