Neuroradiology/Head and Neck Imaging Original Research

Transcription

1 Neuroradiology/Head and Neck Imaging Original Research Ryu et al. Cervical Lymph Node Imaging Reporting and Data System Neuroradiology/Head and Neck Imaging Original Research Kyeong Hwa Ryu 1, 2 Kwang Hwi Lee 1, 3 JiHwa Ryu 1 Hye Jin Baek 4 Suk Jung Kim 1 Hyun Kyung Jung 1 Sung Mok Kim 5 Ryu KH, Lee KH, Ryu J, et al. Keywords: interobserver agreement, lymph node, real-time elastography, structured reporting system, ultrasound DOI: /AJR Received August 8, 2015; accepted after revision December 2, Department of Radiology, Inje University College of Medicine, Haeundae Paik Hospital, Haeundae-ro 875, Haeundae-gu, Busan , Republic of Korea. Address correspondence to K. H. Lee (lkh770429@naver.com). 2 Department of Radiology and Research Institute of Radiology, University of Ulsan College of Medicine, Asan Medical Center, Seoul, Republic of Korea. 3 Department of Radiology, Newoori Namsan Hospital, Busan, Republic of Korea. 4 Department of Radiology, Gyeongsang National University School of Medicine, Gyeongsang National University Changwon Hospital, Changwon, Republic of Korea. 5 Department of Radiology, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea. AJR 2016; 206: X/16/ American Roentgen Ray Society Cervical Lymph Node Imaging Reporting and Data System for Ultrasound of Cervical Lymphadenopathy: A Pilot Study OBJECTIVE. The objective of our study was to compare ultrasound (US) and real-time elastography (RTE) features of benign and malignant cervical lymphadenopathies and propose a structured reporting system for lymph nodes. MATERIALS AND METHODS. The study population for this retrospective study consisted of 291 consecutive patients who underwent US-guided biopsies for cervical lymphadenopathy between 2013 and The following imaging features were analyzed: shape, margin, echogenicity, echogenic hilum, gross necrosis, calcification, matting, intranodal vascular pattern, elasticity scores (four categories), and strain ratio. A score was assigned for each significant factor from a logistic regression analysis and was multiplied by the beta coefficient. The fitted probability of malignancy was calculated. The risk of malignancy was determined on the basis of the number of suspicious features. Interobserver agreement of the imaging features was retrospectively analyzed using a coefficient of interrater agreement. RESULTS. The imaging features that were significantly associated with malignant lymphadenopathy were round shape, noncircumscribed margin, hyperechogenicity, absence of hilum, gross necrosis, calcification, peripheral or mixed vascularity, high elasticity scores, and high level of strain ratio (p < 0.05). The fitted probability and risk of malignancy increased as the number of suspicious features increased. The risk of malignancy according to the Cervical Lymph Node Imaging Reporting and Data System categories was as follows: category 1, 3.3%; category 2, 10.9%; category 3, 26.7%; category 4, %; and category 5, %. An analysis of the overall interobserver agreement revealed that interobserver agreement was moderate to good. CONCLUSION. We propose the Cervical Lymph Node Imaging Reporting and Data System, which uses the number of suspicious US and RTE features to assess the risk of malignancy in cervical lymph nodes. C ervical lymphadenopathy is frequently involved in various disease entities, including tuberculosis, lymphoma, and metastasis [1]. The diagnosis of malignant cervical lymphadenopathy is crucial for determining therapeutic strategies for patients with a suspicion of malignancy arising from different organs and for preoperative staging of patients with primary malignancies originating in the head and neck [2]. Ultrasound (US) is considered a primary diagnostic modality to identify and characterize cervical lymph nodes because US is cost-effective and involves no radiation hazard. US reveals a higher sensitivity, up to 96.8%, than other imaging modalities for detecting malignant cervical lymphadenopathy in patients with head and neck malignancies [3, 4]. Despite these strengths, the application of US is limited because US depends on the operator s experience and subjective impression [5]. The examiner s description of his or her subjective impression of the US findings can result in issues related to the transmission of information from the examiner to the physician who then determines the patient s management. A standardized reporting system for US examinations of lymph nodes is necessary to overcome potential communication problems between examiners and physicians that could confuse physicians who receive radiologic reports describing US features. In breast radiology, this problem was solved by the clinical application of BI-RADS, which was developed by the American College of Radiology [6]. BI-RADS was first established to standardize the reporting system 1286 AJR:206, June 2016

2 Cervical Lymph Node Imaging Reporting and Data System Patients with cervical lymphadenopathy who underwent ultrasound-guided procedures during the study period (n = 399) Benign lymphadenopathy (n = 182) Reactive hyperplasia (n = 166) Tuberculous lymphadenitis (n = 16) of mammography and has subsequently been adopted for breast US. BI-RADS includes management recommendations and final assessment categories for each breast lesion type. Recently, structured reporting systems of other imaging modalities have been investigated in various organs including the liver, ovary, prostate, and lung. However, a structured reporting system for lymph nodes has not been reported to our knowledge. Real-time elastography (RTE) is considered a complementary diagnostic technique to conventional US for cervical lymph node imaging [7, 8]. RTE is used to detect malignant lymphadenopathy and to select suspicious lymph nodes for biopsy. Although the efficiency of RTE in differentiating malignancies from benign diseases in many organs has been investigated, structured reporting systems other than BI-RADS [9] do not include the application of RTE. This study aimed to compare the US and RTE features of benign and malignant cervical lymphadenopathies, to clarify the risk stratification of malignant lymphadenopathy, and to propose a structured imaging reporting system for reporting the US and RTE features of cervical lymph nodes called the Cervical Lymph Node Imaging Reporting and Data System (CLN-RADS). Patients with a final diagnosis (n = 291) Exclusion (n = 108) Nondiagnostic or indeterminate results on cytopathologic examination (n = 17) Insufficient follow-up in cases with a benign cytopathologic result (n = 45) Suboptimal image quality (n = 46) Malignant lymphadenopathy (n = 109) Metastasis (n = 89) Thyroid cancer (n = 36) Lung cancer (n = 29) Head and neck cancer (n = 8) Stomach cancer (n = 3) Other cancers (n = 12) Metastasis of unknown origin (n = 1) Lymphoma (n = 20) Fig. 1 Flowchart shows selection of study population and histopathologic diagnoses of cervical lymphadenopathies. Materials and Methods Study Population This retrospective study was approved by our institutional review board. The requirement to obtain informed consent was waived. Between March 2013 and October 2014, 2720 US examinations were performed in a single institution to survey cervical lymphadenopathy. Of these cases, 399 patients underwent US-guided fine-needle aspiration (FNA, n = 137) or core needle biopsy (CNB, n = 262) because lymph nodes revealed malignant features on conventional US or because other infectious causes such as tuberculous lymphadenitis or suppurative lymphadenitis were clinically suspected. The inclusion criteria of this study were as follows: conclusive result of either benign or malignant lymphadenopathies on cytopathologic examination, clinical or radiologic follow-up of at least 6 months in cases with a benign result on cytopathologic examination, and optimal image quality of US and RTE examinations. Of the 399 patients, 108 patients were excluded for the following reasons: nondiagnostic or indeterminate results on cytopathologic examination (n = 17), insufficient follow-up period in cases with a benign cytopathologic result (n = 45), and suboptimal image quality (n = 46). Finally, 291 patients (male-female ratio, 108:183; mean age, 49.7 years; age range, 5 91 years) were enrolled in this study (Fig. 1). Ultrasound Examination, Fine-Needle Aspiration, and Core Needle Biopsy Neck US examinations were performed by two board-certified radiologists who had 11 and 4 years of experience in head and neck imaging using a 7-15 MHz linear-array transducer (Ascendus, Hitachi Aloka Medical). After surveying the entire neck using gray-scale US, the radiologists selected a target lymph node, one that either presented malignant features or the largest one on gray-scale US; they then performed power Doppler US and RTE of the target lesion. US-guided FNA or CNB of the target lesion was sequentially performed by one of the two radiologists. US-guided FNA was performed using a freehand technique and direct US visualization. A 21- or 23-gauge needle tip attached to a 2- or 10-mL syringe was advanced in the target lymph node below the center of the probe and was advanced to the lesion along the image plane with a to-and-fro technique. Aspirated materials were placed on labeled glass slides, smeared, and fixed in 95% ethyl alcohol. US-guided CNB was performed using an 18-gauge, 1.1- or 1.6-cm excursion, double-action spring-activated needle (Acecut, TSK Laboratory). The core needle was inserted toward the target lymph node with a freehand technique. The stylet and cutting cannula were fired sequentially after the needle tip was advanced to the central or peripheral portion of the lymph node. At least two cores were obtained in each session. The biopsy specimens were fixed in formalin solution. After the US-guided procedures, all specimens were sent to the pathology department and assessed by one of four experienced pathologists. After the procedures, all patients were observed for minutes while firm local compression was applied to the biopsy site. Ultrasound Features All images were analyzed retrospectively by the same two experienced radiologists mentioned earlier. They analyzed the images in consensus in a blinded fashion. The interval between imaging acquisition and retrospective imaging analysis was at least 3 months. The readers were instructed to capture the target lymph node with an annotation on the representative US images. After 6 months, each reader independently reinterpreted the imaging features of the target lymph node. AJR:206, June

3 Ryu et al. The following US features of cervical lymph nodes were analyzed: shape (oval or round), margin (circumscribed or noncircumscribed), echogenicity (hypoechoic, isoechoic, or hyperechoic), echogenic hilum (present or absent), gross necrosis (present or absent), calcification (present or absent), matting (present or absent), and intranodal vascular pattern (hilar, peripheral, mixed, or avascular); the shape was determined by the shortest-to-longest axis ratio in the longitudinal scan of the lymph node. The margin was determined by the sharpness of the nodal border. The echogenicity of the lymph nodes was determined by comparison with adjacent muscles. Echogenic hilum was considered a linear or elliptic hyperechoic central structure connected to adjacent perinodal fat. Gross necrosis included cystic and coagulation necrosis. Cystic necrosis presented as irregular prominent hypoechoic areas; in contrast, coagulation necrosis presented as less echogenic than a hilum without a posterior acoustic shadow and discontinuous with the surrounding fat. Punctate hyperechoic foci with or without posterior acoustic shadows were considered calcifications. Matting was defined by clumping of multiple lymph nodes with abnormal features or perinodal hypoechoic lesions with disruption of the nodal border. Based on power Doppler examination, the intranodal vascular pattern was determined by the distribution of vascular flow signals within the lymph node. Hilar vascularity was defined as central flow signals or flow signals branching radially from the hilar structure. In contrast, peripheral A vascularity was defined as flow signal along the periphery or capsular portion of the lymph node. Mixed vascularity presented as a combination of hilar and peripheral vascular patterns. An avascular pattern was defined as no vascular signal. Real-Time Elastography After conventional US, RTE was subsequently performed using the strain method. The target lesion was manually compressed by the transducer under light repetitive pressure. When the color image was stable, an ROI was selected. The ROI included subcutaneous fat on the top and posterior neck muscles on the bottom and was chosen to encompass as much soft tissue around the target lymph node as possible, while avoiding hard tissues including bone that might disturb the appro- C D Fig year-old man with metastatic cervical lymphadenopathy originating from lung cancer. Because lymph node has three suspicious features on ultrasound (US) and real-time elastography (RTE), it was assigned Cervical Lymph Node Imaging Reporting and Data System category 4. A and B, On gray-scale US images in transverse plane (A) and longitudinal plane (B), 3-cm lymph node (cursors, B) shows noncircumscribed margin and absence of echogenic hilum. Matting (arrowheads) is present. C, On power Doppler US image, lymph node shows mixed vascular patterns that include hilar and peripheral vascularity (arrows). D, On RTE image, lymph node shows blue-dominant pattern (i.e., stiff component = 50% to < 90% of target lesion), which is regarded as elasticity score of 3. RTE assessment is suspicious for malignancy. Strain ratio is B 1288 AJR:206, June 2016

4 Cervical Lymph Node Imaging Reporting and Data System priate analysis of the relative stiffness of the target lymph node [10, 11] (Fig. 2D). The RTE strain images were displayed on a monitor through 256-color mapping. The red area was the area with the greatest strain (softest component); the green area, average strain (intermediate component); and the blue area, no strain (stiffest component). The elasticity score was categorized using the following 4-point scale based on a study by Ying et al. [7]: score of 1, stiff component is less than 10% of target lesion; 2, stiff component is 10% to less than 50% of target lesion; 3, stiff component is 50% to less than 90% of target lesion; and 4, stiff component is 90% or greater of target lesion. The RTE assessment was categorized as probably benign (elasticity scores of 1 and 2) and as suspicious for malignancy (elasticity scores of 3 and 4). For evaluation of the strain ratio, the average strain value was obtained from four continuous frames of strain images. The average strain ratio was calculated as the mean strain of adjacent muscle divided by the mean strain of the target lymph node. Statistical Analysis Data were tested for a normal distribution using a Kolmogorov-Smirnov test. Group comparisons of categoric variables were performed using the chi-square test. Continuous variables including patient age, node size, and average strain ratio were expressed as means ± SDs, and mean differences between the two groups (benign vs malignant) were compared using an independent t test. A ROC curve analysis was applied to obtain the optimal cutoff value for the shortest-to-longest axis ratio and the average strain ratio. The largest area under the ROC curve (A z ) value was used for differentiation between the two groups. A univariate logistic regression analysis was used first to evaluate the predictive power of each variable. The US variables with high predictive power (p < 0.20, Wald test) were selected and fed into a multivariate logistic regression analysis to investigate an optimal logistic regression model for distinguishing between benign and malignant cervical lymphadenopathies. The results of this analysis are presented as odds ratio (OR) estimates with corresponding 95% CIs and p values from the Wald test. After multivariate analyses, we obtained a regression equation for the fitted probability of malignancy in cervical lymph nodes. The scores for each significant factor were multiplied by the beta coefficient obtained for each significant factor from the multivariate logistic regression analysis with generalized estimating equations. To evaluate the distribution of fitted probabilities associated with the number of suspicious US features, we estimated the logit (as the intercept plus the sum of the beta values multiplied by the given level of each feature variable), which was subsequently used for estimating the fitted probabilities. The linear-by-linear association test was used to evaluate the linear association between the number of suspicious imaging features and the probability of malignancy in a given lymph node. The interobserver agreement to interpret each imaging feature was analyzed using coefficient of interrater agreement (Cohen kappa). The kappa values were interpreted as follows: less than 0.20, poor agreement; , fair agreement; , moderate agreement; , good agreement; and , very good agreement. All statistical analyses were performed using statistics software (SPSS, version 13.0, SPSS; and MedCalc, version 14.10, MedCalc Software). A p value of < 0.05 was considered statistically significant. Results The final diagnoses of cervical lymphadenopathy are listed in Figure 1. Of the 291 patients, 182 (male-female ratio, 52:130; mean age, 43.5 years; age range, 5 82 years) had benign lymphadenopathy and 109 (male-female ratio, 56:53; mean age, 60.0 years; age range, years) had malignant lymphadenopathy. Patient age and sex of the two groups were significantly different (p < 0.001). The mean size of the lymph nodes was significantly larger in patients with malignant lymphadenopathy than in those with benign lymphadenopathy (shortest diameter: 1.12 ± 0.68 [SD] vs 0.72 ± 0.30 cm, respectively; longest diameter: 1.87 ± 1.01 vs 1.58 ± 0.69 cm) (p < 0.001). In the ROC curve analysis, the optimal cutoff value of the shortest-to-longest axis ratio in the longitudinal scan of a lymph node was 0.5. A shortest-to-longest axis ratio of more than 0.5 had the largest A z value (A z = 0.683; 95% CI, ; sensitivity, 71.6%; specificity, 59.3%). In addition, the optimal cutoff value of the strain ratio was A strain ratio of more than 2.41 also had the largest A z value (A z = 0.687; 95% CI, ; sensitivity, 54.1%; specificity, 75.3%). In the univariate analysis, the following imaging features showed a significant association with malignancy compared with benign lymphadenopathy: round shape (71.6% vs 40.7%, respectively; p < 0.001), noncircumscribed margin (16.5% vs 6%; p = 0.007), hyperechogenicity (50.5% vs 11.5%; p < 0.001), absence of echogenic hilum (81.7% vs 28.6%; p < 0.001), presence of gross necrosis (18.3% vs 7.7%; p = 0.011), presence of calcification (19.3% vs 3.8%; p < 0.001), peripheral or mixed vascular pattern (78% vs 26.4%; p < 0.001), malignant RTE assessment (64.2% vs 24.7%; p < 0.001), and higher strain ratio (54.1% vs 24.2%; p < 0.001) (Fig. 2). Matting was not significantly different between the two groups (12.8% vs 6.6%; p = 0.11). In the multivariate analysis, the following US features showed a significant and independent association with malignant lymphadenopathy: round shape (OR = 2.734; 95% CI, ; p = 0.004), hyperechogenicity (OR = 3.224; 95% CI, ; p = 0.002), absence of echogenic hilum (OR = 3.361; 95% CI, ; p = 0.001), presence of calcification (OR = 7.657; 95% CI, ; p = 0.001), peripheral or mixed vascular pattern (OR = 5.653; 95% CI, ; p < 0.001), and malignant RTE assessment (OR = 2.118; 95% CI, ; p = 0.03). Final regression analysis of selected imaging features allowed an elaboration of a regression equation to generate a logit of malignant lymphadenopathy (z) as follows: z = ( US1) + ( US2) + ( US3) + ( US4) US5) + ( US6), where US1 is shape (round = 1, oval = 0), US2 is echogenicity (hyperechogenicity = 1, isoechogenicity or hypoechogenicity = 0), US3 is the presence or absence of an echogenic hilum (absence of echogenic hilum = 1, presence of echogenic hilum = 0), US4 is the presence or absence of calcification (presence of calcification = 1, absence of calcification = 0); US5 is the presence or absence of peripheral and mixed vascular patterns (presence of peripheral and mixed vascular patterns = 1, hilar and avascular vascular patterns = 0); and US6 is the RTE assessment (malignant RTE assessment = 1, benign RTE assessment = 0). Using the value of z, we calculated the fitted probability of malignant lymphadenopathy according to the number of suspicious features as follows: fitted probability = 1 / (1 + e z ), where e is a mathematic constant of The values of fitted probabilities were in the lymph nodes with no suspicious features, in those with one suspicious feature, in those with two suspicious features, in those with three suspicious features, AJR:206, June

5 Ryu et al. in those with four suspicious features, in those with five suspicious features, and in those with all suspicious features (Fig. 3). The linear-by-linear association test showed that the risk of malignant lymphadenopathy increased as the number of suspicious features increased (p < 0.001). With these findings, we suggested CLN-RADS, which is based on the US and RTE features of a cervical lymph node: category 1, probably benign (no suspicious features); 2, low suspicion for malignancy (one suspicious feature); 3, moderate suspicion for malignancy (two suspicious features); 4, high suspicion for malignancy (three or four suspicious features); and 5, highly suggestive of malignancy (five or more suspicious features). According to the CLN-RADS categories, the risk of malignant lymphadenopathy was as follows: category 1, 3.3%; category 2, 10.9%; category 3, 26.7%; category 4, %; and category 5, %. The overall interobserver agreement of shape, calcification, RTE assessment, and elasticity score was good (κ = 0.72, 0.73, 0.75, and 0.71, respectively). The overall interobserver agreement of margin, echogenicity, echogenic hilum, gross necrosis, and intranodal vascular pattern was moderate (κ = 0.58, 0.49, 0.51, 0.54, and 0.47, respectively). Among the imaging features, interobserver agreement was the worst for matting (κ = 0.37). Discussion An attempt to establish a standardized reporting system has been made for several organs including breast, liver, prostate, and ovary. In the current study, we propose CLN-RADS the first structured reporting system for cervical lymph nodes to facilitate patient management by physicians and improve communication between radiologists and physicians. A 5-point scale is applied for CLN-RADS to determine the risk stratification of cervical lymphadenopathy according to the number of suspicious features based on the combination of US and RTE. Therefore, CLN-RADS can be applied conveniently in clinical practice. The use of a structured reporting system has several benefits. The most important benefit is to improve communication between radiologists and physicians [5]. In most institutions, one person performs the US examinations, and another person makes decisions regarding patient care and treatment. Occasionally, the radiologic reports contain vague descriptions or do not include an assessment Fig. 3 Box plot shows fitted probabilities of malignancy calculated with logistic regression using conventional ultrasound and real-time elastography features of 291 cervical lymph nodes. Upper horizontal line of each box is 75th percentile of dataset, and lower horizontal line of each box is 25th percentile. Horizontal bar within some boxes represents median. Lower and upper horizontal lines outside boxes represent 10th percentile and 90th percentile of dataset, respectively. Circles represent outliers. Fitted Probability No. of Suspicious Findings of the radiologic findings and recommendations. Physicians select a management strategy using the information in the radiologic reports, but they may be confused by or may misinterpret radiologic reports. CLN-RADS is intended to reduce variability and error in imaging interpretation by providing an unified language for reporting. Additionally, this structured reporting system can provide a common approach to the interpretation of imaging for our colleagues in different medical institutions, for various third-party payers, and for the general public. CLN-RADS can be also used to improve communication among radiologists at academic meetings. The first structured reporting system established in radiology was BI-RADS, which was developed in 1993 for the assessment of breast lesions on mammography [12]. BI-RADS has proved to be a very successful tool that has influenced the management of patients with breast disease. The current classification system for cervical lymphadenopathy, CLN-RADS, is similar to BI-RADS; CLN-RADS is composed of five categories that are based on the risk of malignancy. The BI-RADS category is determined by the comprehensive judgment of radiologists. In contrast, the CLN-RADS category is determined by the number of suspicious features on US and RTE images. Previous studies on risk stratification for malignant lymphadenopathy based on US features are sparse; therefore, application of comprehensive judgment is limited for the structured reporting system of cervical lymph nodes [13]. The malignant features of cervical lymph nodes on conventional US have been investigated in previous studies [14 16]. Most US features associated with malignant lymphadenopathy in previous studies corresponded with our results. However, matting was not a predictive factor for malignant lymphadenopathy in our study. Matting of cervical lymph nodes has been reported to be a characteristic feature of metastasis and tuberculous lymphadenitis [4, 17]. Matting is believed to result from extracapsular spread to adjacent soft tissue in patients with metastatic lymph nodes or from perinodal inflammatory reactions in patients with tuberculous lymphadenitis. In the current study, cases of benign lymphadenopathy included patients with tuberculous lymphadenitis (8.8%, 16/182); this finding is not surprising because tuberculosis is prevalent in our country. The inclusion of tuberculous lymphadenitis can affect matting to reveal no statistical significance between benign and malignant lymphadenopathies. A few investigations have reported that the use of elastography improved diagnostic performance when combined with conventional US to determine malignancy in patients with cervical lymphadenopathy [11, 18]. Alam et al. [11] reported that the combination of elastography with B-mode US revealed diagnostic performance superior to that of B-mode US alone (sensitivity, 92% vs 98%, respectively; specificity, 94% vs 59%; accuracy, 93% vs 84%). Similarly, other researchers reported the qualitative evaluation of elastography in patients with oral squamous cell carcinoma [18]; they reported that the combination of elastography with conventional US showed higher diagnostic values than either modality alone [18]. Additionally, RTE can help to increase the detection rate of malignancy and identify suspicious lymph nodes requiring aspiration or biopsy [7, 11]. Therefore, RTE can be included in a structured re AJR:206, June 2016

6 Cervical Lymph Node Imaging Reporting and Data System porting system for cervical lymph nodes to diagnose malignant lymphadenopathy. The overall interobserver agreement for the imaging features of the cervical lymph nodes was relatively high in this study for all features except matting. Our results for the imaging features of cervical lymph nodes are comparable with those of other studies for the US features of different organs [19, 20]. Interobserver agreement for the interpretation of breast masses using the BI-RADS lexicon for US was fair to moderate (κ = ) [19]. For thyroid nodules, interobserver agreement for the interpretation of conventional US was moderate to good (κ = ) [20]. For the establishment of the widespread application of a structured reporting system, the reproducibility of the imaging interpretation is an important factor. There are several limitations to this study. First, this study was retrospectively designed. Because our hospital is a referral medical center, the proportion of patients with malignant lymphadenopathy might be relatively higher than that seen in the general population. Cases of benign lymphadenopathy with no suspicious features on conventional US and no clinical suspicion of other infectious causes were not included in this study because these patients did not undergo FNA or CNB. Therefore, patient selection bias may have occurred. Second, shear-wave elastography was not available, and the absolute values of the elasticity measurements could not be evaluated in this study. Strain elastography has the disadvantage that it is operator-dependent because the operator performs compression. Excessive compression can alter tissue stiffness and result in nonaxial displacement. Strain elastography tends to reveal lower intraobserver and interobserver reproducibility than shear-wave elastography. Third, our results were obtained from a single institution with a small population. Most structured reporting systems of imaging of different organs have been established through the deliberations of consensus committees in established professional organizations. CLN-RADS is not yet ready to be widely applied to practice. More prospective studies in an independent cohort and establishment through professional organizations are needed to validate the efficacy of CLN-RADS. Fourth, each suspicious feature was regarded with the same weight of predicted probability of malignancy. A predictive categoric system derived from a total risk score is proposed that reflects the different malignancy probabilities from each suspicious feature of US. To conclude, we propose CLN-RADS as a structured reporting system for cervical lymph nodes. CLN-RADS has been established using the risk stratification of malignant lymphadenopathy according to the number of suspicious conventional US and RTE features. CLN-RADS can be reproducible to interpret imaging features with a relatively high degree of interobserver agreement. References 1. Khanna R, Sharma AD, Khanna S, Kumar M, Shukla RC. Usefulness of ultrasonography for the evaluation of cervical lymphadenopathy. World J Surg Oncol 2011; 9:29 2. Na DG, Lim HK, Byun HS, Kim HD, Ko YH, Baek JH. Differential diagnosis of cervical lymphadenopathy: usefulness of color Doppler sonography. AJR 1997; 168: Bruneton JN. Ultrasonography of the neck. Berlin, Germany: Springer-Verlag, 1987:81 4. Ahuja AT, Ying M, Ho SY, et al. Ultrasound of malignant cervical lymph nodes. Cancer Imaging 2008; 8: Amor F, Vaccaro H, Alcazar JL, Leon M, Craig JM, Martinez J. Gynecologic imaging reporting and data system: a new proposal for classifying adnexal masses on the basis of sonographic findings. J Ultrasound Med 2009; 28: D Orsi CJ, Mendelson, EB, Ikeda DM, et al. Breast Imaging Reporting and Data System: ACR BI-RADS breast imaging atlas. Reston, VA: American College of Radiology, 2003:14 7. Ying L, Hou Y, Zheng HM, Lin X, Xie ZL, Hu YP. Real-time elastography for the differentiation of benign and malignant superficial lymph nodes: a meta-analysis. Eur J Radiol 2012; 81: Ghajarzadeh M, Mohammadifar M, Azarkhish K, Emami-Razavi SH. Sono-elastography for differentiating benign and malignant cervical lymph nodes: a systematic review and meta-analysis. Int J Prev Med 2014; 5: Mendelson EB, Böhm-Vélez M, Berg WA, et al. ACR BI-RADS ultrasound. In: D Orsi CJ, Sickles EA, Mendelson EB, Morris EA, et al. ACR BI-RADS Atlas, Breast Imaging Reporting and Data System. Reston, VA: American College of Radiology, 2013: Lenghel LM, Bolboaca SD, Botar-Jid C, Baciut G, Dudea SM. The value of a new score for sonoelastographic differentiation between benign and malignant cervical lymph nodes. Med Ultrason 2012; 14: Alam F, Naito K, Horiguchi J, Fukuda H, Tachikake T, Ito K. Accuracy of sonographic elastography in the differential diagnosis of enlarged cervical lymph nodes: comparison with conventional B-mode sonography. AJR 2008; 191: American College of Radiology. Breast Imaging Reporting and Data System (BI-RADS). Reston, VA: American College of Radiology, Liao LJ, Wang CT, Young YH, Cheng PW. Realtime and computerized sonographic scoring system for predicting malignant cervical lymphadenopathy. Head Neck 2010; 32: Dudea SM, Lenghel M, Botar-Jid C, Vasilescu D, Duma M. Ultrasonography of superficial lymph nodes: benign vs. malignant. Med Ultrason 2012; 14: Ying M, Bhatia KS, Lee YP, Yuen HY, Ahuja AT. Review of ultrasonography of malignant neck nodes: greyscale, Doppler, contrast enhancement and elastography. Cancer Imaging 2013; 13: Gupta A, Rahman K, Shahid M, et al. Sonographic assessment of cervical lymphadenopathy: role of high-resolution and color Doppler imaging. Head Neck 2011; 33: Ahuja A, Ying M. Sonography of neck lymph nodes. Part II. Abnormal lymph nodes. Clin Radiol 2003; 58: Ishibashi N, Yamagata K, Sasaki H, et al. Realtime tissue elastography for the diagnosis of lymph node metastasis in oral squamous cell carcinoma. Ultrasound Med Biol 2012; 38: Abdullah N, Mesurolle B, El-Khoury M, Kao E. Breast Imaging Reporting and Data System lexicon for US: interobserver agreement for assessment of breast masses. Radiology 2009; 252: Koh J, Moon HJ, Park JS, et al. Variability in interpretation of ultrasound elastography and grayscale ultrasound in assessing thyroid nodules. Ultrasound Med Biol 2016; 42:51 59 AJR:206, June