Diagnostic Accuracy of Sonographic Criteria for Evaluation of Cervical Lymphadenopathy



Similar documents
Sonography of Neck Lymph Nodes. Part I: Normal Lymph Nodes

Ultrasonography of the Adrenal Glands CVM 6105 Kari L. Anderson, DVM, Diplomate ACVR Associate Clinical Professor of Veterinary Radiology

IV. DEFINITION OF LYMPH NODE GROUPS (FIGURE 1) Level IA: Submental Group

Cervical lymphadenopathy

Power Doppler Sonography to Differentiate Tuberculous Cervical Lymphadenopathy from Nasopharyngeal Carcinoma

How to Detect a Thyroid Pyramidal Lobes

Lymph Node Sonography

Ultrasonography of superficial lymph nodes: benign vs. malignant

Incidence of Incidental Thyroid Nodules on Computed Tomography (CT) Scan of the Chest Performed for Reasons Other than Thyroid Disease

VI. FREQUENTLY ASKED QUESTIONS CONCERNING BREAST IMAGING AUDITS

Recommendations for cross-sectional imaging in cancer management, Second edition

Pediatric Oncology for Otolaryngologists

Sonographic Evaluation of Isolated Abnormal Axillary Lymph Nodes Identified on Mammograms

Laparoscopic Ultrasonography Assisted Retroperitoneal Lymph Node Sampling in Patients Evaluated for Stomach Cancer Recurrence

Gray scale assessment of axillary lymph nodes in women suspected of breast cancer.

Recurrent & Persistent Papillary Thyroid Cancer Central Nodal Dissection vs. Node-Picking Patterns of Nodal Metastases Recurrent Laryngeal Nerve,

TNM Staging of Head and Neck Cancer and Neck Dissection Classification

SCD Case Study. Most malignant lesions of the tonsil are either lymphosarcoma or carcinoma.

Update on thyroid cancer surveillance and management of recurrent disease. Minimally invasive thyroid surgery

HOVON Staging and Response Criteria for Non-Hodgkin s Lymphomas Page 1

Patterns of nodal spread in thoracic malignancies

Common and Uncommon Sonographic Features of Papillary Thyroid Carcinoma

Advances in Differentiated Thyroid Cancer

D. FREQUENTLY ASKED QUESTIONS

Il percorso diagnostico del nodulo tiroideo: il ruolo dell analisi molecolare

Sonographic Findings in the Surgical Bed After Thyroidectomy

Ovarian Torsion: Sonographic Evaluation

General Rules SEER Summary Stage Objectives. What is Staging? 5/8/2014

9. Discuss guidelines for follow-up post-thyroidectomy for cancer (labs/tests) HH

Multi-slice Helical CT Scanning of the Chest

SUNY DOWNSTATE MEDICAL CENTER SURGERY GRAND ROUNDS February 28, 2013 VERENA LIU, MD ROSEANNA LEE, MD

The TV Series. INFORMATION TELEVISION NETWORK

Staging Head and Neck Cancers Transitioning to the Seventh Edition of The AJCC Cancer Staging Manual

Duplication Images in Vascular Sonography

Kidney Cancer OVERVIEW

Sonography of Wrist Ganglion Cysts

Medullary Renal Cell Carcinoma Case Report

Measures of diagnostic accuracy: basic definitions

BRAF in the diagnostic evaluation of thyroid nodules

GENERAL CODING. When you review old cases that were coded to unknown, make corrections based on guidelines in effect at the time of diagnosis.

OBJECTIVES By the end of this segment, the community participant will be able to:

Disease/Illness GUIDE TO ASBESTOS LUNG CANCER. What Is Asbestos Lung Cancer? Telephone

Sonographic Diagnosis of Ureteral Tumors

Chapter 2 Staging of Breast Cancer

Surgical Management of Papillary Microcarcinoma 趙 子 傑 長 庚 紀 念 醫 院 林 口 總 院 一 般 外 科

Thyroid and Adrenal Gland

First floor, Main Hospital North Services provided 24/7 365 days per year

NICE Pathways bring together all NICE guidance, quality standards and other NICE information on a specific topic.

Thyroid Fine-Needle Aspiration Indications and Technique. Subcommittee members Zubair W. Baloch, MD, PhD Martha Bishop Pitman, MD

Metastatic renal cell carcinoma to the left maxillary sinus

Controlling recurrent papillary thyroid carcinoma in the neck by ultrasonographyguided

The diagnostic usefulness of tumour markers CEA and CA-125 in pleural effusion

Soft Tissue Neck CT Anatomy

Sonographic Features of Medullary Thyroid Carcinomas According to Tumor Size

Breast cancer close to the nipple: Does this carry a higher risk ofaxillary node metastasesupon diagnosis?

Primary -Benign - Malignant Secondary

Report series: General cancer information

Thyroid Cancer: Resection, Dissection, Surveillance and Recurrence. Cord Sturgeon, MD

Sonographic Features Related to Volvulus in Neonatal Intestinal Malrotation

THYROID CANCER. I. Introduction

Diagnostic Sensitivity of Ultrasound-Guided Needle Biopsy in Soft Tissue Masses About Superficial Bone Lesions

The Role of Genetic Testing in the Evaluation of Thyroid Nodules. Thyroid Cancer and FNA. Thyroid Cancer. Pure Follicular Cancers.

Ultrasound Examinations of the Head and Neck

Sonographic Demonstration of Couinaud s Liver Segments

Rotation Specific Goals & Objectives: University Health Network-Princess Margaret Hospital/ Sunnybrook Breast/Melanoma

LYMPHOMA IN DOGS. Diagnosis/Initial evaluation. Treatment and Prognosis

Neoplasms of the LUNG and PLEURA

CHAPTER 2. Neoplasms (C00-D49) March MVP Health Care, Inc.

Releasing Nuclear Medicine Patients to the Public: Dose Calculations and Discharge Instructions

The recommendations made throughout this book are by the National Health and Medical Research Council (NHMRC).

Diagnosis and Prognosis of Pancreatic Cancer

School of Diagnostic Medical Sonography

Early Colonoscopy in Patients with Acute Diverticulitis Simon Bar-Meir, M.D.

Sonographic Findings in Skeletal Muscle Metastasis From Renal Cell Carcinoma

Guideline for the Imaging of Patients Presenting with Breast Symptoms incorporating the guideline for the use of MRI in breast cancer

LIVER CANCER AND TUMOURS

Ultrasonographic Determination of Equine Fetal Gender (31 Mar 2000)

Image SW Review the anatomy of the EAC and how this plays a role in the spread of tumors.

Bayes Theorem & Diagnostic Tests Screening Tests

Small cell lung cancer

Prediction of Pregnancy Outcome Using HCG, CA125 and Progesterone in Cases of Habitual Abortions

In Practice Whole Body MR for Visualizing Metastatic Prostate Cancer

Test Request Tip Sheet

Metastatic Prostate Cancer Causing Complete Obstruction of the IVC

YOUR LUNG CANCER PATHOLOGY REPORT

Oncology. Objectives. Cancer Nomenclature. Cancer is a disease of the cell Cancer develops when certain cells begin to grow out of control

Latest Oncologic Strategies for Well-Differentiated Thyroid Carcinoma

2011 Radiology Diagnosis Coding Update Questions and Answers

Lung Cancer: Diagnosis, Staging and Treatment

Ultrasonography of the Accessory Nerve

The Need for Accurate Lung Cancer Staging

Crosswalk for Positron Emission Tomography (PET) Imaging Codes G0230 G0030, G0032, G0034, G0036, G0038, G0040, G0042, G0044, G0046

95% of childhood kidney cancer cases are Wilms tumours. Childhood kidney cancer is extremely rare, with only 90 cases a year in

A912: Kidney, Renal cell carcinoma

Breast Ultrasound: Benign vs. Malignant Lesions

Detection and staging of recurrent prostate cancer is still one of the important clinical problems in prostate cancer. A rise in PSA or biochemical

Prognosis of Very Large First-Trimester Hematomas

Brain Cancer. This reference summary will help you understand how brain tumors are diagnosed and what options are available to treat them.

Lung cancer is not just one disease. There are two main types of lung cancer:

Transcription:

Diagnostic Accuracy of Sonographic Criteria for Evaluation of Cervical Lymphadenopathy Michael Ying, MPhil, Anil Ahuja, FRCR, Constantine Metreweli, FRCR Although ultrasonographic criteria for abnormal nodes are used routinely in the evaluation of cervical lymphadenopathy, the diagnostic accuracy of these criteria in different areas has not been documented. This study evaluated 977 normal cervical nodes from 80 normal subjects and 1419 abnormal cervical nodes from 277 patients with proven cervical lymphadenopathy. The diagnostic accuracy (sensitivity, specificity, positive predictive value, negative predictive value, and accuracy) of size, shape (short axis to long axis ratio), echogenic hilus, and nodal border for regional lymph nodes are evaluated and discussed. KEY WORDS: Lymph nodes, cervical; Lymphadenopathy; Ultrasonography, diagnostic criteria. ABBREVIATIONS UGFNAC, Ultrasonographically guided fine-needle aspiration cytology; S/L, short axis to long axis; CCA, Common carotid artery; IJV, Internal jugular vein; PPV, Positive predictive value; NPV, Negative predictive value; ROC, Receiver operating characteristic; CT, Computed tomography Received November 28, 1997, from the Department of Optometry and Radiography (M.Y.), Hong Kong Polytechnic University, Hung Hom, Kowloon; and the Department of Diagnostic Radiology and Organ Imaging (A.A., C.M.), Prince of Wales Hospital, Shatin, New Territories, Hong Kong. Revised manuscript accepted for publication March 16, 1998. Address correspondence and reprint requests to Michael Ying, Department of Optometry and Radiography, Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong. Sonographic criteria for distinguishing normal from abnormal lymph nodes have been described previously. These include size, shape, echogenic hilus, nodal border, and distribution. 1 4 However, the individual reliability of these criteria has not been assessed. Some of the previously described criteria for assessing cervical lymph nodes can specifically identify abnormal nodes (such as intranodal necrosis, matting, and surrounding soft tissue edema), whereas others are not specific for either normal or abnormal nodes (such as size, shape, echogenic hilus, and nodal border). Moreover, as reported earlier, 5 we found that the normal appearance of cervical lymph nodes is not the same in different regions of the neck. Thus, this study was undertaken to assess the accuracy of these nonspecific criteria in differentiating normal nodes from abnormal nodes in different regions of the neck. 1998 by the American Institute of Ultrasound in Medicine J Ultrasound Med 17:437 445, 1998 0278-4297/98/$3.50

438 CERVICAL LYMPHADENOPATHY J Ultrasound Med 17:437 445, 1998 MATERIALS AND METHODS Eighty healthy subjects (37 men, 43 women) with no history of malignancy, tuberculosis, chronic tonsillitis, glandular fever, or neck surgery were included in the study. All subjects were Cantonese Chinese, and none of them had any history of recent acute inflammation of the upper respiratory tract or ear, nose, or throat disease. The subjects ranged in age from 12 to 76 years (mean, 32.2 years). Scans were performed with a 5 to 10 MHz linear transducer on a General Electric LOGIQ 500 (General Electric, Milwaukee, WI) ultrasound system. The subjects lay supine on the examination couch with the shoulders supported by a pillow. The neck was hyperextended. As the shape of the lymph nodes depends on the scan plane, scans were obtained with the transducer placed transversely and longitudinally, and measurements were made in the plane that showed a maximum cross-sectional area of the node. In each subject, all cervical lymph nodes detected in the examination were included in the study. A review of the ultrasonograms of 277 patients (151 men, 126 women) with proven cervical lymphadenopathy also was conducted as a part of this study. The patients were classified into seven etiologic groups: 1. Pharyngeal, laryngeal, and esophageal carcinomas, 33 patients 2. Oral cavity carcinomas, 23 patients 3. Infraclavicular carcinomas, 16 patients 4. Papillary carcinoma of the thyroid, 55 patients 5. Nasopharyngeal carcinoma, 84 patients 6. Non-Hodgkin lymphoma, 19 patients 7. Tuberculous lymphadenitis, 47 patients None of these patients had had previous radiotherapy or chemotherapy. Their age range was 12 to 94 years (mean, 47.6 years). All the scans were performed on a commercially available ultrasound system (Aloka 650, Tokyo, Japan) with a linear 7.5 and a sector 10 MHz transducer with a built-in water bath. All the patients underwent UGFNAC on the largest node. In each patient, all other lymph nodes with a ultrasonographic appearance similar to that of the aspirated node were considered to have a similar pathologic result. Cervical lymph nodes were assessed for their size, shape, echogenic hilus, and sharpness of border. As the longitudinal diameter of the nodes was found to be an unreliable criterion in assessment of cervical nodes, 6 only the short axis of the nodes was evaluated. Six cut-off points of the short axis of nodes were evaluated for their accuracy: (1) 5 mm; (2) 6 mm; (3) 7 mm; (4) 8 mm; (5) 9 mm; and (6) 10 mm. These cut-off points of the short axis were selected because they were the sizes that were reported previously to be useful for distinguishing normal from abnormal cervical nodes. 2,3,7,8 In this study, a normal lymph node with a short axis greater than the selected cutoff point was considered a false-positive finding, whereas a short axis less than or equal to the selected cut-off point was considered a true-negative finding. Similarly, an abnormal lymph node with a short axis greater than the selected cut-off point was considered a true-positive result, whereas a short axis less than or equal to the selected cut-off point was considered a false-negative result. The shape of the cervical nodes was assessed by the S/L ratio. 4,9 An S/L ratio equal to or greater than 0.5 indicates a round node, whereas an S/L ratio less than 0.5 indicates an oval or elongated node. In this study, a normal lymph node with an S/L ratio equal to or greater than 0.5 was considered a false-positive finding, whereas an S/L ratio less than 0.5 was considered a true-negative finding. An abnormal lymph node with an S/L ratio equal to or greater than 0.5 was considered a true-positive result, whereas an S/L ratio less than 0.5 was considered a falsenegative result. The nodal hilus is identified as an intranodal echogenic structure continuous with the surrounding fat. 10 It should not be confused with coagulation necrosis, which may also be seen as echogenic area within the node. However, these echogenic foci are not continuous with surrounding fat. In the present study, a normal lymph node without an echogenic hilus was considered to be a false-positive finding, whereas a normal node with an echogenic hilus was considered a true-negative result. An abnormal lymph node without an echogenic hilus was considered a true-positive result, whereas an abnormal node with an echogenic hilus was considered a falsenegative finding. It has been reported that lymph nodes with a sharp border were considered to be abnormal, whereas those with an unsharp border tended to be normal. 7 In this study, a normal lymph node with a sharp border was considered a false-positive result, whereas a normal node with an unsharp border was considered a true-negative result. An abnormal lymph node with a sharp border was considered a true-positive finding, whereas an abnormal node with an unsharp border was considered a false-negative finding. The distribution of cervical lymph nodes was made into eight regions in the neck, similar to the

J Ultrasound Med 17:437 445, 1998 YING ET AL 439 method described by Hajek and coworkers 2 : (1) submental, (2) submandibular, (3) parotid, (4) upper cervical: above the hyoid bone and along the CCA and the IJV, (5) middle cervical: between the hyoid bone and the cricoid cartilage and along the CCA and IJV, (6) lower cervical: below the cricoid cartilage and along the CCA and IJV, (7) supraclavicular fossa, and (8) posterior triangle (also known as accessory chain) (Fig. 1). The sensitivity, specificity, PPV, NPV, and accuracy were calculated. The ROC curves for the size of the lymph nodes in different regions of the neck were plotted. Graphs showing the relationship between the sensitivity and specificity for the shape, echogenic hilus, and nodal border of the lymph nodes in different regions of the neck also were plotted. RESULTS In the 80 normal subjects, 977 cervical lymph nodes were detected. In the 277 patients, 1419 lymph nodes were detected. The distribution of normal and abnormal lymph nodes is shown in Table 1. As the numbers of normal nodes in regions 6 and 7 were low, the calculated statistical parameters may not be reliable in these areas. Therefore, only lymph nodes in submental (region 1), submandibular (region 2), parotid (region 3), upper cervical (region 4), middle cervical (region 5), and posterior triangle (region 8) were evaluated. How well the different sonographic criteria perform in these regions is shown in Tables 2 through 7. In the parotid region, using 10 mm as the size cut-off, the PPV was invalid as no true-positive or false-positive findings were obtained. The sensitivity and specificity for size in different regional nodes were charted as ROC curves (Fig. 2). The curve that is highest and near the Y axis (which is nearest the top left-hand corner) shows the best sensitivity with a high specificity. In Figure 2, the curve for the parotid nodes is the one nearest the top left-hand corner when compared with other curves. The sensitivity and specificity for shape (S/L ratio), echogenic hilus, and nodal border in different regional nodes were plotted in Figures 3 to 5, respectively. The point that is nearest the top left-hand corner shows that the criterion for the corresponding regional nodes has a high sensitivity together with a high specificity. In Figure 3 (graph for the nodal shape), the point for the submental nodes and the point for the upper cervical nodes have a similar distance, both being nearest to the top left-hand corner. The submental nodes show a higher sensitivity (97.6%), whereas the upper cervical nodes have a higher specificity (98.3%). Figure 4 (graph for the echogenic hilus) shows that the point for the submandibular nodes is nearest to the top left-hand corner. In Figure 5 (graph for the nodal border), the point for the submental nodes is nearest to the top left-hand corner. DISCUSSION Different sonographic criteria have been established by which normal lymph nodes can be distinguished from abnormal nodes. 4,10 12 Unlike other criteria, size, shape, echogenic hilus, and sharpness of nodal border are not specific for either normal or abnormal nodes. Thus, it is essential for sonographers to be Figure 1 Schematic diagram of the neck shows the distribution of the cervical lymph nodes.

440 ACERVICAL LYMPHADENOPATHY J Ultrasound Med 17:437 445, 1998 Table 1: Distribution of Normal and Abnormal Cervical Lymph Nodes Number of Lymph Nodes (Percentage) Region Nodes Normal Nodes Abnormal Nodes 1 Submental 45 (4.6%) 41 (2.9%) 2 Submandibular 237 (24.3%) 124 (8.7%) 3 Parotid 72 (7.4%) 22 (1.6%) 4 Upper cervical 230 (23.5%) 283 (19.9%) 5 Middle cervical 30 (3.1%) 75 (5.3%) 6 Lower cervical 2 (0.2%) 69 (4.9%) 7 Supraclavicular 7 (0.7%) 132 (9.3%) fossa 8 Posterior triangle 354 (36.2%) 673 (47.4%) Total 977 1419 familiar with the accuracy of these criteria in differentiating normal from abnormal nodes. The ROC curves for size in different regional nodes (Fig. 2) showed that the curve for the parotid nodes had the highest sensitivity together with a high specificity. It indicated that the short axis of lymph nodes was more accurate in assessment of the parotid nodes than in assessment of other regional nodes. In Figure 3, as the point for the submental nodes and the point for the upper cervical nodes had a similar distance to the top left-hand corner; thus we assumed that the nodal shape (S/L ratio) had a similar accuracy in differential diagnosis of the submental and upper cervical nodes and was more accurate when these nodes were examined. Similarly, nodal shape had a higher reliability when the echogenic hilus was used to assess the submandibular nodes (Fig. 4), whereas the nodal border was more accurate when the submental nodes were examined (Fig. 5). In assessing nodal size, we found that with increasing size, the sensitivity decreased whereas the specificity increased (Tables 2 through 7). The decreasing sensitivity was probably due to the fact that smaller positive nodes (abnormal nodes) were detected when increasing the cut-off point. However, the specificity increased because more large negative nodes (normal nodes) were detected. The result was consistent with that of previous reports in the literature. 6,11 Similar to these reports, 11 the accuracy did not constantly change with the cut-off point. The PPV and NPV are actually important indicators of the diagnostic usefulness of the criteria, since they show the probability of the criteria that give a correct diagnosis, whereas the sensitivity and specificity do not give this information. In order to make an objective evaluation of the nodal size, the cut-off point was selected when it obtained a highest accuracy. For the submental nodes (region 1), a short axis of 5 mm is the optimum size criterion for diagnosis as it shows the highest sensitivity (58%), specificity (100%), PPV (100%), NPV (73%), and accuracy (80%) among the cut-off points studied. The 100% for specificity and PPV was probably due to the fact that no false-positive nodes were found in this cut-off point (5 mm). Results showed that all normal submental nodes had a short axis of less than 5 mm, and any submental nodes with a short axis greater than 5 mm could indicate an abnormality (either inflammation or malignancy). The S/L ratio had an accuracy of about 84%, which makes it a better criterion than the size and the presence of an echogenic hilus for the diagnosis. The high NPV (97%) indicated that the S/L ratio was highly valuable in diagnosis of negative submental nodes, since only a few abnormal nodes show an S/L ratio less than 0.5 (one node in this study). Although the echogenic hilus was the least accurate criterion, it had about 77% accuracy, PPV, and NPV in the diagnosis of submental nodes overall. The nodal border is the most accurate among the criteria studied, as it has accuracy, PPV, and NPV of 85%. For the submandibular nodes (region 2), 9 mm is the most accurate size criterion (accuracy = 76%). The PPV and the NPV are 77% and 76%, respectively. This indicates that about 77% of nodes with a short axis greater than 9 mm were abnormal, and about Table 2: Performance of the for Submental Nodes (Region 1) Sensitivity 58.5 41.5 34.1 31.7 31.7 12.2 97.6 73.2 82.9 Specificity 100.0 100.0 100.0 100.0 100.0 100.0 71.1 80.0 86.7 PPV 100.0 100.0 100.0 100.0 100.0 100.0 75.5 76.9 85.0 NPV 72.6 65.2 62.5 61.6 61.6 55.6 97.0 76.6 84.8 Accuracy 80.2 72.1 68.6 67.4 67.4 58.1 83.7 76.7 84.9

J Ultrasound Med 17:437 445, 1998 YING ET AL 441 Table 3: Performance of the for Submandibular Nodes (Region 2) Sensitivity 75.0 71.8 54.0 49.2 43.5 29.0 78.2 71.0 83.1 Specificity 30.0 51.5 72.6 83.1 93.2 98.7 5.1 97.5 75.5 PPV 35.9 43.6 50.8 60.4 77.1 92.3 30.1 93.6 64.0 NPV 69.6 77.7 75.1 75.8 75.9 72.7 30.8 86.5 89.5 Accuracy 45.4 58.4 66.2 71.5 76.2 74.8 30.2 88.4 78.1 Table 4: Performance of the for Parotid Nodes (Region 3) Sensitivity 72.7 72.7 50.0 40.9 27.3 0.0 81.8 81.8 81.8 Specificity 95.8 97.2 100.0 100.0 100.0 100.0 43.1 88.9 70.8 PPV 84.2 88.9 100.0 100.0 100.0 30.5 69.2 46.2 NPV 92.0 92.1 86.7 84.7 81.8 76.6 88.6 94.1 92.7 Accuracy 90.4 91.5 88.3 86.2 83.0 76.6 52.1 87.2 73.4 Table 5: Performance of the for Upper Cervical Nodes (Region 4) Sensitivity 79.9 66.8 58.3 48.4 43.5 36.0 69.3 80.6 79.5 Specificity 53.5 74.8 87.0 95.7 97.4 99.1 98.3 83.0 56.1 PPV 67.9 76.5 84.6 93.2 95.3 98.1 98.0 85.4 69.0 NPV 68.3 64.7 62.9 60.1 58.3 55.7 72.2 77.6 69.0 Accuracy 68.0 70.4 71.2 69.6 67.6 64.3 82.3 81.7 69.0 Table 6: Performance of the for Middle Cervical Nodes (Region 5) Sensitivity 62.7 54.7 41.3 32.0 29.3 21.3 58.7 77.3 88.0 Specificity 96.7 96.7 96.7 96.7 96.7 96.7 96.7 80.0 3.3 PPV 97.9 97.6 96.9 96.0 95.7 94.1 97.8 90.6 69.5 NPV 50.9 46.0 39.7 36.3 35.4 33.0 48.3 58.5 10.0 Accuracy 72.4 66.7 57.1 50.5 48.6 42.9 69.5 78.1 63.8 Table 7: Performance of the for Posterior Triangle Nodes (Region 8) Sensitivity 72.5 67.5 60.8 48.7 40.3 32.5 60.5 85.7 77.6 Specificity 91.0 94.9 98.3 98.3 99.4 99.4 98.6 80.2 31.4 PPV 93.8 96.2 98.6 98.2 99.3 99.1 98.8 89.2 68.2 NPV 63.5 60.5 56.9 50.2 46.7 43.7 56.7 74.7 42.4 Accuracy 78.9 76.9 73.7 65.8 60.7 55.6 73.6 83.8 61.6

442 ACERVICAL LYMPHADENOPATHY J Ultrasound Med 17:437 445, 1998 Figure 2 ROC curves show the relationship between the sensitivity and the specificity of the size in different regional nodes (see also Tables 2 to 7). 76% of nodes with a short axis less than 9 mm were normal. The size criterion for submandibular nodes (9 mm) is higher than the size criterion of other regional nodes (5 to 7 mm). This probably is due to the large size of normal submandibular nodes, 5 and thus a higher size criterion is required to differentiate normal from abnormal nodes. Previous literature has also reported that the size criterion for submandibular nodes is higher than the size criterion for the cervical nodes in other areas. 1 The S/L ratio is not helpful for differentiating normal from abnormal nodes since the accuracy, PPV, and NPV are only about 30%. The specificity of the S/L ratio for submandibular nodes also is low (5%). The unsatisfactory result of the S/L ratio in submandibular nodes is due mainly to the fact that the majority of normal submandibular nodes have an S/L ratio greater than 0.5. 5 The echogenic hilus is the most accurate criterion, with an accuracy of 88%, PPV of 94%, and NPV of 86%. The high PPV indicated that submandibular nodes with absent echogenic hilus are highly suggestive of abnormality. The nodal border is the second most accurate criterion for the diagnosis. The accuracy, PPV, and NPV are 78%, 64%, and 89%, respectively. The relatively low PPV for the nodal border probably is due to the high prevalence of normal nodes with a sharp border (24%). For the parotid nodes (region 3), a short axis of 6 mm is the most accurate size criterion for the differentiation. It showed the highest sensitivity (73%), specificity (97%), PPV (89%), NPV (92%), and accuracy (91%) when compared with other size criteria studied. Similar to submandibular nodes, the S/L ratio is not accurate in differentiating normal parotid nodes from abnormal parotid nodes, since only 52% of lymph nodes were diagnosed correctly as normal or abnormal using this criterion. The low PPV (30%) probably is due to the high prevalence of normal parotid nodes with an S/L greater than 0.5 (57%). Although the S/L was not accurate for the diagnosis of parotid nodes, it was useful to predict negative nodes since about 89% of lymph nodes with an S/L less than 0.5 were normal nodes (NPV = 89%). The echogenic hilus is the second most accurate criterion for differentiating normal from abnormal parotid nodes (accuracy = 87%). The echogenic hilus is the criterion that shows the highest NPV (94%) among the criteria studied (i.e., it indicated that the presence of an echogenic hilus in parotid nodes is highly suggestive of normality). The nodal border is less accurate than the size and echogenic hilus but is more accurate than the S/L ratio. The low PPV of the nodal border (46%) occurs because of the high prevalence of normal nodes with a sharp border (29%). For the upper cervical nodes (region 4), a short axis of 7 mm is the most accurate size criterion (71%) when compared with other size criteria studied. Except for submandibular nodes, the selected size criterion for upper cervical nodes is higher than that for lymph nodes in other areas (5 to 6 mm). Som 1 also found a larger size criterion more accurate for submandibular and upper cervical nodes (both are 1.5 cm); however, the size criteria are different from those in the present study. The difference in the result may be due to the use of different imaging modalities. Som 1 used CT scanning, whereas ultrasonography was used in the present study. CT is reported to have a limited reliability in detecting lymph nodes less than 5 mm in diameter, 13,14 and, therefore, small nodes may be missed. However, ultrasonography can demonstrate lymph nodes as small as 2 mm in diameter. 3,15 The S/L ratio (82.3%) and the echogenic hilus (81.7%) had similar accuracies for the differentiation in the upper cervical region, which were higher than those of size (71%) and nodal border (69%). The high PPV of the S/L ratio indicated that an upper cervical node with an S/L ratio greater than 0.5 is highly suggestive of abnormality. However, the relatively low diagnostic accuracy of nodal border sharpness may be due to the high prevalence of sharp borders in normal upper cervical nodes (44%). For the middle cervical nodes (region 5), the most accurate size criterion for the diagnosis was 5 mm in the short axis, with an accuracy of 72%. The high PPV (98%) indicated that a 5 mm short axis was valuable in predicting a positive node. However, the low NPV (51%) showed that a cut-off point of 5 mm in the short axis was not accurate for predicting negative nodes, since only about half of the nodes with

J Ultrasound Med 17:437 445, 1998 YING ET AL 443 Figure 3 Graph shows the relationship between the sensitivity and the specificity of the shape in different regional nodes (see also Tables 2 to 7). Figure 4 Graph shows the relationship between the sensitivity and the specificity of the echogenic hilus in different regional nodes (see also Tables 2 to 7). a short axis less than 5 mm were normal. The S/L ratio and the nodal border had a similar diagnostic accuracy (69% and 64%, respectively). The high PPV of the S/L ratio (98%) showed that an S/L ratio greater than 0.5 in middle cervical nodes was highly suggestive of abnormality. The low PPV and NPV of the nodal border in middle cervical nodes probably occurred because the majority of normal nodes had a sharp border (97%). The echogenic hilus is the most accurate criterion for differentiating normal from abnormal nodes in the middle cervical region (accuracy = 78%). The high PPV (91%) makes it a useful criterion in predicting abnormal nodes. In the middle cervical region, the NPV of the four criteria studied tends to be low (10% to 58%), which is not seen in other areas and indicates that none of the criteria studied can identify a normal middle cervical node accurately. For the posterior triangle nodes (region 8), the most accurate cut-off point of the short axis was 5 mm, with an accuracy of 79%. Lymph nodes greater than 5 mm are highly likely to be abnormal (PPV = 94%). However, the probability that lymph nodes with a short axis less than 5 mm were normal was 63% (NPV = 63%). The S/L ratio had a high PPV (99%), which showed that posterior triangle lymph nodes with an S/L ratio greater than 0.5 were highly suggestive of abnormality. The relatively low NPV (57%) probably was due to the high prevalence of abnormal nodes with an S/L ratio less than 0.5. The echogenic hilus is the most accurate diagnostic criterion in the posterior triangle region, with an accuracy of 88%, PPV of 89%, and NPV of 75%. The nodal border is the least accurate diagnostic criterion in the posterior triangle, with the lowest accuracy (62%), PPV (68%), and NPV (42%) when compared with other criteria. The inaccuracy probably is due to the high prevalence of unsharp borders in abnormal nodes (22%) and the high rate of sharp borders in normal nodes (69%). Normal posterior triangle nodes with sharp borders also were noted in our previous report. 5 CONCLUSION In the assessment of the cervical lymph nodes, the short axis is more accurate for the parotid nodes than for the lymph nodes in other regions. The nodal shape is more reliable in the differential diagnosis when the submental and upper cervical nodes are assessed. The echogenic hilus is more accurate when it is used to assess the submandibular nodes. The

444 ACERVICAL LYMPHADENOPATHY J Ultrasound Med 17:437 445, 1998 accurate criterion is the nodal border (69%). A short axis of 7 mm is considered the most accurate size criterion (71%). Upper cervical nodes with an S/L ratio greater than 0.5 are highly suggestive of abnormality. For the middle cervical nodes, the most accurate criterion is the echogenic hilus (78%), and the least accurate criterion is the nodal border (64%). A short axis of 5 mm is the most accurate size criterion (72%). A short axis greater than 5 mm, an S/L ratio greater than 0.5, and absence of an echogenic hilus in middle cervical nodes are highly suggestive of abnormality. For the posterior triangle nodes, the most accurate criterion is the echogenic hilus (84%), whereas the least accurate criterion is the nodal border (62%). A short axis of 5 mm is considered the most accurate size criterion (79%). A short axis greater than 5 mm or an S/L ratio greater than 0.5 in the posterior triangle nodes is highly suggestive of abnormality. In conclusion, ultrasonography is an ideal initial examination for cervical lymphadenopathy if appropriate sonographic criteria are used. Figure 5 Graph shows the relationship between the sensitivity and the specificity of the nodal border in different regional nodes (see also Tables 2 to 7). nodal border is more accurate when the submental nodes are assessed. For the submental nodes, the nodal border is the most accurate criterion (85%), whereas the echogenic hilus is the least accurate criterion (77%). A short axis of 5 mm is the most accurate size criterion (80%). Submental nodes with a short axis greater than 5 mm should be considered abnormal. Submental nodes with an S/L ratio less than 0.5 are highly likely to be normal. For the submandibular nodes, the most accurate criterion is the echogenic hilus (88%); the least accurate criterion is the shape of lymph nodes (S/L) (30%). A short axis of 9 mm is the most accurate size criterion (76%). An absent hilus in the submandibular nodes is highly suggestive of abnormality. For the parotid nodes, the most accurate criterion is the size of lymph nodes (short axis) (91%), whereas the least accurate criterion is their shape (52%). A short axis of 6 mm is found to be the most accurate size criterion. Parotid nodes with a short axis less than 6 mm, with an echogenic hilus, or with an unsharp border are strongly suspected to be normal. For the upper cervical nodes, the most accurate criterion is the shape of lymph nodes (82%); the least REFERENCES 1. Som PM: Lymph nodes of the neck. Radiology 165:593, 1987 2. Hajek PC, Salomonowitz E, Turk R, et al: Lymph nodes of the neck: Evaluation with US. Radiology 158:739, 1986 3. Solbiati L, Cioffi V, Ballarati E: Ultrasonography of the neck. Radiol Clin North Am 30:941, 1992 4. Tohnosu N, Onoda S, Isono K: Ultrasonographic evaluation of cervical lymph node metastases in esophageal cancer with special reference to the relationship between the short to long axis ratio (S/L) and the cancer content. J Clin Ultrasound 17:101, 1989 5. Ying M, Ahuja A, Brook F, et al: Sonographic appearance and distribution of normal cervical lymph nodes in a Chinese population. J Ultrasound Med 15:431, 1996 6. van den Brekel MW, Stel HV, Castelijns JA, et al: Cervical lymph node metastasis: Assessment of radiologic criteria. Radiology 177:379, 1990 7. Shozushima M, Suzuki M, Nakasima T, et al: Ultrasound diagnosis of lymph node metastasis in head and neck cancer. Dentomaxillofac Radiol 19:165, 1990 8. Bruneton JN, Balu-Maestro C, Marcy PY, et al: Very high frequency (13 MHz) ultrasonographic examination of the normal neck: Detection of normal lymph nodes and thyroid nodules. J Ultrasound Med 13:87, 1994 9. Solbiati L, Arsizio B, Rizzatto G, et al: High-resolution sonography of cervical lymph nodes in head and neck cancer: Criteria for differentiation of reactive versus malignant nodes. Radiology 169(P):113, 1988 10. Sakai F, Kiyono K, Sone S, et al: Ultrasonic evaluation of cervical metastatic lymphadenopathy. J Ultrasound Med 7:305, 1988

J Ultrasound Med 17:437 445, 1998 YING ET AL 445 11. van den Brekel MW, Castelijns JA, Stel HV, et al: Modern imaging techniques and ultrasound-guided aspiration cytology for the assessment of neck node metastases: A prospective comparative study. Eur Arch Otorhinolaryngol 250:11, 1993 12. Ahuja A, Ying M, Yang WT, et al: The use of sonography in differentiating cervical lymphomatous lymph nodes from cervical metastatic lymph nodes. Clin Radiol 51:186, 1996 13. Ishii J, Amagasa T, Tachibana T, et al: US and CT evaluation of cervical lymph node metastasis from oral cancer. J Craniomaxillofac Surg 19:123, 1991 14. Mancuso AA, Maceri D, Rice D, et al: CT of cervical lymph node cancer. AJR 136:381, 1981 15. Sugama Y, Kitamura S: Ultrasonographic evaluation of neck and supraclavicular lymph nodes metastasized from lung cancer. Intern Med 31:160, 1992