| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Head and Neck Imaging |
1 From the Departments of Diagnostic Radiology and Organ Imaging (A.D.K., A.T.A., E.H.Y.Y.), Anatomical and Cellular Pathology (G.M.K.T.), and Surgery (A.C.V., A.C.v.H.), Faculty of Medicine, Chinese University of Hong Kong, Prince of Wales Hospital, Ngan Shing St, Shatin, New Territories, Hong Kong SAR, China; and Oral Maxillofacial Surgery Center, St Teresa’s Hospital, Kowloon, Hong Kong SAR, China (E.W.H.T.). Received January 29, 2003; revision requested April 21; final revision received July 17; accepted August 13. Address correspondence to A.D.K. (e-mail: b834756@mailserv.cuhk.hk).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Twenty-seven patients (age range, 39–85 years; mean age, 62 years) with squamous cell carcinoma in the head and neck underwent CT, MR imaging, and US. Three radiologists evaluated the images for nodal necrosis. The results of each modality were analyzed for sensitivity, specificity, and accuracy. Pathologic analysis of the surgical resection served as the reference standard. The three modalities were compared for specificity and sensitivity with the McNemar test.
RESULTS: Pathologic examination revealed 903 nodes, of which 89 were malignant. Of the malignant nodes, 43 were necrotic. Analysis of the detection of necrosis in the 89 malignant nodes showed an accuracy, sensitivity, and specificity of 92%, 91%, and 93% for CT; 91%, 93%, and 89% for MR imaging; and 85%, 77%, and 93% for US, respectively. All imaging modalities failed to depict necrotic areas of 3 mm or smaller in three nodes, and necrosis was missed in an additional seven nodes with US and in one node with CT. Necrosis could not be distinguished from other components of malignancy, such as viable tumor and scar tissue, in seven nodes (CT, 3; MR imaging, 5; US, 3). The sensitivity of both MR imaging and CT was significantly better than that of US (P = .0082 and P = .0339, respectively). There was no significant difference in sensitivity (P = .3173) between MR imaging and CT, or in the specificity of the three modalities.
CONCLUSION: MR imaging is comparable to CT for the detection of necrosis. The sensitivity of MR imaging and CT is better than that of US.
© RSNA, 2004
Index terms: Head and neck neoplasms, CT, 262.12112, 271.12112 • Head and neck neoplasms, MR, 262.1214, 271.1214 • Head and neck neoplasms, US, 262.12981, 271.12981 • Lymphatic system, diseases, 276.37
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
In most institutions, imaging of metastases in lymph nodes of the neck is performed at the same time as imaging of the primary tumor; hence, the literature is skewed toward use of computed tomography (CT) and magnetic resonance (MR) imaging. At present, CT is considered the best technique for detection of nodal necrosis (3,4); however, the use of surface coils and matrices with smaller pixels (512 pixels) produces high-quality MR images with the potential to improve the detection of necrosis. In addition, ultrasonography (US) is being used more widely for staging nodal metastases in the neck, but to our knowledge no study exists that compares US with MR imaging and CT. Thus, the aim of this study was to compare the diagnostic accuracy of CT, MR imaging, and US in the detection of necrosis in metastatic cervical nodes from patients with head and neck squamous cell carcinoma.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
The study group included 27 patients (age range, 39–85 years; mean age, 62 years). Of these patients, 22 were men (age range, 39–85 years; mean age, 65 years), and five were women (age range, 43–73 years; mean age, 62 years). Primary head and neck tumors were located in the hypopharynx, oropharynx, or both in 10 patients, in the tongue in seven patients, in the oral cavity in seven patients, and in the larynx in three patients. The study lasted from August 1999 through August 2002. Twenty-four patients in the study group underwent CT, MR imaging, and US within a 10-day interval, and three patients underwent CT, MR imaging, and US within 12, 14, and 16 days of each other.
Imaging Protocol
All patients underwent MR imaging, CT, and US. Real-time US was performed with a 5–12-MHz transducer (HDI 5000; ATL, Bothell, Wash) to obtain gray-scale images of the neck in the transverse plane. In addition, the nodes were scanned in multiple planes. MR imaging was performed with a 1.5-T magnet (Gyroscan; Philips, Eindhoven, the Netherlands). In all patients, the imaging protocol included transverse T2-weighted fast spin-echo imaging with fat suppression (repetition time msec/echo time msec, 2,000–2,800/80–120; echo train length, 15–18; section thickness, 4 mm with no intersection gap), transverse T1-weighted spin-echo imaging (425–500/12–15; section thickness, 4 mm with no intersection gap), and contrast material– enhanced transverse T1-weighted spin-echo imaging after a bolus injection of 0.1 mmol/kg of bodyweight gadodiamide (Omniscan; Nycomed, Oslo, Norway) with use of a 512 matrix. Contrast-enhanced T1-weighted imaging with fat suppression (425–500/12; section thickness, 4 mm with no intersection gap) was performed in 16 patients. Because of the length of time required for complete MR imaging, not all patients were able to undergo the whole examination. In all patients, at least one of the sequences— the contrast-enhanced T1-weighted sequence—was performed with a 14- or 20-cm surface coil. Transverse helical CT imaging (HiSpeed Advantage 9800; GE Medical Systems, Milwaukee, Wis) (collimation, 3–5 mm; pitch, 1.0 or 1.5) was performed after intravenous injection of 100 mL of iohexol (Omnipaque; Nycomed) (240 mg iodine per milliliter) or iodixanol (Visipaque; Nycomed) (240 mg iodine per milliliter) at a rate of 2 mL per second with a 40-second delay before scanning commenced.
Imaging Assessment
Images were assessed by three radiologists (A.T.A., A.D.K., E.H.Y.Y.) with 4–12 years experience in head and neck imaging, and all radiologists had at least 4 years of experience with all three imaging modalities. For any given modality any one of the three radiologists performed the evaluation, but for any given patient the images obtained with each modality were evaluated by a different radiologist who was blinded to the results of the other two tests. Criteria for the diagnosis of necrosis on CT, MR, and US images are shown in Table 1.
|
Statistical Analysis
The sensitivity, specificity, positive and negative predictive values, and accuracy of each modality for the detection of necrosis were calculated from the total group of benign and malignant nodes and from the subgroup of malignant nodes only. The results of the three modalities were compared (US vs CT, US vs MR imaging, and MR imaging vs CT). The McNemar test was used for the results restricted to the true-positive nodes to compare sensitivity and was used for the results restricted to the true-negative nodes to compare specificity. A P value of less than .05 was considered to indicate a statistically significant difference. To take into account possible dependency issues of multiple nodes per patient influencing the results, sensitivity and specificity of the three imaging modalities were compared again by using analysis of patients with malignant nodes rather than analysis of total number of malignant nodes.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Results with Each Imaging Modality
The results of CT, MR imaging, and US in the detection of necrosis in the group of 89 malignant nodes and in the total group of 903 benign and malignant nodes are shown in Table 2.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Analysis of the detection of necrosis in the group of 89 malignant nodes showed the sensitivity of CT was significantly better than that of US (P = .0339), but there was no significant difference in the specificity (P > .99). Analysis of the detection of necrosis in the group of 19 patients with malignant nodes showed the sensitivity of CT was no longer significantly better than that of US (P = .1025) and that there was no significant difference in the specificity (P > .99).
MR imaging versus US.—Among the 89 malignant nodes in 19 patients, concordant results for necrosis (Fig 5) were present in 33 true-positive nodes in 14 patients, 41 true-negative nodes in 16 patients, three false-positive nodes in three patients, and three false-negative nodes in three patients. Discordant results for necrosis were present in seven nodes in five patients in which MR images were true-positive and US images were false-negative, in no nodes or patients in which MR images were false-negative and US images were true-positive, in two nodes and patients in which MR images were false-positive and US images were true-negative, and in no nodes or patients in which MR images were true-negative and US images were false-positive.
|
|
MR imaging versus CT.—Among the 89 malignant nodes and 19 patients, concordant results for necrosis (Fig 5) were present in 39 true-positive nodes in 14 patients, 40 true-negative nodes in 15 patients, two false-positive nodes in two patients, and three false-negative nodes in three patients. Discordant results for necrosis were present in one node in one patient in which MR images were true-positive and CT scans were false-negative, in no nodes or patients in which MR images were false-negative and CT scans were true-positive, in three nodes in three patients in which MR images were false-positive and CT scans were true-negative, and in one node in one patient in which MR images were true-negative and CT scans were false-positive.
Analysis of the detection of necrosis in the group of 89 malignant nodes and in the group of 19 patients with malignant nodes showed there was no significant difference in the sensitivity (P = .3173 and P = .3173, respectively) or specificity (P = .3173 and P = .3173, respectively).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
The aim of this study was to look at the detection of necrosis in malignant nodes and not to compare the different imaging modalities for the detection of malignant nodes. The reason for this is because many studies have compared different modalities for the detection of malignant nodes, but they used the presence of any one of several criteria—usually nodal size—to determine if a node was malignant. As a result, it may not have been clear why one technique was better than another; therefore, the strengths and potential pitfalls may not have been appreciated. This is particularly important because some centers, and indeed some regions of the world, have distinct preferences for a given imaging modality. Necrosis was chosen because it is frequently found in nodal metastases from squamous cell carcinoma of the head and neck and because the identification of necrosis with imaging is a reliable sign of a metastatic node. In addition, while the frequency of necrosis increases with nodal size, 56%–63% of malignant nodes larger than 1.5 cm in diameter show necrosis (12,13) and 10%–33% of malignant nodes smaller than 1 cm in diameter show necrosis. These are nodes that may appear to be normal with other imaging criteria (12,13).
To our knowledge, no published reports have compared the diagnostic accuracy of CT, MR imaging, and US for the detection of nodal necrosis. This may be because in some centers the expertise lies in only one or two of these modalities. In addition, we found previously that it can be difficult to match individual nodes on images with the dissected pathologic specimen. For this reason, where there were multiple nodes, the radiologist scanned the surgical specimen with US to allow correct orientation of all nodes. The pathologist then dissected each node with US guidance, ensuring correct radiologic and pathologic correlation.
Contrast-enhanced CT is considered to be the best modality for identification of necrosis. A sensitivity of 74% and a specificity of 94% have been reported for areas of necrosis larger than 3 mm (1). This study showed an even greater sensitivity for necrosis, with a 91% rate of detection for all areas of necrosis. The specificity (93%–100%) was similar to that reported previously. Specificity of 100% was not reached, because only nodes with true necrosis rather than islands of tumor were considered to yield true-positive results. Fatty hilar metaplasia, which is a further cause of false-positive results at CT (1), was not encountered in this study.
MR imaging has advantages over CT for staging many of the primary squamous cell carcinomas in the head and neck. Thus, MR imaging is being used frequently to stage neck nodes during the same examination. Contrast-enhanced T1-weighted imaging sequences are required to improve the detection of necrosis at MR imaging (2,4,14,15), with some authors advocating the need for the addition of fat suppression (16). Initially, it was hoped that MR imaging would improve the detection of necrosis in malignant nodes, but the results have been disappointing to date, with several studies showing that MR imaging is inferior to CT (3,4,17,18). In a study by Yousem et al (3) the accuracy and sensitivity of MR imaging were only 86%–87% and 60%–67%, respectively, compared with 91%–96% and 83%–100%, respectively, for CT. Some of these studies used a low-strength magnet, while in others there was no pathologic confirmation of the results. We found the use of surface coils and smaller matrices (512 pixels) improved the quality of the MR images and the depiction of small foci of necrosis in malignant nodes.
In this study, we have shown that MR imaging can perform as well as CT in the detection of necrosis in metastases in lymph nodes of the neck. Diagnostic accuracy and sensitivity were similar (91%–99% and 93%, respectively, for MR imaging compared with 92%–99% and 91%, respectively, for CT), with no significant difference between the two modalities. The specificity for the total group of 903 benign and malignant nodes was also similar. When considering only the group of malignant nodes, the specificity of MR dropped to 89% compared with 93% for CT because of the ability of MR imaging to depict small foci of tumor more readily, thus increasing the number of false-positive results. This difference, however, was not statistically significant.
US is being used more widely to assess the lymph nodes of the neck, especially when indeterminate nodes are found with MR imaging or CT. The criteria for diagnosis of a malignant node with US are similar to those with other imaging modalities, but US has several additional advantages. First, US is able to depict abnormalities of hilar architecture more clearly. Second, color Doppler US provides a method for examining nodal vascularity in detail. Malignant nodes may show prominent capsular vessels in addition to or in place of the normal hilar vessels, and the vascular resistance within the malignant node is increased (19,20). Third, US can be used to guide fine-needle aspiration to provide cytologic analysis from nodes as small as 3–5 mm in diameter (21). While several studies have shown that US and US-guided fine-needle aspiration are similar or superior to MR imaging or CT for the detection of malignant nodes (18,22,23), to our knowledge no study has compared the sensitivity and specificity of US for the detection of necrosis.
Despite the routine use of US for many years at The Prince of Wales Hospital, we found that while the specificity of US was similar, the sensitivity of US (77%) was lower than that of MR imaging (93%) or CT (91%). The results for sensitivity were statistically significant for the group of 89 malignant nodes, but to avoid the potential bias resulting from patients with multiple nodes, analysis of patients with malignant nodes was also performed. These results showed that the difference between MR imaging and US remained statistically significant; however, as a result of the smaller numbers, the difference between CT and US was no longer significant, although the trend remained the same.
Pathologic-radiologic correlation of each node provided the means to compare imaging appearances directly. All three modalities were used to identify 74% of the necrotic nodes correctly. False-negative results occurred with all three modalities in nodes in which the areas of necrosis were smaller than 3 mm. This finding is in accordance with findings in previous studies (1,12), in which imaging was unable to depict necrosis in nodes smaller than 3 mm. CT did not depict one further necrotic node in a malignant node at the level of the clavicle, an area that is prone to diagnostic difficulties. The largest number of false-negative results occurred with US, accounting for its lower sensitivity. Areas of necrosis may be readily depicted on CT scans and MR images, but because they are isoechoic on US images they are rendered inconspicuous, unless there is movement within the necrotic area caused by compression with the probe. Detection with US is a particular problem when the foci of necrosis are small. While US retains several distinct advantages over MR imaging and CT in the identification of a malignant node, the sonographer should be aware of this pitfall in the diagnosis of necrosis.
False-positive results occurred where the focal areas of abnormality were due to tumor, keratinization, or fibrous scars rather than necrosis. This feature has been depicted previously with MR imaging and CT where the central area of necrosis has been shown to be caused by a mixture of necrosis, keratin, fibrous tissue, interstitial fluid, and viable tumor cells (1,16). No single imaging technique was able to distinguish between these different entities with the criteria set out in this study. At present, when staging neck nodes it is irrelevant if the focal area of abnormality is due to tumor or necrosis, because in both instances, the correct diagnosis of a malignant node will be made. The correct identification of a necrotic process within a malignant node, however, may become important in the future when the effects of tumor oxygenation with respect to tumor propagation, malignant progression, and resistance to therapy become clearer.
There are several limitations to the study. First, when planning the study, US was the limiting factor. Unlike MR images and CT scans, US images could not be evaluated after the examination, but instead had to be evaluated in real time by the radiologist performing the examination. Logistically, it was not possible for one radiologist to perform all the US examinations throughout the entire study; therefore, we could not designate one radiologist per modality for this study. Furthermore, it was not possible to have two radiologists review each US examination; therefore, interobserver variations could not be assessed. The second limiting factor was the small number of patients with malignant nodes. As a result, when dependency issues were taken into account by analyzing patients rather than total number of nodes, the sample was too small to show that the higher sensitivity of CT versus US was statistically significant.
We believe, however, that these limitations do not detract from the three main findings of the study. First, MR imaging with surface coils and matrices with smaller pixel size improves the depiction of necrosis, with results being comparable to those achieved with CT. Second, the sensitivity of US for the detection of necrosis is lower than that of MR imaging and CT because small isoechoic areas of necrosis are difficult to detect with US. Third, by using the imaging criteria outlined in this study, all imaging modalities may fail to depict areas of necrosis that are smaller than 3 mm, and all imaging modalities may be unable to distinguish tumor necrosis from other elements of malignant nodes such as the tumor itself, keratinization, and scarring.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  V. Vandecaveye, F. De Keyzer, V. Vander Poorten, P. Dirix, E. Verbeken, S. Nuyts, and R. Hermans Head and Neck Squamous Cell Carcinoma: Value of Diffusion-weighted MR Imaging for Nodal Staging Radiology, April 1, 2009; 251(1): 134 - 146. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. Tagaya, N. Kurimoto, H. Osada, and A. Kobayashi Automatic Objective Diagnosis of Lymph Nodal Disease by B-Mode Images From Convex-Type Echobronchoscopy Chest, January 1, 2008; 133(1): 137 - 142. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. Rubaltelli, S. Corradin, A. Dorigo, A. Tregnaghi, F. Adami, C. R. Rossi, and R. Stramare Automated Quantitative Evaluation of Lymph Node Perfusion on Contrast-Enhanced Sonography Am. J. Roentgenol., April 1, 2007; 188(4): 977 - 983. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. H. Choi, M. H. Han, W. K. Moon, K. R. Son, J.-K. Won, J.-H. Kim, B. J. Kwon, D. G. Na, H.-J. Weinmann, and K.-H. Chang Cervical Lymph Node Metastases: MR Imaging of Gadofluorine M and Monocrystalline Iron Oxide Nanoparticle-47 in a Rabbit Model of Head and Neck Cancer Radiology, December 1, 2006; 241(3): 753 - 762. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. Kawai, M. Sumi, and T. Nakamura Turbo Short {tau} Inversion Recovery Imaging for Metastatic Node Screening in Patients with Head and Neck Cancer AJNR Am. J. Neuroradiol., June 1, 2006; 27(6): 1283 - 1287. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Sumi, M. Van Cauteren, and T. Nakamura MR Microimaging of Benign and Malignant Nodes in the Neck. Am. J. Roentgenol., March 1, 2006; 186(3): 749 - 757. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
