604 AJR:191, August 2008

from these two echo lines undergo process­ing, and ultimately an elastographic image (elasto gram) appears on the monitor. There are two types of elastograms, gray­scale and color. The hard and soft areas (i.e., areas of high and low elasticity, respectively) appear in the gray­scale elastogram as dark and bright [5, 6]. In the color elastogram, in­creasing tissue hard ness appears in ascending order as red, yel low, green, and blue. These colors represent the relative hardness of the tissues in the elastogram.

Sonographic elastography has been used to examine several organs: the breast [7, 8], thyroid [5], prostate [9], cervix [10], liver [11], and so forth. Recently, an article relating their initial experience in the diag­nosis of cervical lymph node metastasis us ing gray­scale sonographic elastography was published by Lyshchik et al. [6]. To our knowledge, however, no English­language articles have documented the evaluation of cervical lymph nodes using color sonograph­ic elastography.

Accuracy of Sonographic Elastography in the Differential Diagnosis of Enlarged Cervical Lymph Nodes: Comparison with Conventional B-Mode Sonography

Farzana Alam1

Kumiko Naito2,3

Jun Horiguchi2 Hiroshi Fukuda2

Toshihiro Tachikake1

Katsuhide Ito1

Alam F, Naito K, Horiguchi J, Fukuda H, Tachikake T, Ito K

1Department of Radiology, Division of Medical Intelligence and Informatics, Graduate School of Biomedical Sciences, Hiroshima University, Hiroshima, Japan.

2Department of Clinical Radiology, Hiroshima University Hospital, 1-2-3, Kasumi-cho, Minami-ku, Hiroshima, 734-8551, Japan. Address correspondence to J. Horiguchi (horiguch@hiroshima-u.ac.jp).

3Present address: Hiroshima Atomic Bomb Casualty Council, Hiroshima, Japan.

Head and Neck Imaging • Or ig ina l Research

AJR 2008; 191:604–610

0361–803X/08/1912–604

© American Roentgen Ray Society

Elasticity (hardness) is a mech­anical tissue characteristic that prevents tissue displacement un­der pressure. It varies in different

types of tissue (fat, collagen, and so forth) and in the same tissue in different pathologic states (inflammatory, malignant). During the past few years, sonographic elastography, MR elastography [1, 2], and some other tech­niques have performed digital measurements of tissue hardness. In sonographic elasto­graphy, image represent ation of tissue hard­ness can be obtained using a conventional sonography machine with special software and a conventional ultrasound probe [3, 4].

In brief, sonographic elastography works in the following steps: first, elastography receives digitized radiofrequency echo lines from the tissue; second, it gives slight com­pression to the tissue by the transducer along the radiation axis to make some displace­ment; and third, it receives a second, post­compression digitized radiofrequency echo line from the same tissue [4]. Then the data

Keywords: elastography, lymph nodes, metastasis, sonography

DOI:10.2214/AJR.07.3401

Received November 9, 2007; accepted after revision February 17, 2008.

OBJECTIVE. The purpose of our study was to evaluate the diagnostic performance of sonographic elastography and B­mode sonography individually and combined in the differentiation of reactively and metastatically enlarged cervical lymph nodes.

SUBJECTS AND METHODS. Eighty­five lymph nodes (metastatic, n = 53; reactive, n = 32) from 37 patients were examined by both elastography and B­mode sonography in this prospective study. Elastographic patterns were determined on the distribution and percentage of the lymph node area with high elasticity (hard), with pattern 1 being an absent or very small hard area to pattern 5, a hard area occupying the entire lymph node. The cutoff line for reactive versus metastatic was set between patterns 2 and 3; patterns 3–5 were considered metastatic. B­mode sonographic diagnosis was based on the sum of scores for five criteria: short­axis diameter, shape, border (regular or irregular), echogenicity (homogeneous or inhomo geneous), and hilum (present or absent). The cutoff line for reactive versus metastatic was set between scores 6 and 7; scores 5 and 6 were considered reactive, and scores 7–10, metastatic.

RESULTS. Sensitivity, specificity, and accuracy of B­mode sonography were 98%, 59%, and 84%, respectively; 83%, 100%, and 89% for elastography; and 92%, 94%, and 93% for the combined evaluation.

CONCLUSION. The combination of highly specific elastography with highly sensitive conventional B­mode sonography has the potential to further improve the diagnosis of metastatic enlarged cervical lymph nodes.

Alam et al.Sonographic Elastography of Cervical Lymph Nodes

Head and Neck ImagingOriginal Research

Dow

nloa

ded

from

ww

w.a

jron

line.

org

by U

nive

rsity

of

Was

hing

ton-

Inte

rnet

on

08/2

5/14

fro

m I

P ad

dres

s 12

8.95

.104

.66.

Cop

yrig

ht A

RR

S. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved

AJR:191, August 2008 605

Sonographic Elastography of Cervical Lymph Nodes

Of the 400–450 lymph nodes in the human body, the head and neck contain 60–70. These nodes show reactive enlargement due to infection (e.g., infection of upper aero­digestive tract). They also undergo enlarge­ment when they are secondarily involved in head and neck cancer. Sometimes metastatic cervical lymphadenopathy appears as the first symptom in patients having malignancy in the head and neck, lung, breast, and so forth. Differentiation between reactive and metastatic lymph adenopathy is vital, and one of the differentiating criteria is hardness (elasticity) of the lymph node.

The purpose of this study was to investigate the accuracy of conventional sonography, sonographic elastography, and their combined evaluation for the differentiation between reactive and metastatic enlarged cervical lymph nodes.

Subjects and MethodsPatients

This was a prospective study. From January to November 2006, 109 lymph nodes of 59 con sec­utive patients who were referred for sonography of enlarged cervical lymph nodes were enrolled. Written informed consent was received from all pa tients for undergoing both conventional B­ mode sonography and elastography. The 85 lymph nodes (reactive, n = 32; metastatic, n = 53) from 37 patients (25 males, 12 females; mean age ± SD, 55 ± 22 years old; age range, 11–92 years) with (n = 25) and with out (n = 12) malignancy in which the final diag nosis was known, were used for evaluation.

Final DiagnosisLymph nodes were determined to be reactive

on the basis of histopathologic findings or on the basis of clinical findings such as enlarged tender lymph node, increased C­reactive protein, leuko­cytosis, decreased size of the lymph node after antibiotic treatment, and the absence of known malignancy. The reference standard for the final diagnosis of metastatic lymph nodes was histo­pathologic findings or imaging findings suggest­ing central necrosis—that is, hypodensity on contrast­enhanced CT [12] or hypointensity on contrast­enhanced T1­weighted MRI in patients with malignancy.

Equipment and ScanningOne radiologist who had 18 years’ experience

with conventional sonography and was a novice in elastography performed the sonography. The pa­tients underwent both B­mode and elastographic sonography in the supine position with a digital sonography scanner (EUB 8500, Hitachi Medical) equipped with the Real­time Tissue Elastography software (Hitachi Medical); the probe was a conventional linear probe with a 13­MHz trans­ducer (EUP­L65, Hitachi Medical).

For each lymph node, B­mode images were obtained first. Then, changing the system into the elastography mode, real­time freehand elasto­graphy was performed using the same probe for an additional 1–2 minutes. For elastography, com­pression with light pressure followed by de­compression was repeated until a stable image was obtained—that is, nearly the same size and color distribution of the target area in several consecutive images. Direction of the compression

was upward and downward (along the radiation axis). Frequency of compression appeared on a scale on the monitor. Real­time elastographic and B­mode images simultaneously appeared as a two­panel image. Figure 1 shows a typical image displayed on the monitor during elastography, in which the elastogram appears in a region­of­interest (ROI) box. The size of the box was determined to be the target lymph node including surrounding tissue in almost the same proportion while avoiding tissues (bone, blood vessel) that might disturb the appropriate analysis of the relative hardness of the target lymph node. To maintain unbiased reading, 85 sets of elasto ­grams and corresponding B­mode images with­out any patient information were collected into two separate slide shows (Microsoft PowerPoint, 2003) in random order.

EvaluationTo minimize recall bias, 1 month after data

collection was completed, score­based reevalua­tion of B­mode and elastographic images was done by one of the authors who was blinded to other information. Less than 1 minute was requir­ed for the evaluation of each image (elastography and B­mode).

B-mode evaluation—B­mode images were evaluated on the criteria of size, shape, border, echogenicity, and hilum. Size and shape were evaluated per region. Regional distribution was determined according to the 1997 American Joint Committee on Cancer (AJCC) criteria for lymph nodes [13], in which neck lymph nodes are divided into seven levels. Lymph nodes in level 7 were excluded because scanning them was not possible with sonography. For the criterion of size, the short­axis diameter (in mm) of the lymph nodes was evaluated. For shape, the ratio of the short­axis diameter to the long­axis diameter (S/L axis ratio) was used. For each lymph node, we deter­mined the cutoff values of short­axis diameter and S/L axis ratio that gave the best accuracy. In cases of level 4 (AJCC classification) lymph nodes in which two cutoff values had the highest accuracy, we used receiver operating characteristic (ROC) curve analysis to select one for our purposes. Scores were determined for five criteria: the short­axis diameter (diameter ≤ cutoff value, score of 1; diameter > cutoff value, score of 2), the S/L axis

Fig. 1—Appearance of monitor during elastography examination of 52-year-old man. Elastogram appears on left side of box and is superimposed on corresponding B-mode image. Corresponding B-mode image is displayed on right side of monitor. Color scale of tissue elasticity and frequency scale appear in center (upper and lower, respectively). Final diagnosis was metastasis.

Dow

nloa

ded

from

ww

w.a

jron

line.

org

by U

nive

rsity

of

Was

hing

ton-

Inte

rnet

on

08/2

5/14

fro

m I

P ad

dres

s 12

8.95

.104

.66.

Cop

yrig

ht A

RR

S. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved

606 AJR:191, August 2008

Alam et al.

ratio (ratio ≤ 0.6, score of 1; > 0.6, score of 2), the border (regular, 1; irregular, 2), echogenicity (homo geneous, 1; inhomogeneous, 2), and hilum (present, 1; absent, 2) [6, 12–21]. The score for each lymph node was determined by summing scores for all criteria. A statistically supported cutoff line between reactive and metastatic was set between scores 6 and 7, depending on the best accuracy. Scores 5 and 6 were determined to be reactive, and scores 7–10, metastatic.

Elastogram evaluation—After observing all the elastograms, two authors together subjectively decided five patterns for the elastograms depend­ing on the distribution of the blue (i.e., hard) area in the lymph node. Elastographic patterns were deter mined on the distribution and percentage of the lymph node area with high elasticity (hard): pattern 1, an absent or a very small hard (i.e., blue) area; pattern 2, hard area < 45% of the lymph node; pattern 3, hard area ≥ 45%; pattern 4, peripheral hard and central soft areas; pattern 5, hard area occupying entire lymph node with or without a soft rim. The patterns 1, 2, 3, 4, and 5 were assigned scores of 2, 4, 6, 8, and 10, respectively. The outline of the lymph node was drawn, and the percentage of blue area was measured on an off­line computer using image analysis software “Image J” developed by the National Institutes of Health [22]. On average, approximately 1–2 min utes was required for the measurement of the blue area of each lymph node. A statistically sup ported cutoff line at 45% blue between patterns 2 and 3 was obtained from the ROC (Table 1 and Figs. 2–7).

Combined evaluation—The combined score for each lymph node was the sum of the B­mode

and elastographic scores. A statistically supported cutoff line between metastatic and reactive was set between scores 12 and 13, depending on the best accuracy. Scores 7–12 were determined to be reactive, and scores 13–20, metastatic.

For B­mode, elastography, and the combined evaluations, the precision of the diagnostic parameters (sensitivity, specificity, accuracy, and positive and negative predictive values) was expressed using 95% CI. The areas under the ROC curves were calculated.

ResultsFinal Diagnosis

The prevalence of lymph node metastasis was 62% (53/85). Among the metastatic lymph nodes, 36 had primary squamous cell car ci­

noma (SCC), 13 had primary thyroid carci­noma, two had their primary site in the breast, one was in the lung, and one was unknown. The lymph nodes were finally diagnosed as reactive histopathologically (n = 2) and clinically (n = 30). The lymph nodes were diagnosed as metastatic by histo pathology (n = 26), CT (n = 24), and MRI (n = 3). By AJCC levels, 17 nodes were in level 1, 26 in level 2, 21 in level 3, nine in level 4, and 12 in level 5.

B-Mode SonographyCutoff short­axis diameters according to

AJCC levels were 11 mm in level 1, 13 mm in level 2, 5 mm in level 3, 9 mm in level 4, and 7 mm in level 5. The cutoff S/L axis ratio was 0.6; ≤ 0.6 was reactive and > 0.6 was metastatic. The accuracy for individual B­ mode criteria was 84% for size, 56% for S/L axis ratio, 67% for border, 86% for hilum, and 78% for echogenicity (Table 2).

ElastographyDuring elastography, compression at a fre­

quency of 3 or 4 on the frequency scale was needed to obtain stable elastographic fea­tures. In elastogram evaluation, all 32 reac­tive lymph nodes showed patterns 1 and 2 (re active). Among 53 metastatic lymph nodes,

TABLE 1: Patterns and Scoring System on Elastographic Findings

Pattern Score Description Elastographic Diagnosis Figure

1 2 Absent or very small blue area(s) Reactive 2, 3

2 4 Small scattered blue areas, total blue area < 45%

Reactive 2, 4

3 6 Large blue area(s), total blue area ≥ 45% Malignant 2, 5

4 8 Peripheral blue area and central green area, suggesting central necrosis

Malignant 2, 6

5 10 Blue area with or without a green rim Malignant 2, 7

Fig. 2—Drawings show typical diagrammatic appearance of five patterns of lymph nodes. Elastographic patterns were determined on distribution and percentage of lymph node area having high elasticity (hard): pattern 1, absent or small hard area; pattern 2, hard area < 45% of lymph node; pattern 3, hard area ≥ 45%; pattern 4, peripheral hard and central soft areas; pattern 5, hard area occupying entire lymph node. Increasing tissue hardness appears in ascending order as red, yellow, green, and blue.

Dow

nloa

ded

from

ww

w.a

jron

line.

org

by U

nive

rsity

of

Was

hing

ton-

Inte

rnet

on

08/2

5/14

fro

m I

P ad

dres

s 12

8.95

.104

.66.

Cop

yrig

ht A

RR

S. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved

AJR:191, August 2008 607

Sonographic Elastography of Cervical Lymph Nodes

44 (83%) had patterns 3, 4, and 5 (metastatic). Two (4%) and seven (13%) metastatic lymph nodes showed patterns 1 and 2, respectively. Con cerning elasticity pattern 4 (i.e., sug­gestive of central necrosis), 12 of 13 (92%) lymph nodes were metastasized from SCC and one (8%) from an unknown origin.

Diagnostic PerformanceThe diagnostic performances of B­mode

sonography, elastography, and the combined test are shown in Table 3. B­mode sono­graphy showed sensitivity of 98% (95% CI, 94–100%), specificity of 59% (42–76%), and accuracy of 84% (76–91%). Elasto graphy

showed sensitivity of 83% (73–93%), speci­ficity of 100% (100–100%), and accu racy of 89% (83–96%). The combined evaluation showed sensitivity of 92% (85–100%), specificity of 94% (85–100%), and accuracy of 93% (88–98%). The ROC curves for B­ mode, elastography, and the combined evalu­

Fig. 3—Transverse sonogram of level 1 lymph node in 30-year-old man with left submandibular lymphadenopathy. Elastography image on left shows pattern 1, absent or small hard area. B-mode sonographic image on right shows score of 5, reactive. Final diagnosis from clinical and serologic findings was reactive lymph node.

Fig. 4—Longitudinal sonogram of level 2 lymph node in 33-year-old woman with left posterior auricular lymphadenopathy. Elastography image on left shows pattern 2, hard area < 45% of node. B-mode sonographic image on right shows score 5, reactive. Final diagnosis from clinical and serologic findings was reactive lymph node.

Fig. 6—Longitudinal sonogram of level 5 lymph node in 52-year-old man with nasopharyngeal carcinoma. Elastography image on left shows pattern 4, peripheral hard and central soft areas. B-mode sonographic image on right shows score 7, metastatic. Final diagnosis by CT was metastatic lymph node.

Fig. 5—Transverse sonogram of level 2 lymph node in 79-year-old man with laryngeal carcinoma. Elastography image on left shows pattern 3, hard area ≥ 45%. B-mode sonographic image on right shows score 7, metastatic. Final diagnosis by histopathology was metastatic lymph node.

Dow

nloa

ded

from

ww

w.a

jron

line.

org

by U

nive

rsity

of

Was

hing

ton-

Inte

rnet

on

08/2

5/14

fro

m I

P ad

dres

s 12

8.95

.104

.66.

Cop

yrig

ht A

RR

S. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved

608 AJR:191, August 2008

Alam et al.

ation in the differentiation of reactive and metastatic lymph nodes are shown in Figure 8. The areas under the curves for B­mode sonography, elastography, and the combined evaluation were 0.901, 0.873, and 0.970, respectively.

DiscussionTo our knowledge, this is the first English­

language study documenting the accuracy of color sonographic elastography individually and combined with B­mode sonography in the differential diagnosis of reactive versus meta­static cervical lymph nodes. Elasto graphy provides high specificity; especially when combined with B­mode sonography, elas tog­raphy increases the accuracy of sono graphy for enlarged cervical lymph nodes.

B-Mode SonographySeveral studies have been published on the

diagnosis of metastatic cervical lymph node by B­mode sonography [12, 14–21]; however, specific criteria for distinguishing reactive from metastatic cervical lymph nodes are not yet clear. The cutoff short­axis diameter for level 2 lymph nodes was two times higher than that for level 3. This suggests that, for the evaluation of cervical lymph nodes, a common cutoff diameter should not be set for all levels. In a study by Lyshchik et al. [6] using the fixed cutoff diameter of 8 mm, the accuracy of the short­axis diameter was 65%, whereas in our study it was 84%. With respect to shape, our calculation gave the best accuracy at S/L axis ratio ≤ 0.6 for reactive and > 0.6 for metastatic nodes, which supports the findings of round shape for metastatic nodes in previous studies [6, 14, 15]. Very few studies have evaluated lymph node borders. Ying et al. [16], Ahuja and Ying [17], and Ying et al. [21] evaluated lymph node borders as sharp and unsharp. In this study, we evaluated regular and irregular lymph node borders as a criterion of reactive and metastatic lymph nodes, respectively, that showed 67% accuracy. The presence or absence of a hilum has been reported to be an important criterion for lymph node diagnosis [6, 18]. In our study also, this showed the best accuracy (86%).

Fig. 7—Longitudinal sonogram of level 3 lymph node in 59-year-old man with thyroid carcinoma. Elastography image on left shows pattern 5, hard area occupying entire node. B-mode sonographic image on right shows score 8, metastatic. Final diagnosis by histopathology was metastatic lymph node.

TABLE 3: Diagnostic Performance of Elastography, B-Mode Sonography, and Combined Evaluation

Performance Measure B-Mode Sonography Elastography Combined

Sensitivity (%) 98 (94–100) 83 (73–93) 92 (85–100)

Specificity (%) 59 (42–76) 100 (100–100) 94 (85–100)

Accuracy (%) 84 (76–91) 89 (83–96) 93 (88–98)

Positive predictive value (%) 80 (70–89) 100 (100–100) 96 (91–100)

Negative predictive value (%) 95 (85–100) 78 (65–91) 88 (77–99)

Note—Values in parentheses are 95% CIs (in percentages).

TABLE 2: Results According to Criteria for Reactive and Metastatic Lymph Nodes

Sonography Criteria

Lymph Nodes

Accuracy (%)Reactive Metastatic

B-mode

Size of short-axis diameter 84

≤ Cutoff value 28 (88) 10 (19)

> Cutoff value 4 (12) 43 (81)

S/L axis ratio 56

≤ 0.6 18 (56) 23 (43)

> 0.6 14 (44) 30 (57)

Border 67

Regular 24 (75) 20 (38)

Irregular 8 (25) 33 (62)

Hilum 86

Present 26 (81) 6 (11)

Absent 6 (19) 47 (89)

Echogenicity 78

Homo geneous 21 (66) 8 (15)

Inhomo geneous 11 (34) 45 (85)

Elastography pattern 89

1 19 (59) 2 (4)

2 13 (41) 7 (13)

3 0 21 (40)

4 0 13 (24)

5 0 10 (19)

Note—Numbers in parentheses are percentages. S/L axis ratio = ratio of short-axis diameter to long-axis diameter.

Dow

nloa

ded

from

ww

w.a

jron

line.

org

by U

nive

rsity

of

Was

hing

ton-

Inte

rnet

on

08/2

5/14

fro

m I

P ad

dres

s 12

8.95

.104

.66.

Cop

yrig

ht A

RR

S. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved

AJR:191, August 2008 609

Sonographic Elastography of Cervical Lymph Nodes

ElastographyStiffness of cervical lymph nodes was

independently assessed both quantitatively and qualitatively by Lyshchik et al. [6] using external sonographic gray­scale elastog­raphy. Those authors assessed the diag nostic potential of qualitative criteria—lymph node visibility, relative brightness, margin regu­larity, and margin definition—as well as the quantitative criterion strain index, which was obtained by comparing the absolute values of lymph node strain with the absolute values of surrounding muscle strain. Among these criteria, strain index greater than 1.5 was the most useful in metastatic lymph node classification, having 98% specificity, 85% sensitivity, and 92% accuracy. We obtained a similar level of results (100% specificity, 83% sensitivity, and 89% accuracy) using the combination of both qualitative and quanti­tative criteria. However, we propose the combined criteria, for the following reasons: The presence of a central green area in pattern 4 was considered to be central necrosis. Our result showing 12 of the 13 pattern 4 lymph nodes had primary SCC supports this con­sideration. Furthermore, we considered a large blue area in a scattered distribution to suggest the manifestation of focal cortical metastatic invasion and there fore categorized this appearance as pattern 3. This criterion did not show any false­positive results.

In endoscopic sonographic elastography of lymph nodes, Giovannini et al. [23] re­ported 100% sensitivity and 50% specificity without using any qualitative pattern analy­sis, whereas Saftoiu et al. [24] reported 91.7% sensitivity and 94.4% specificity using a pattern analysis. These facts also indicate the importance of qualitative pattern analysis. However, another pattern of elastographic lymph node analysis has been reported in which only the percentage of blue area was used as a criterion [25].

The greatest advantage of elastography is its high specificity, which has been found not only in our study but also in other studies [6, 24]. Because of this, elastography may re duce unnecessary invasive procedures for the diag­nosis of cervical lymph node metastasis.

Combined EvaluationAccuracy and area under the ROC curve

showed that the diagnostic power of the com­bined evaluation was higher than that of the individual evaluations. This suggests that the examination can be best performed when B­mode sonography, with its high sensitivity, and elastography, with its high specificity, are complementary. The scope of our study was to test the performance of elastography and the combined test compared with B­mode sonography alone. In clinical practice, how­ever, sonography practitioners should use both methods in real time.

Gray-Scale and Color ElastographyIn the gray­scale elastography study by

Lyshchik et al. [6], postprocessing of radio­frequency images was done on an off­line computer. In our color elastogram study, postprocessing was done in real time, which made the process quicker. Although our study was performed on an off­line computer, in clinical practice real­time color elastog raphy seems to have an advantage over gray­scale. Real­time diagnosis needs further investigation.

LimitationsThis study had some limitations. First, our

data sample was small; we used the lymph node as our unit, not the patient. Second, we included only enlarged lymph nodes, whereas elastographic findings of normal­sized lymph nodes should also be explored. Most meta­static lymph nodes in our study were in patients who had either primary SCC or

thyroid cancer. Lymph nodes with primary malignancy or metastases from other prim­ary malignancies such as melanoma or breast cancer were not included. Third, the final diagnosis of almost all reactive lymph nodes was done based on follow­up findings instead of histopathologic findings. Fourth, evalu­ation was done on an off­line computer from single images, which would not be practical for real­time evaluation.

Although elastography is a good improve­ment in the field of sonography, we still need to address one limitation for the future. Our evaluation had to be done in only one section of the lymph node. Assessment of the stiff­ness of the whole lymph node, by volu metric measurement, may increase the diag nostic performance of elastography.

In conclusion, elastography, with its high specificity, can improve the performance of sonography for the diagnosis of enlarged metastatic cervical lymph nodes. The com­bined evaluation of elastography and B­ mode sonography offers the strongest diag­nostic power, and we recommend it in clini­cal practice.

AcknowledgmentsWe thank professors Megu Ohtaki and

Keiko Otani, from the Department of Envi­ronmetrics and Bio­Medical Informatics, Research Institute for Radiation Biology and Medicine, for their contribution in sta­tistical analysis.

References 1. Glaser KJ, Felmlee JP, Manduca A, Kannan Mar­

iappan Y, Ehman RL. Stiffness­weighted mag­

netic resonance imaging. Magn Reson Med 2006;

55:59–67

2. Huwart L, Peeters F, Sinkus R, et al. Liver fibro­

sis: non­invasive assessment with MR elastogra­

phy. NMR Biomed 2006; 19:173–179

3. Ophir J, Cespedes I, Ponnekanti H, Yazdi Y, Li X.

Elastography: a quantitative method for imaging

the elasticity of biological tissues. Ultrason Imag-

ing 1991; 13:111–134

4. Ophir J, Garra B, Kallel F, et al. Elastographic

imaging. Ultrasound Med Biol 2000; 26[suppl

1]:S23–S29

5. Lyshchik A, Higashi T, Asato R, et al. Thyroid

gland tumor diagnosis at US elastography. Radi-

ology 2005; 237:202–211

6. Lyshchik A, Higashi T, Asato R, et al. Cervical

lymph node metastases: diagnosis at sonoelastog­

raphy—initial experience. Radiology 2007; 243:

258–267

7. Itoh A, Ueno E, Tohno E, et al. Breast disease:

1

0.8

0 0.2 0.4

1 – Specificity

Sen

siti

vity

0.6 0.8 1

0.6

0.4

0.2

0

Fig. 8—Receiver operating characteristic curve for elastography (), B-mode sonography (), and combined evaluation (). Area under the curve for combined evaluation (0.970) is higher than that for B-mode sonography (0.901) or elastography (0.873).

Dow

nloa

ded

from

ww

w.a

jron

line.

org

by U

nive

rsity

of

Was

hing

ton-

Inte

rnet

on

08/2

5/14

fro

m I

P ad

dres

s 12

8.95

.104

.66.

Cop

yrig

ht A

RR

S. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved

610 AJR:191, August 2008

Alam et al.

clinical application of US elastography for diag­

nosis. Radiology 2006; 239:341–350

8. Thomas A, Fischer T, Frey H, et al. Real­time

elastography: an advanced method of ultrasound—

first results in 108 patients with breast lesions.

Ultrasound Obstet Gynecol 2006; 28: 335–340

9. Taylor LS, Rubens DJ, Porter BC, et al. Prostate

cancer: three­dimensional sonoelastography for

in vitro detection. Radiology 2005; 237:981–985

10. Thomas A, Kummel S, Gemeinhardt O, Fischer T.

Real­time sonoelastography of the cervix: tissue

elasticity of the normal and abnormal cervix.

Acad Radiol 2007; 14:193–200

11. Friedrich­Rust M, Ong MF, Herrmann E, et al.

Real­time elastography for noninvasive assess­

ment of liver fibrosis in chronic viral hepatitis.

AJR 2007; 188:758–764

12. van den Brekel MW, Stel HV, Castelijns JA, et al.

Cervical lymph node metastasis: assessment of

radiologic criteria. Radiology 1990; 177:379–384

13. Castelins JA, van den Brekel MW, Mukherji SK,

Lameris JS. Ultrasound of the neck. In: Som PM,

Curtin HD, eds. Head and neck imaging. Boston,

MA: Mosby, 2003:1935–1953

14. Vassallo P, Wernecke K, Roos N, Peters PE. Dif­

ferentiation of benign from malignant superficial

lymphadenopathy: the role of high­resolution US.

Radiology 1992; 183:215–220

15. Takashima S, Sone S, Nomura N, Tomiyama N,

Kobayashi T, Nakamura H. Nonpalpable lymph

nodes of the neck: assessment with US and US­

guided fine­needle aspiration biopsy. J Clin Ul-

trasound 1997; 25:283–292

16. Ying M, Ahuja A, Metreweli C. Diagnostic accu­

racy of sonographic criteria for evaluation of cer­

vical lymphadenopathy. J Ultrasound Med 1998;

17:437–445

17. Ahuja A, Ying M. An overview of neck node

sonography. Invest Radiol 2002; 37:333–342

18. Leboulleux S, Girard E, Rose M, et al. Ultrasound

criteria of malignancy for cervical lymph nodes in

patients followed up for differentiated thyroid

cancer. J Clin Endocrinol Metab 2007; 92: 3590–

3594

19. Jank S, Robatscher P, Emshoff R, Strobl H, Gojer

G, Norer B. The diagnostic value of ultrasonogra­

phy to detect occult lymph node involvement at

different levels in patients with squamous cell car­

cinoma in the maxillofacial region. Int J Oral

Maxillofac Surg 2003; 32:39–42

20. van den Brekel MW, Castelijns JA, Snow GB. The

size of lymph nodes in the neck on sonograms as

a radiologic criterion for metastasis: how reliable

is it? Am J Neuroradiol 1998; 19:695–700

21. Ying M, Ahuja A, Brook F. Sonographic appear­

ances of cervical lymph nodes: variations by age

and sex. J Clin Ultrasound 2002; 30:1–11

22. Image J: image processing and analysis in Java.

rsb.info.nih.gov/ij/. Accessed January 11, 2008

23. Giovannini M, Hookey LC, Bories E, Pesenti C,

Monges G, Delpero JR. Endoscopic ultrasound

elastography: the first step towards virtual biopsy?

Preliminary results in 49 patients. Endoscopy

2006; 38:344–348

24. Saftoiu A, Vilmann P, Hassan H, Gorunescu F.

Analysis of endoscopic ultrasound elastography

used for characterisation and differentiation of

benign and malignant lymph nodes. Ultraschall

Med 2006; 27:535–542

25. Furukawa MK, Kubota A, Hanamura H, Furuka­

wa M. Clinical application of real­time tissue

elastography to head and neck cancer: evaluation

of cervical lymph node metastasis with real­time

tissue elastography. Nippon Jibiinkoka Gakkai

Kaiho 2007; 110:503–505

Dow

nloa

ded

from

ww

w.a

jron

line.

org

by U

nive

rsity

of

Was

hing

ton-

Inte

rnet

on

08/2

5/14

fro

m I

P ad

dres

s 12

8.95

.104

.66.

Cop

yrig

ht A

RR

S. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved