Neuroradiology/Head and Neck Imaging
Original Research
Ultrasound Elastography Using Carotid Artery Pulsation in the Differential Diagnosis of Sonographically Indeterminate Thyroid Nodules
Audio Available | 
Share
OBJECTIVE. The purpose of this study was to evaluate the diagnostic performance of gray-scale ultrasound and a new method of thyroid ultrasound elastography using carotid artery pulsation in the differential diagnosis of sonographically indeterminate thyroid nodules.
MATERIALS AND METHODS. A total of 102 thyroid nodules with indeterminate gray-scale ultrasound features from 102 patients (20 males and 82 females; age range, 16–74 years; mean age, 51 years) were included. The gray-scale ultrasound images of each nodule were reviewed and assigned a score from 1 (low) to 5 (high) according to the possibility of malignancy. Ultrasound elastography was performed using carotid pulsation as the compression source. The elasticity contrast index (ECI), which quantifies local strain contrast within a nodule, was automatically calculated. The radiologist reassessed the scores after concurrently reviewing gray-scale ultrasound and elastography. ROC curve analysis was used to evaluate the diagnostic performances of each dataset and to compare the AUC (Az) values of gray-scale ultrasound score alone, ECI alone, and a combined assessment.
RESULTS. Significantly more malignant thyroid nodules were hypoechoic than benign nodules (p = 0.014). The ECI was significantly higher in malignant nodules than in benign thyroid nodules. The Az values of each dataset were 0.755 (95% CI, 0.660–0.835) for gray-scale ultrasound score, 0.835 (0.748–0.901) for ECI, and 0.853 (0.769–0.915) for a combined assessment. The Az value for a combined assessment of the gray-scale ultrasound score and the ECI was significantly higher than that for the gray-scale ultrasound score alone (p = 0.022).
CONCLUSION. Combined assessment with gray-scale ultrasound and elastography using carotid artery pulsation is helpful for characterizing sonographically indeterminate thyroid nodules as benign or malignant.
Keywords: carotid artery pulsation, elastography, thyroid nodule, thyroid ultrasound
Thyroid nodules are a common finding with a prevalence that ranges up to 50% according to autopsy data [1]. Although most thyroid nodules are benign, the overall incidence of thyroid cancer is as high as 5–15% [2]. Several ultrasound features are widely used to diagnose malignant thyroid nodules, such as taller-than-wide shape, spiculated margin, marked hypoechogenicity, microcalcification, and macrocalcification [3, 4]. The sensitivity and specificity of ultrasound alone are in the ranges of 52–81% and 54–83%, respectively [5]. For thyroid nodules with suspicious findings on ultrasound, ultrasound-guided fine-needle aspiration (FNA) using a minimally invasive procedure is currently the standard test for diagnosis. However, ultrasound-guided FNA results are persistently nondiagnostic for 5–15% of thyroid nodules, and 15–25% of aspirates yield indeterminate or suspicious results [6].
If solid thyroid nodules had no suspicious ultrasound findings, they underwent selective FNA or observation. Whether FNA should be performed for benign-looking thyroid nodules is determined according to nodule size, patient age, and guidelines [7–9]. There has been little investigation about whether nodules should be biopsied if there are more predictive markers in thyroid nodules with indeterminate ultrasound findings.
Ultrasound elastography is a relatively new technique developed to noninvasively obtain information about tissue stiffness [10]. Two representative techniques have been used for thyroid lesions: strain elastography and shear-wave elastography. Strain elastography requires manual compression, and the subjective power could affect the elasticity. Shear-wave elastography is operator-independent because it does not require manual compression, but arterial pulsation pushing the thyroid gland independently may cause noise and may influence the signal. A new thyroid elastography technique that uses carotid artery pulsation as the internal compression source has been developed to overcome these limitations. This new technique eliminates the variability of different compression levels applied by different operators and produces similar signal intensities between frames [11, 12]. There have been few reports regarding the diagnostic performance of this new technique.
The purpose of our study was to evaluate the diagnostic performance of gray-scale ultrasound and a new thyroid ultrasound elastography technique that uses carotid artery pulsation for the differential diagnosis of sonographically indeterminate thyroid nodules as benign or malignant.
| Materials and Methods |   |
|---|
The institutional review board approved this retrospective study and did not require informed consent from patients to review images and clinical records.
Between March 2012 and August 2012, 419 consecutive patients with thyroid nodules were imaged with gray-scale ultrasound and elastography. Of these patients, 317 were excluded because the cytohistologic diagnosis was incomplete (n = 144), cysts or spongiform nodules were present (n = 47), or the nodules had ultrasound features suspicious for malignancy(n = 126). Thyroid nodules that met one of the following criteria were included: thyroid surgery was performed or there were at least two benign cytologic results without interval change in follow-up ultrasound after at least 1 year. Finally, 102 nodules from 102 patients (20 males and 82 females; age range, 16–74 years; mean age ± SD, 51 ± 11 years) were included in this study.
Conventional ultrasound and elastographic imaging were performed using an ultrasound system (Accuvix V20, Samsung) with a 5-13–MHz linear transducer. One board-certified radiologist who specializes in thyroid imaging with 11 years of experience performed all imaging examinations. According to institutional guidelines, at least two orthogonal gray-scale ultrasound images were saved from each examination, and the final radiologic reports included information regarding the location, longest diameter, echogenicity, internal content, margin, and shape of the nodule; the presence or absence of calcifications; and the final assessment. When a thyroid nodule showed one or more gray-scale ultrasound features suspicious for malignancy (i.e., marked hypoechogenicity, microlobulated or spiculated margin, taller-than-wide shape, and micro- or macrocalcifications), it was considered suspicious for malignancy. When a thyroid nodule was cystic or spongiform and did not show any gray-scale ultrasound features suspicious for malignancy, it was considered benign. Nodules that did not meet these criteria for suspicious for malignancy or benign were considered indeterminate.
Elastography was routinely performed by the same radiologist who performed gray-scale ultrasound. The optimal representative elastography study was one that was acquired for approximately 5 seconds while the patient held his or her breath. No external compression was applied because carotid artery pulsation was used as the compression source. An elasticity parameter termed the “elasticity contrast index” or “ECI” was determined by manually setting the ROI within the lesion; once the ROI was manually set, the ECI was automatically calculated. To minimize possible observer variability and environmental factors, the operator obtained ECI values from both the transverse and longitudinal planes, and the mean of these two ECI values was used as the parameter of elasticity of each thyroid nodule.
One radiologist reviewed the gray-scale ultrasound images of the included data without knowledge of the clinical information or elastography findings and assigned a score of the likelihood of malignancy ranging from 1 (low) to 5 (high). The gray-scale images were classified by modifying the scores used in previous report by Kwak et al. [13]. A score of 1 represented a benign finding (no suspicious ultrasound features with a < 2% likelihood of malignancy); 2, a low suspicion for malignancy (one suspicious ultrasound feature with a 3–10% likelihood of malignancy); 3, intermediate suspicion (two suspicious ultrasound features with an 11–50% likelihood of malignancy); 4, moderate suspicion (three suspicious ultrasound features with a 51–90% likelihood of malignancy); and 5, high suspicion (four suspicious ultrasound features with a 90–100% likelihood of malignancy). The ultrasound findings of microcalcifications, irregular or microlobulated margins, marked hypoechogenicity, and taller-than-wide shape were considered to be indicative of malignancy.
One month after assessing the gray-scale ultrasound images alone, the radiologist reviewed the gray-scale ultrasound and elastography studies concurrently and again assigned a score from 1 to 5 for the likelihood of malignancy. The elastography results were classified by ECI values as follows: An elasticity score of 0 indicated an ECI value of less than 1; a score of 1, an ECI value of 1–3; and a score of 2, an ECI value of more than 3. The likelihood of malignancy on gray-scale ultrasound and elastography combined was then reviewed. The reader was asked to downgrade the likelihood of malignancy score when an elasticity score of 0 was assigned and to upgrade the likelihood of malignancy when an elasticity score of 2 was assigned. For the lesions with an elasticity score of 1, the likelihood of malignancy score was not changed.
Differences in the demographic characteristics and ECIs of patients with benign or malignant thyroid nodules were compared using the independent-sample Student t test. The ultrasound characteristics of benign and malignant thyroid nodules were compared using the Fisher exact test. The diagnostic performances of the gray-scale ultrasound score alone, ECI alone, and combined assessment of gray-scale ultrasound score and elastography were evaluated. The AUC (Az) values were compared by ROC curve analysis. The diagnostic performance of ECI with various cutoff values was evaluated along with sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). The diagnostic performances of the gray-scale ultrasound score, ECI, and a combined assessment using the optimal ECI cutoff value were evaluated, and p values < 0.05 were considered statistically significant. All statistical analyses were performed using a software package (SPSS, version 21.0, IBM) for Microsoft Windows.
| Results | |
|---|
Of 102 lesions, 12 lesions (11.8%) were confirmed to be malignant at surgery, and all 12 were papillary thyroid carcinoma. Of the 90 benign lesions, the diagnosis of four lesions was confirmed at surgery and the diagnosis of the other 86 lesions was confirmed at ultrasound-guided FNA. There were no significant differences in the demographic characteristics of patients with benign lesions and those with malignant lesions (Table 1).
The ultrasound characteristics of benign and malignant thyroid nodules are summarized in Table 2. The mean nodule size of the benign nodules was larger (16 mm) than that of the malignant nodules (9 mm) (p = 0.008). Significantly more malignant thyroid nodules were more hypoechoic than benign nodules (p = 0.014).
The gray-scale ultrasound score was 2 in 68 cases (66.7%), 3 in 31 cases (30.4%), and 4 in three cases (2.9%). The reassessed score of combined gray-scale ultrasound and elastography was 1 in 15 cases (14.7%), 2 in 56 cases (54.9%), 3 in 27 cases (26.5%), 4 in three cases (2.9%), and 5 in one case (1.0%). The mean ECI for the malignant nodules was 3.07 ± 2.01 (SD) with a range of 1.1–7.8 (Fig. 1), and the mean ECI for the benign nodules was 1.44 ± 0.47 with a range of 0.75–2.95 (Fig. 2). The ECI was significantly higher in malignant nodules than in benign thyroid nodules (p = 0.019). Figure 3 shows the ROC curves of the gray-scale ultrasound score, ECI, and combined assessment for the differentiation of benign from malignant nodules. The Az values of each dataset were 0.755 (95% CI, 0.660–0.835) for gray-scale ultrasound score alone, 0.835 (0.748–0.901) for ECI alone, and 0.853 (0.769–0.915) for a combined assessment. The Az value for the combined assessment was significantly higher than that of the gray-scale ultrasound score alone (p = 0.022). There were no significant differences in the Az values of gray-scale ultrasound alone versus ECI alone (p = 0.477) or in the Az values of ECI alone versus a combined assessment (p = 0.843).
 View larger version (282K) | Fig. 1A —29-year-old woman with 1.2-cm nodule in right thyroid lobe. A, Gray-scale ultrasound images show hypoechoic mass with smooth margin. Gray-scale ultrasound score was 2. |
 View larger version (283K) | Fig. 1B —29-year-old woman with 1.2-cm nodule in right thyroid lobe. B, Gray-scale ultrasound images show hypoechoic mass with smooth margin. Gray-scale ultrasound score was 2. |
 View larger version (247K) | Fig. 1C —29-year-old woman with 1.2-cm nodule in right thyroid lobe. C, Elasticity contrast index (ECI) is 5.48, and combination gray-scale ultrasound and elastography score is 3. This nodule (circles) was confirmed to be papillary carcinoma at histopathology. ⋄ = ROI to automatically calculate ECI. |
 View larger version (291K) | Fig. 2A —30-year-old woman with 2.1-cm nodule in right thyroid lobe. A, Gray-scale ultrasound images show 2.1-cm hypoechoic mass with ill-defined margin. Gray-scale ultrasound score was 4. |
 View larger version (263K) | Fig. 2B —30-year-old woman with 2.1-cm nodule in right thyroid lobe. B, Gray-scale ultrasound images show 2.1-cm hypoechoic mass with ill-defined margin. Gray-scale ultrasound score was 4. |
 View larger version (235K) | Fig. 2C —30-year-old woman with 2.1-cm nodule in right thyroid lobe. C, Elasticity contrast index (ECI) is 1.17, and combination gray-scale ultrasound and elastography score is 3. This nodule (circles) was confirmed to be adenomatous hyperplasia at histopathology. ⋄ = ROI to automatically calculate ECI>. |
 View larger version (35K) | Fig. 3 —ROC curves of gray-scale ultrasound score, elasticity contrast index (ECI), and combined gray-scale ultrasound with elastography for differentiation of benign from malignant nodules. Az value of combined assessment is significantly higher than that of gray-scale ultrasound score alone (p = 0.022). |
ECI had 100.0% sensitivity, 22.2% specificity, 14.6% PPV, and 100.0% NPV when a cutoff ECI value of 1.05 was used. Table 3 lists the diagnostic performances of ECI with various cutoff values for the differential diagnosis of thyroid nodules as benign or malignant. No benign nodules had ECI values of greater than 3.0, and no malignant nodules had ECI values of less than 1.05.
Table 4 lists the diagnostic performances of the gray-scale ultrasound score, ECI, and combined assessment using the optimal cutoff value for the differential diagnosis of thyroid nodules. When a cutoff ECI value of 2.45 was used, the specificity of ECI (97.8%) was higher than the specificity of the gray-scale ultrasound score (72.2%) and of the combined assessment (76.7%).
| Discussion | |
|---|
Ultrasound is the most popular method for differentiating benign from malignant thyroid nodules. However, there is little research regarding the role and accuracy of ultrasound for the diagnosis of thyroid nodules without suspicious ultrasound findings. Our retrospective review of sonographically indeterminate thyroid nodules resulted in an Az value of 0.755 for gray-scale ultrasound.
The management of solid thyroid nodules without suspicious ultrasound findings has been dependent on physician decisions. According to the American Thyroid Association guidelines [8], nodules less than 1 cm in diameter may require evaluation if there are pertinent historical factors that predict malignancy. The Korean Society of Radiology [7] suggests that thyroid nodules with indeterminate findings on ultrasound undergo observation if less than 1 cm in diameter or undergo selective FNA if larger than 1 cm in diameter. Selective FNA is recommended because the possibility of malignancy may be difficult to discern on gray-scale ultrasound. In terms of echogenicity, the ultrasound criteria for performing FNA vary with different research results and guidelines [7, 8, 14]. The descriptor “marked hypoechogenicity” has been used as a more specific criterion to differentiate malignant nodules from benign nodules. Among the various gray-scale parameters in our results, only hypoechogenicity resulted in significantly more malignant thyroid nodules with indeterminate ultrasound findings. Therefore, hypoechogenicity can be used as an adjunctive criterion when deciding whether to perform selective FNA of thyroid nodules with indeterminate findings on ultrasound.
Tissue stiffness may be an important reference index for the clinical differentiation of benign from malignant tumors. In previous studies, the diagnostic performance of elastography for differentiating benign from malignant thyroid nodules was variable. In a recent meta-analysis [15], elastography criteria had higher sensitivities (82% for elasticity score, 89% for strain ratio), higher specificities (82% for elasticity score, 82% for strain score), and higher Az values (0.89 for elasticity score, 0.93 for strain ratio) than gray-scale ultrasound features. Some investigators have reported that there was no additional role for thyroid elastography in the differentiation of malignant nodules from benign nodules [16, 17]. Recently, thyroid elastography has been used as an adjunctive tool to manage thyroid nodules with indeterminate or atypical cytology results [18, 19]. Thyroid nodules without suspicious findings were categorized as having indeterminate ultrasound findings [7]. The diagnostic performance of a combined assessment consisting of the gray-scale ultrasound score and ECI was significantly higher (Az = 0.853) than that of the gray-scale ultrasound score alone (Az = 0.755) (p = 0.022).
There are two different types of ultrasound elastography: Strain elastography is evaluated with extrinsic compression [20, 21], and shear-wave elastography generates shear waves. In our study, we used a new elastographic technique that uses carotid artery pulsation as a compression source for thyroid elastography (Fig. 4). Based on a previous study [22], the clinical applicability of thyroid elastography with external compression has been limited because of carotid artery pulsation, out-of-plane motion, and artifact caused by swallowing, and breathing. Furthermore, various factors have influenced thyroid elastography results such as nodule size, exophytic location of the nodule, the presence of rim calcification, and thyroiditis [12, 20–23]. In this study, carotid arterial pulsation had a different elastographic stimulus that was more regular and objective than manual external compression. This technique could eliminate one of the factors that influences elastography results. However, acquiring elastography data failed in a few cases in our study because of excessive carotid arterial pulsation in patients with an unusually high-positioned aortic arch.
 View larger version (20K) | Fig. 4A —Schematic drawings show principles of ultrasound elastography. Strain elastography, shear-wave elastography, and elastography using carotid artery pulsation are shown. A, Strain elastography applies manual compression to tissue with transducer. Malignant nodule is usually harder and shows less strain than benign nodules. A = nodule before compression force is applied, B = nodule with compression force applied, E = elasticity. |
 View larger version (11K) | Fig. 4B —Schematic drawings show principles of ultrasound elastography. Strain elastography, shear-wave elastography, and elastography using carotid artery pulsation are shown. B, Shear-wave elastography measures propagation speed of shear waves within tissue, which is induced by automatic pulse from transducer. Malignant nodule is usually stiffer and higher in propagation speed. |
 View larger version (36K) | Fig. 4C —Schematic drawings show principles of ultrasound elastography. Strain elastography, shear-wave elastography, and elastography using carotid artery pulsation are shown. C, Elastography using carotid artery pulsation evaluates heterogeneity of elasticity within nodule. Malignant nodule usually shows more heterogeneous elasticity and higher elasticity contrast index. |
The parameter for elastography is ECI, which quantifies the local strain contrast within a nodule. Larger ECI values suggest a stiffer nodule, indicating an increased likelihood of malignancy [24]. Similarly, our study showed that ECI values for malignant nodules (3.07 ± 2.01) were significantly higher than those for benign nodules (1.44 ± 0.47). The specificity of ECI (97.8%) was higher than that of the gray-scale ultrasound score (72.2%). High ECI specificity could play an adjunctive role in enhancing the diagnostic performance of gray-scale ultrasound [20, 23].
Many clinical researchers have reported on the observer variability of thyroid elastography. Park et al. [25] reported that there was no significant concordance among reviewers about thyroid elastography features unlike the significant concordance seen among reviewers for most gray-scale ultrasound findings. The subjective compression and learning curves of a relatively new technique could have affected the results of that study. Lim et al. [26] reported good interobserver agreement and good intraobserver agreement with high Pearson correlation coefficients (0.73–0.79) using the same elastography technique as our study.
Our study had several limitations. First, although this study was performed using consecutive cases, this retrospective study may have inherent patient selection bias because a similar ultrasound diagnosis category and biopsied cases were selected. Second, interobserver variability was not evaluated in the assessment of ultrasound and elastography. In addition, all thyroid malignancies were of the same papillary histologic type.
This study showed that a combined assessment of sonographically indeterminate thyroid nodules that consists of gray-scale ultrasound and thyroid elastography with carotid artery pulsation is useful for differentiating benign from malignant thyroid nodules.
Supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (NRF = 2012R1A1A2041860).
| References | |
|---|
