- Email: [email protected]
The diagnostic value of dynamic contrast-enhanced MRI for thyroid tumors
European Journal of Radiology 81 (2012) 3313–3318
Contents lists available at SciVerse ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
The diagnostic value of dynamic contrast-enhanced MRI for thyroid tumors Ying Yuan a,1 , Xiu-Hui Yue b,1 , Xiao-Feng Tao a,∗ a b
Department of Radiology, Shanghai Ninth People’s Hospital, Affiliated to JiaoTong University School of Medicine, Shanghai 200011, China Department of Radiology, Changzheng Hospital, Affiliated to Second Military Medical University, Shanghai 200003, China
a r t i c l e
i n f o
Article history: Received 13 January 2012 Received in revised form 20 April 2012 Accepted 23 April 2012 Keywords: Thyroid Neoplasm Dynamic Magnetic resonance imaging
a b s t r a c t Background and purpose: The exact place for dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) in the diagnosis and management of thyroid tumors is still under debate. We performed the study to analyze and compare the parameters generated from DCE-MRI for thyroid lesions. Materials and methods: For each thyroid lesion, time intensity curves (TIC), time of peak enhancement (Tpeak ), maximum enhancement ratio (ERmax ) and maximum rise slope (Slopemax ) were plotted and calculated. Receiver operator characteristics (ROC) analysis was conducted to assess the diagnostic ability and appropriate cut-off value. The area under the ROC curve (AUC) and the confidence intervals (CIs) were also assessed. Results: Forty-two patients were consecutively included. All 21 lesions demonstrated the rapid inflow and washout pattern (type-I) were benign. The 12 cases with delayed inflow pattern (type-III) were all malignant. When compared with the benign lesions, the thyroid carcinoma showed significantly lower Slopemax and higher Tpeak (P < 0.05). No statistical difference of ERmax was found between malignant and benign ones (P = 0.15). The AUC of ERmax , Slopemax and Tpeak in differentiating benign thyroid lesions from malignant ones were 0.63, 0.93and 1, respectively. The ERmax cut-off value of 73.86 (sensitivity, 71.4%; specificity, 64.3%), Slopemax cut-off value of 2.4126 (sensitivity, 92.9%; specificity, 82.1%) and Tpeak value of 28 (sensitivity, 100%; specificity, 100%) offered the best diagnostic performances. Conclusions: DCE-MRI, especially the pattern of TIC and the value of Slopemax and Tpeak , could be helpful in differentiating thyroid carcinoma from benign thyroid lesions. © 2012 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Thyroid carcinoma is the most common endocrine malignancy. Differentiating benign thyroid nodules from malignant ones is important in planning further therapeutic approach and the extent of surgical intervention. Different tools have been proposed to identify thyroid carcinoma at the onset. Ultrasound-guided fine-needle aspiration biopsy (US-guided FNAB) represents an essential tool; however this technique is still limited due to the invasiveness and 5–10% non-diagnostic rate [1,2]. Although not common and routine, magnetic resonance imaging (MRI), especially dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), has recently been used for the diagnosis and evaluation of thyroid lesions [3,4]. DCE-MRI provides information
∗ Corresponding author at: Department of Radiology, Shanghai Ninth People’s Hospital, 639 Zhizaoju Road, Shanghai, China. Tel.: +86 1381666309; fax: +86 21 64156886. E-mail address: [email protected] (X.-F. Tao). 1 Ying Yuan and Xiu-Hui Yue have contributed equally to this work. 0720-048X/$ – see front matter © 2012 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ejrad.2012.04.029
related to the vascularity of lesions and has the potential to detect and characterize tumors and evaluate treatment response. It was reported that dynamic MRI was helpful for the preoperative determination of cell proliferative ability and the grade of tumor differentiation of thyroid carcinoma [3], and was superior to US-guided FNAB in excluding thyroid carcinoma in patients with multinodular goiter [4]. However, to our knowledge, the exact place for DCE-MRI in the diagnosis and management of thyroid patients is still debated. Depending on the methods used for measurement and the status of patients, controversies persist concerning certain issues, including the cut-off values of data generating from DCE-MRI for accurate diagnosis of thyroid nodules. Therefore, the aim of this study was to: (1) determine the time intensity curves (TICs) pattern, time of peak enhancement (Tpeak ), maximum enhancement ratio (ERmax ) and maximum rise slope (Slopemax ) for each thyroid mass; (2) compare the TIC patterns, Tpeak , ERmax and Slopemax between benign and malignant thyroid lesions; (3) evaluate the diagnostic ability and determine the appropriately diagnostic threshold values for thyroid carcinoma using DCE-MRI.
3314
Y. Yuan et al. / European Journal of Radiology 81 (2012) 3313–3318
Table 1 MR imaging parameters. Parameter
Axial T1WI
Axial T2WI
Coronal T2WI
DCE-MR
Repetition time (msec) echo time (msec) Section thickness (mm) Section gap (mm) NEX FOV (cm) Matrix
520 14.3 4 1 4 14 × 14 320 × 192
3500 95 4 1 4 14 × 14 320 × 256
3000 85.6 4 1 4 14 × 14 320 × 224
3.6 0.9 8 −4 1 16 × 12 256 × 160
DCE-MRI, dynamic contrast-enhanced magnetic resonance imaging; FOV, field of view; NEX, number of excitations; T1WI, T1-weighted imaging; T2WI, T2-weighted imaging.
2. Materials and methods 2.1. Patients Patients with dominant thyroid nodules and/or nodules of clinical or US suspicion of malignancy, who underwent MRI before surgery and histopathological confirmation between July 2010 and January 2011 were consecutively included in the study. Nodules with features such as microcalcifications, coarse or macrocalcifications, marked hypoechogenicity, infiltrative margins, and taller-than-wide shape on US were treated as US suspicion of malignancy. Patients were excluded when an electrically powered device was implanted or a head or neck cancer was previously diagnosed. In patients with multiple nodules, the dominant nodules with the most clinical or US suspicious of malignancy were excised and evaluated in our study. 2.2. MR imaging MR images were obtained using 1.5 Tesla MR scanner (GE, Signa HD) with DUAL coil, and performed 3–8 days prior to surgery. All patients were placed in the magnet in the supine position. To avoid the motion artifact, the patients were told not to swallow during the acquisition. All patients underwent conventional nonenhanced axial T1-weighted (T1W) sequence, axial T2-weighted imaging (T2WI) and coronal T2WI. Axial images were obtained in each patient from the top of the mandibular angle to the sternal notch. The dynamic images were acquired with contrast enhanced fast spoiled gradient echo (FSPGR) sequence. The sections were acquired in the axial plane, centered on the focal thyroid mass and covering the peritumoral thyroid tissue and the external reference. After the first 10 s of dynamic acquisitions, Gadopentetate dimeglumine was intravenously bolus injected via a power injector at the rate of 2 mL/s at the dose of 0.1 mmol/kg of body weight. One hundred dynamic phases were acquired for each investigation, which were obtained every 3.6 s over 6 min in a serial manner. The MRI protocol was based on previous studies of thyroid cancer [4–6] and was further adjusted to enable better image quality and sound-tonoise ratio. The imaging parameters of non-enhanced and dynamic MR imaging are given in Table 1. 2.3. Evaluation of dynamic MRI and data analysis After a color-coded mapping of the entire tumor region, we placed 2 mm × 2 mm regions of interest (ROIs) voxel by voxel over the entire lesion on multiple slices. We chose to place multiple ROIs voxel by voxel instead of manually delineating the lesion and calculating the DCE parameters over the entire tumor in order to avoid cystic areas and calcifications in the lesions. Thereafter, a multiple of ROIs were obtained and the number of acquired ROIs changed with the lesion size. The signal intensity (SI) of a ROI was calculated from the mean pixel value on each acquired dynamic image using MRI console and then time-signal intensity curves (TICs) were generated. By comparing the TICs from each ROI, the ROI with maximal
enhancement was selected as a representation for the lesion [7]. The selection of representative ROI was performed by two readers (X.-H. Y. and Y. Y.) independently. To resolve possible disagreement and to reduce intra-observer variability, a third reviewer (X.-F. T.) assessed discrepant items, and the majority opinion was used for analysis. The corresponding TICs were obtained, from which signal intensity (SIpre , SImax ) and time (Tpre , Tpeak ) were derived. SIpre was defined as the pre-contrast signal intensity, and SImax was defined as the signal intensity at maximal contrast enhancement. Tpre and Tpeak were defined as the time corresponding to the SIpre and SImax . According to the parameters previously reported in an MRI study of parotid gland tumors [8], ERmax and Slopemax were calculated using the following two formulas: ERmax =
SImax − SIpre × 100 SIpre
Slopemax =
SImax − SIpre × 100 SIpre × (Tpeak − Tpre )
(1)
(2)
Statistical analysis was then performed to determine the relationship between enhancement pattern and malignancy. We reviewed the distribution of Tpeak , ERmax and Slopemax levels according to the final diagnosis of each specimen. Data of the benign and malignant lesions were compared. To find the optimal cut-off level to distinguish malignant tumors from benign ones, we calculated the sensitivity and specificity of various Tpeak , ERmax and Slopemax cutoff values. Receiver operator characteristics (ROC) analysis was conducted using SPSS software Version 16.0 (SPSS, Chicago, IL) to evaluate the diagnostic ability and assess the appropriate threshold value of Tpeak , ERmax and Slopemax levels in all cases. We plotted the sensitivity versus (1-specificity) for each cut-off value across the range of Tpeak , ERmax and Slopemax for diagnosing thyroid carcinoma. The areas under the ROC curve (AUCs) and the confidence intervals (CIs) were assessed. The cut-off values which maximize the sum of sensitivity and specificity were determined as the points in the upper left hand corner. P < 0.05 was considered as statistically significant. 3. Results Forty-two patients (10 males, 32 females) aged 20–72 years (mean, 45 years) were finally included. Twenty-six of them were accidentally detected during physical examination of US. The most common clinical symptom was neck mass, and 6 patients had already clinical signs and symptoms suspicious for malignancy, including hoarseness (n = 1), suspicious lymph nodes (n = 2), and fixed and firm palpable thyroid masses (n = 2). All 42 lesions were surgically removed and histopathologically confirmed. The spectrum of thyroid diseases included: (1) malignant lesions (n = 14): recorded as papillary carcinoma (n = 12) and follicular carcinoma (n = 2); (2) benign lesions (n = 28): recorded as adenoma (n = 20), adenomatous goiter (n = 6) and Hashimoto thyroiditis/chronic lymphocytic thyroiditis (n = 2). (see Table 2) The mean (range) nodule diameter was 2.3 cm (0.9–4.4) cm.
Y. Yuan et al. / European Journal of Radiology 81 (2012) 3313–3318
3315
Fig. 1. (a) The nodule showed contrast enhancement and rapid inflow and washout pattern (type-I) on DCE-MRI. The final histopathologic diagnosis was benign adenoma. (A, B) Non-enhanced axial T1W and T2W MR image; (C) contrast enhanced axial T1W image; (D) contrast enhanced coronal T1W image; (E, F) ROI on dynamic MR mage and corresponding time intensity curve. (b) The nodule showed contrast enhancement and plateau pattern (type-II) on DCE-MRI. The final histopathologic diagnosis was follicular thyroid carcinoma. (A, B) Non-enhanced axial T1W and T2W MR image; (C) contrast enhanced axial T1W image; (D) contrast enhanced coronal T1W image; (E, F) ROI on dynamic MR mage and corresponding time-intensity curve. (c) The nodule showed contrast enhancement and delayed inflow pattern (type-III) on DCE-MRI. The final histopathologic diagnosis was papillary thyroid carcinoma. (A, B) Non-enhanced axial T1W and T2W MR image; (C) contrast enhanced axial T1W image; (D) contrast enhanced coronal T1W image; (E, F) ROI on dynamic MR mage and corresponding time-intensity curve.
3316
Y. Yuan et al. / European Journal of Radiology 81 (2012) 3313–3318
Table 2 Time-intensity curves of lesions. Number of cases
Benign Malignant
28 14
TIC pattern I
II
III
21 0
7 2
0 12
TIC, time-intensity curve.
Fig. 2. The distribution of Tpeak , ERmax , and Slopemax according to the final diagnosis (benign or malignant).
Twenty-nine nodules were solitary. Twenty cases were patients with multiple nodules, and the dominant nodules with the most clinical suspicious of malignancy were excised and evaluated. Preoperative FNAB was performed in one patient. TICs (axis coordinate, time; vertical, signal intensity) exhibited three patterns: (1) the rapid inflow and washout pattern (typeI) with TICs displayed a rapid increase slope and a final intensity lower than 90% of peak grade (Fig. 1a); (2) the plateau pattern (typeII) with a relatively prominent increase slope and a final intensity 90–100% of peak grade (Fig. 1b); or (3) the delayed inflow pattern (type-III) with a peak enhancement reached during the second half of the dynamic acquisition (Fig. 1c). All of the lesions that demonstrated the rapid inflow and washout pattern (type-I) were benign (n = 21). Among nine cases presenting the plateau pattern (type-II), seven cases were benign and the other two cases were malignant, which were both follicular carcinoma. All of the 12 cases with the delayed inflow pattern (type-III) were malignant (see Table 2). The results of ERmax , Slopemax and Tpeak are plotted in Fig. 2. The Tpeak of malignant lesions (mean ± SD, 50.50 ± 11.61) was statistically higher than benign ones (14.50 ± 3.66) (P < 0.0001). The mean ± SD Slopemax was 4.99 ± 3.14 for benign lesions, and 0.94 ± 0.93 for malignant tumors. Significant difference was also found between them (P = 0.0003 < 0.05) (see Table 3). No statistical difference was found between the ERmax of malignant lesions (130.37 ± 119.92) and those of benign lesions (78.47 ± 74.36) (P = 0.15 > 0.05) (see Table 3). The ROC curves are shown in Fig. 3. The AUC for ERmax , Slopemax and Tpeak in differentiating benign thyroid lesions from malignant ones were 0.63, 0.93and 1, respectively. The ERmax cut-off value of Table 3 Tpeak , Slopemax , ERmax value of lesions.
Benign Malignant P value*
Number of cases
Tpeak
ERmax
Slopemax
28 14 –
14.50 ± 3.66 50.50 ± 11.61 P < 0.0001
78.47 ± 74.36 130.37 ± 119.92 P = 0.15
4.99 ± 3.14 0.94 ± 0.93 P = 0.0003
* P value for the comparison of Tpeak , ERmax , and Slopemax values between benign and malignant lesions. P < 0.05 was considered statistically significant.
Fig. 3. Receiver operator characteristics (ROC) curve for Tpeak , ERmax , and Slopemax measurements in this study. The dotted red line represents the ROC curve of Tpeak , the solid blue line the ROC curve for Slopemax , and solid black line the ROC curve for ERmax . The area under the ROC curve (AUCs) for Tpeak , ERmax , and Slopemax are 1.00 (standard error [SE] 0; 95% confidence interval [CI] 1.000–1.000), 0.63 (SE 0.094; CI 0.449–0.816), and 0.93 (SE 0.040; CI 0.849–1.004).
73.86 (sensitivity, 71.4%; specificity, 64.3%), Slopemax cut-off value of 2.4126 (sensitivity, 92.9%; specificity, 82.1%) and Tpeak value of 28 (sensitivity, 100%; specificity, 100%) offered the best diagnostic performances. 4. Discussion MRI offers the advantages of higher contrast resolution and the lack of ionizing radiation. With development in MR technique and thriving of dynamic enhancement technology, it could be expected that MRI might contribute to improve the diagnostic ability. DCEMRI is an emerging tool for functional assessment of a specific target tissue [9,10]. Kusunoki et al. [3] and Tezelman et al. [11] have reported 100% sensitivity for DCE-MRI to diagnose thyroid disease. With the information provided by TIC, the diagnostic accuracy and specificity of DCE-MRI were also further improved [4,12]. Thyroid is an endocrine organ and consists of all sizes of acinus, which are filled with colloid with a main component of thyroid hormone and thyroglobulin (Tg) [13]. Even under normal condition, the vascular distribution is abundant in the thyroid gland in adapt to its function. Once pathological changes take place, corresponding changes may be detected in thyroglobulin level, vascular permeability and perfusion, which eventually may alter TIC pattern. Kusunoki et al. [3] found that in most of malignant thyroid diseases and a few benign diseases that had marked cell proliferative activity, the TICs displayed the plateau pattern, and almost all benign diseases and a few well differentiated carcinomas displayed the rapid washout pattern. It was concluded that Gd-DTPA may be well take in lesions that show marked cell proliferative activity with hypervascularity, and pooled in these lesions, which resulted to exhibit the plateau pattern. Tezelman et al. [11] assess the results from 33 patients with thyroid lesions and found that malignant lesions tended to display the plateau pattern on DCE-MRI and a correlation was found between them. Different opinions were also proposed. In a recently published study of 28 patients with solitary thyroid nodule, rapid enhancement and washout pattern was seen in most malignant cases [14]. However, results of another
Y. Yuan et al. / European Journal of Radiology 81 (2012) 3313–3318
study reported by Tunca et al. [4] has demonstrated that thyroid carcinoma presented delayed washout pattern, but no plateau or rapid washout pattern. In the current study, most (85.17%, 12/14) of the malignant tumors, that was all of the 12 cases of papillary carcinoma, demonstrated the delayed inflow pattern (type-III). The TICs of the other two cases of thyroid carcinoma exhibited a plateau pattern (type-II), and both of them were follicular carcinoma. While benign lesions mainly presented a rapid inflow and washout pattern (type-I) or plateau pattern (type-II). The pathological specificity of the TIC pattern in our study is interesting and should be worth of further study. Possible explanation for variations of curves and the results among the studies may be related to the following aspects. Firstly, increased vascular permeability of malignant lesions, due to the larger space between vascular endothelial cells and incomplete basilar membrane, provides the pathophysiological foundation of the rapid inflow and washout pattern. However, the nuclear-cytoplasmic ratio of thyroid carcinoma is high. Thyroid carcinoma or thyroid nodules with a high cell proliferation index tend to demonstrate plateau pattern on DCE-MRI [3]. The component of thyroid lesions matters as well. The component of benign thyroid nodule is complex, including colloid substance, necrosis, hemorrhage, fibrous tissue and calcification, which would alter the TIC pattern. Therefore, in our study, we choose to place multiple small ROIs voxel by voxel and picked up the representative ROI, instead of manually delineating the lesion and calculating the DCE parameters over the entire tumor, in order to avoid cystic areas and calcifications in the lesions, which would introduce bias. Furthermore, the microvessel density (MVD) count is the most reliable method for estimating tumor angiogenesis [15]. Studies have proved that the MVD in follicular is higher than that in papillary carcinoma [16]. In recent years, along with the studies on specific markers of lymphatic endothelium, lymphatic vessel density (LVD) has been adopted as a new biomarker [17,18]. Reduction in the amount of D2-40 positive lymphatic vessel was detected in follicular thyroid carcinoma when compared with papillary carcinoma. Therefore, the two cases of follicular carcinoma presented plateau pattern in our studies could be explained by the relatively high MVD and low LVD in follicular carcinoma. It is also in line with the bionomics that papillary carcinoma tends to have lymphatic metastasis while hematogenous metastasis is more common in follicular carcinoma. Technologic methods, such as the parameter and sequence of MR scans, the way of deriving TICs and placing ROIs, the contrast (dose, rate and delay time) may differed from one study to another, which explains these discrepancies as well. Dynamic MRI relies on the use of a fast imaging technique with high temporal resolution and provides quantitative estimation of physiologic parameters related to perfusion and/or permeability in vivo [19]. The protocol of dynamic MRI partly differed from each other in previous studies. Kusunoki et al. [3] had take DCE-MR scans for 26 patients with thyroid masses. The overall dynamic duration was 10 min, and scans were performed at 30 s intervals for 3, 4, 5, 7 and 10 min after the injection. In another study [11] to evaluate DCE-MR performance of 30 patients with thyroid lesions, dynamic scan was executed at 3, 7, 10, 14, 17, 21 min after contrast injection. Comparing to our study, these two studies employed a longer dynamic MR scanning, which may better reflect the washout pattern. In our study, DCE-MR scans were started 10 s before the injection and 100 sets of dynamic MR images were obtained every 3.5 s over 6 min, in order to guarantee that no data was missed during the dynamic procedure and the TIC patterns obtained were sufficient for data analysis. The more or less difference of imaging protocol might have impact on the TICs generated. The MRI protocol adopted in the current study was based on previous studies of thyroid cancer [4–6] and was only adjusted to enable better image quality and sound-to-noise ratio. All the mentioned components and features would alter the result
3317
of TIC. Further studies including larger number of patients should be conducted to explain the possibility of heterogeneity between studies. The other two biomarkers Tpeak and Slopemax generated from DCE-MRI was significantly higher and lower in malignant thyroid lesions compared with benign lesions, respectively, which was in line with the TIC pattern. No statistical difference of ERmax was found between the two groups, and the results of ROC curve and AUC value of ERmax (AUC, 0.63; CI, 0.449–0.816) confirmed a limited diagnostic value with a confidential interval including 0.5. More optimal diagnostic ability was achieved by Slopemax and Tpeak value. The AUCs for Slopemax and Tpeak in differentiate benign thyroid lesions from malignant ones were 0.93 and 1, respectively. Best diagnostic performances were obtained by adopting the Slopemax cut-off value of 2.4126 (sensitivity, 92.9%; specificity, 82.1%) and Tpeak value of 28 (sensitivity, 100%; specificity, 100%). The results of the ROC analysis were promising; however, it should be mentioned that the method for determining the cutoff value may also differed from one study to another, depending on the sequence of MR imaging, the ROI selection, and so forth. Furthermore, it is not clear whether the same threshold can be proposed for patients before and after thyroidectomy, and for patients with different circulating Tg concentrations. Therefore, the main limitation of the current study should be related to the relatively small number of malignant cases included in the study (n = 14). The number of patients within each subgroup had, to certain extent, limited the subgroup analysis based on different histological types and other patients’ characteristics. For instance, the limited amount of patients with follicular carcinoma (n = 2) deprived the statistically ability of explaining DCE-MRI results. Therefore, further studies are worth to be conducted in larger group of patients, which would further confirm the diagnostic ability of dynamic MR and evaluate cut-off level depending on the characteristics of patients, lesions and imaging techniques. Another limitation is related to the method we used to generate the representative ROI for the evaluation. Most of the previous studies chose to manually delineate the lesion and calculate the DCE parameters over the entire tumor. In our study, we placed multiple small ROIs voxel by voxel over the entire lesions and then picked up the representative ROI with the maximal enhancement. This method would rule out the possibility of including cystic or calcific components along with the solitary tumor, since the lesions analyzed in our study are not only solitary tumors but also those with cyst and calcification. The subjective bias might be reduced as well, as the number of ROIs acquired in the same lesion should be stable. The bias might be introduced during the procedure of selecting the representative ROIs. As a result, two readers independently selected the representative ROI and a third reviewer assessed discrepant items according to the majority opinion was used for analysis. FNAB, especially US-guided FNAB, has an essential role in the evaluation of patients with thyroid nodules to reduce the rate of unnecessary thyroid surgery and triage patients to appropriate surgery. However, there are still a low incidence of both falsepositive and false-negative diagnosis of FNA of thyroid nodules [20]. For example, it is more likely to get specimens inadequate for cytology analysis for nodules smaller than 1.5 cm. False-negative results may cause delay in the surgical treatment of patients with thyroid carcinoma and lead to the development of local invasion in some patients. On the other side, sampling error may be especially relevant for larger nodules even with US guidance [21]. Cystic or calcific changes in lesions can cause further diagnostic problems due to inadequate sampling. In our study, preoperative FNAB was performed in only one patient. This could be attibuted to the abovementioned limitations of FNAB, as well as the patients’ willing to refuse an invasive examination. The cost was also a consideration.
3318
Y. Yuan et al. / European Journal of Radiology 81 (2012) 3313–3318
However, we have to admit that it is a limitation of our study for not comparing the results of DCE-MRI with US guided-FNAB, which are being clinically recommended. 5. Conclusions Dynamic MRI could be clinically useful in the differential diagnosis of benign thyroid lesions from malignant lesions. The rapid inflow and washout pattern of time-intensity curve indicates a benign lesion, and most malignant lesions demonstrated the delayed inflow pattern. The Slopemax and Tpeak performed well in differentiating benign thyroid lesions from malignant ones. Therefore, further studies are worth to be conducted in larger group of patients, which would further confirm the diagnostic ability of dynamic MR and evaluate cut-off level depending on the characteristics of patients, lesions and imaging techniques. References [1] Frasoldati A, Valcavi R. Challenges in neck ultrasonography: lymphadenopathy and parathyroid glands. Endocrine Practice 2004;10(3):261–8. [2] Boland GW, Lee MJ, Mueller PR, Mayo-Smith W, Dawson SL, Simeone JF. Efficacy of sonographically guided biopsy of thyroid masses and cervical lymph nodes. American Journal of Roentgenology 1993;161(5):1053–6. [3] Kusunoki T, Murata K, Nishida S, Tomura T, Inoue M. Histopathological findings of human thyroid tumors and dynamic MRI. Auris, Nasus, Larynx 2002;29(4):357–60. [4] Tunca F, Giles Y, Salmaslioglu A, et al. The preoperative exclusion of thyroid carcinoma in multinodular goiter: dynamic contrast-enhanced magnetic resonance imaging versus ultrasonography-guided fine-needle aspiration biopsy. Surgery 2007;142(6):992–1002. [5] Noworolski SM, Fischbein NJ, Kaplan MJ, et al. Challenges in dynamic contrastenhanced MRI imaging of cervical lymph nodes to detect metastatic disease. Journal of Magnetic Resonance Imaging 2003;17(4):455–62. [6] Schueller-Weidekamm C, Kaserer K, Schueller G, et al. Can quantitative diffusion-weighted MR imaging differentiate benign and malignant cold thyroidnodules initial results in 25 patients. AJNR American Journal of Neuroradiology 2009;30:417–22.
[7] Unlu E, Albayram S, Temizoz O, Caloglu M, Tamer Y. The role of dynamic contrast-enhanced magnetic resonance imaging parameters in head-neck tumors. Rivista di Neuroradiologia 2006;19:245–56. [8] Yabuuchi H, Fukuya T, Tajima T, Hachitanda Y, Tomita K, Koga M. Salivary gland tumors: diagnostic value of gadolinium-enhanced dynamic MR imaging with histopathologic correlation. Radiology 2003;226(2): 345–54. [9] Choyke PL, Dwyer AJ, Knopp MV. Functional tumor imaging with dynamic contrast-enhanced magnetic resonance imaging. Journal of Magnetic Resonance Imaging 2003;17(5):509–20. [10] O’Connor JP, Jackson A, Asselin MC, Buckley DL, Parker GJ, Jayson GC. Quantitative imaging biomarkers in the clinical development of targeted therapeutics: current and future perspectives. Lancet Oncology 2008;9(8): 766–76. [11] Tezelman S, Giles Y, Tunca F, et al. Diagnostic value of dynamic contrast medium-enhanced magnetic resonance imaging in preoperative detection of thyroid carcinoma. Archives of Surgery 2007;142(11): 1036–41. [12] Kusunoki T, Murata K, Hosoi H, Nishida S, Tomura T, Inoue M. Malignancies of human thyroid tumors and dynamic magnetic resonance imaging (MRI). Auris, Nasus, Larynx 1998;25(4):419–24. [13] Wissig SL. The anatomy of secretion in the follicular cells of the thyroid gland. I. The fine structure of the gland in the normal rat. Journal of Biophysical and Biochemical Cytology 1960;7:419–32. [14] Gupta N, Norbu C, Goswami B, et al. Role of dynamic MRI in differentiating benign from malignant follicular thyroid nodule. Auris, Nasus, Larynx 2011;38(6):718–23. [15] Weidner N. Tumor angiogenesis: review of current applications in tumor prognostication. Seminars in Diagnostic Pathology 1993;10(4):302–13. [16] Ramsden JD. Angiogenesis in the thyroid gland. Journal of Endocrinology 2000;166(3):475–80. [17] Giorgadze TA, Baloch ZW, Pasha T, Zhang PJ, Livolsi VA. Lymphatic and blood vessel density in the follicular patterned lesions of thyroid. Modern Pathology 2005;18(11):1424–31. [18] de la Torre NG, Buley I, Wass JA, Turner HE. Angiogenesis and lymphangiogenesis in thyroid proliferative lesions: relationship to type and tumour behaviour. Endocrine Related Cancer 2006;13(3):931–44. [19] Oostendorp M, Post MJ, Backes WH. Vessel growth and function: depiction with contrast-enhanced MR imaging. Radiology 2009;251(2):317–35. [20] Yeh MW, Demircan O, Ituarte P, Clark OH. False-negative fine-needle aspiration cytology results delay treatment and adversely affect outcome in patients with thyroid carcinoma. Thyroid 2004;14(3):207–15. [21] Porterfield Jr JR, Grant CS, Dean DS, et al. Reliability of benign fine needle aspiration cytology of large thyroid nodules. Surgery 2008;144:963–8.
Contents lists available at SciVerse ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
The diagnostic value of dynamic contrast-enhanced MRI for thyroid tumors Ying Yuan a,1 , Xiu-Hui Yue b,1 , Xiao-Feng Tao a,∗ a b
Department of Radiology, Shanghai Ninth People’s Hospital, Affiliated to JiaoTong University School of Medicine, Shanghai 200011, China Department of Radiology, Changzheng Hospital, Affiliated to Second Military Medical University, Shanghai 200003, China
a r t i c l e
i n f o
Article history: Received 13 January 2012 Received in revised form 20 April 2012 Accepted 23 April 2012 Keywords: Thyroid Neoplasm Dynamic Magnetic resonance imaging
a b s t r a c t Background and purpose: The exact place for dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) in the diagnosis and management of thyroid tumors is still under debate. We performed the study to analyze and compare the parameters generated from DCE-MRI for thyroid lesions. Materials and methods: For each thyroid lesion, time intensity curves (TIC), time of peak enhancement (Tpeak ), maximum enhancement ratio (ERmax ) and maximum rise slope (Slopemax ) were plotted and calculated. Receiver operator characteristics (ROC) analysis was conducted to assess the diagnostic ability and appropriate cut-off value. The area under the ROC curve (AUC) and the confidence intervals (CIs) were also assessed. Results: Forty-two patients were consecutively included. All 21 lesions demonstrated the rapid inflow and washout pattern (type-I) were benign. The 12 cases with delayed inflow pattern (type-III) were all malignant. When compared with the benign lesions, the thyroid carcinoma showed significantly lower Slopemax and higher Tpeak (P < 0.05). No statistical difference of ERmax was found between malignant and benign ones (P = 0.15). The AUC of ERmax , Slopemax and Tpeak in differentiating benign thyroid lesions from malignant ones were 0.63, 0.93and 1, respectively. The ERmax cut-off value of 73.86 (sensitivity, 71.4%; specificity, 64.3%), Slopemax cut-off value of 2.4126 (sensitivity, 92.9%; specificity, 82.1%) and Tpeak value of 28 (sensitivity, 100%; specificity, 100%) offered the best diagnostic performances. Conclusions: DCE-MRI, especially the pattern of TIC and the value of Slopemax and Tpeak , could be helpful in differentiating thyroid carcinoma from benign thyroid lesions. © 2012 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Thyroid carcinoma is the most common endocrine malignancy. Differentiating benign thyroid nodules from malignant ones is important in planning further therapeutic approach and the extent of surgical intervention. Different tools have been proposed to identify thyroid carcinoma at the onset. Ultrasound-guided fine-needle aspiration biopsy (US-guided FNAB) represents an essential tool; however this technique is still limited due to the invasiveness and 5–10% non-diagnostic rate [1,2]. Although not common and routine, magnetic resonance imaging (MRI), especially dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), has recently been used for the diagnosis and evaluation of thyroid lesions [3,4]. DCE-MRI provides information
∗ Corresponding author at: Department of Radiology, Shanghai Ninth People’s Hospital, 639 Zhizaoju Road, Shanghai, China. Tel.: +86 1381666309; fax: +86 21 64156886. E-mail address: [email protected] (X.-F. Tao). 1 Ying Yuan and Xiu-Hui Yue have contributed equally to this work. 0720-048X/$ – see front matter © 2012 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ejrad.2012.04.029
related to the vascularity of lesions and has the potential to detect and characterize tumors and evaluate treatment response. It was reported that dynamic MRI was helpful for the preoperative determination of cell proliferative ability and the grade of tumor differentiation of thyroid carcinoma [3], and was superior to US-guided FNAB in excluding thyroid carcinoma in patients with multinodular goiter [4]. However, to our knowledge, the exact place for DCE-MRI in the diagnosis and management of thyroid patients is still debated. Depending on the methods used for measurement and the status of patients, controversies persist concerning certain issues, including the cut-off values of data generating from DCE-MRI for accurate diagnosis of thyroid nodules. Therefore, the aim of this study was to: (1) determine the time intensity curves (TICs) pattern, time of peak enhancement (Tpeak ), maximum enhancement ratio (ERmax ) and maximum rise slope (Slopemax ) for each thyroid mass; (2) compare the TIC patterns, Tpeak , ERmax and Slopemax between benign and malignant thyroid lesions; (3) evaluate the diagnostic ability and determine the appropriately diagnostic threshold values for thyroid carcinoma using DCE-MRI.
3314
Y. Yuan et al. / European Journal of Radiology 81 (2012) 3313–3318
Table 1 MR imaging parameters. Parameter
Axial T1WI
Axial T2WI
Coronal T2WI
DCE-MR
Repetition time (msec) echo time (msec) Section thickness (mm) Section gap (mm) NEX FOV (cm) Matrix
520 14.3 4 1 4 14 × 14 320 × 192
3500 95 4 1 4 14 × 14 320 × 256
3000 85.6 4 1 4 14 × 14 320 × 224
3.6 0.9 8 −4 1 16 × 12 256 × 160
DCE-MRI, dynamic contrast-enhanced magnetic resonance imaging; FOV, field of view; NEX, number of excitations; T1WI, T1-weighted imaging; T2WI, T2-weighted imaging.
2. Materials and methods 2.1. Patients Patients with dominant thyroid nodules and/or nodules of clinical or US suspicion of malignancy, who underwent MRI before surgery and histopathological confirmation between July 2010 and January 2011 were consecutively included in the study. Nodules with features such as microcalcifications, coarse or macrocalcifications, marked hypoechogenicity, infiltrative margins, and taller-than-wide shape on US were treated as US suspicion of malignancy. Patients were excluded when an electrically powered device was implanted or a head or neck cancer was previously diagnosed. In patients with multiple nodules, the dominant nodules with the most clinical or US suspicious of malignancy were excised and evaluated in our study. 2.2. MR imaging MR images were obtained using 1.5 Tesla MR scanner (GE, Signa HD) with DUAL coil, and performed 3–8 days prior to surgery. All patients were placed in the magnet in the supine position. To avoid the motion artifact, the patients were told not to swallow during the acquisition. All patients underwent conventional nonenhanced axial T1-weighted (T1W) sequence, axial T2-weighted imaging (T2WI) and coronal T2WI. Axial images were obtained in each patient from the top of the mandibular angle to the sternal notch. The dynamic images were acquired with contrast enhanced fast spoiled gradient echo (FSPGR) sequence. The sections were acquired in the axial plane, centered on the focal thyroid mass and covering the peritumoral thyroid tissue and the external reference. After the first 10 s of dynamic acquisitions, Gadopentetate dimeglumine was intravenously bolus injected via a power injector at the rate of 2 mL/s at the dose of 0.1 mmol/kg of body weight. One hundred dynamic phases were acquired for each investigation, which were obtained every 3.6 s over 6 min in a serial manner. The MRI protocol was based on previous studies of thyroid cancer [4–6] and was further adjusted to enable better image quality and sound-tonoise ratio. The imaging parameters of non-enhanced and dynamic MR imaging are given in Table 1. 2.3. Evaluation of dynamic MRI and data analysis After a color-coded mapping of the entire tumor region, we placed 2 mm × 2 mm regions of interest (ROIs) voxel by voxel over the entire lesion on multiple slices. We chose to place multiple ROIs voxel by voxel instead of manually delineating the lesion and calculating the DCE parameters over the entire tumor in order to avoid cystic areas and calcifications in the lesions. Thereafter, a multiple of ROIs were obtained and the number of acquired ROIs changed with the lesion size. The signal intensity (SI) of a ROI was calculated from the mean pixel value on each acquired dynamic image using MRI console and then time-signal intensity curves (TICs) were generated. By comparing the TICs from each ROI, the ROI with maximal
enhancement was selected as a representation for the lesion [7]. The selection of representative ROI was performed by two readers (X.-H. Y. and Y. Y.) independently. To resolve possible disagreement and to reduce intra-observer variability, a third reviewer (X.-F. T.) assessed discrepant items, and the majority opinion was used for analysis. The corresponding TICs were obtained, from which signal intensity (SIpre , SImax ) and time (Tpre , Tpeak ) were derived. SIpre was defined as the pre-contrast signal intensity, and SImax was defined as the signal intensity at maximal contrast enhancement. Tpre and Tpeak were defined as the time corresponding to the SIpre and SImax . According to the parameters previously reported in an MRI study of parotid gland tumors [8], ERmax and Slopemax were calculated using the following two formulas: ERmax =
SImax − SIpre × 100 SIpre
Slopemax =
SImax − SIpre × 100 SIpre × (Tpeak − Tpre )
(1)
(2)
Statistical analysis was then performed to determine the relationship between enhancement pattern and malignancy. We reviewed the distribution of Tpeak , ERmax and Slopemax levels according to the final diagnosis of each specimen. Data of the benign and malignant lesions were compared. To find the optimal cut-off level to distinguish malignant tumors from benign ones, we calculated the sensitivity and specificity of various Tpeak , ERmax and Slopemax cutoff values. Receiver operator characteristics (ROC) analysis was conducted using SPSS software Version 16.0 (SPSS, Chicago, IL) to evaluate the diagnostic ability and assess the appropriate threshold value of Tpeak , ERmax and Slopemax levels in all cases. We plotted the sensitivity versus (1-specificity) for each cut-off value across the range of Tpeak , ERmax and Slopemax for diagnosing thyroid carcinoma. The areas under the ROC curve (AUCs) and the confidence intervals (CIs) were assessed. The cut-off values which maximize the sum of sensitivity and specificity were determined as the points in the upper left hand corner. P < 0.05 was considered as statistically significant. 3. Results Forty-two patients (10 males, 32 females) aged 20–72 years (mean, 45 years) were finally included. Twenty-six of them were accidentally detected during physical examination of US. The most common clinical symptom was neck mass, and 6 patients had already clinical signs and symptoms suspicious for malignancy, including hoarseness (n = 1), suspicious lymph nodes (n = 2), and fixed and firm palpable thyroid masses (n = 2). All 42 lesions were surgically removed and histopathologically confirmed. The spectrum of thyroid diseases included: (1) malignant lesions (n = 14): recorded as papillary carcinoma (n = 12) and follicular carcinoma (n = 2); (2) benign lesions (n = 28): recorded as adenoma (n = 20), adenomatous goiter (n = 6) and Hashimoto thyroiditis/chronic lymphocytic thyroiditis (n = 2). (see Table 2) The mean (range) nodule diameter was 2.3 cm (0.9–4.4) cm.
Y. Yuan et al. / European Journal of Radiology 81 (2012) 3313–3318
3315
Fig. 1. (a) The nodule showed contrast enhancement and rapid inflow and washout pattern (type-I) on DCE-MRI. The final histopathologic diagnosis was benign adenoma. (A, B) Non-enhanced axial T1W and T2W MR image; (C) contrast enhanced axial T1W image; (D) contrast enhanced coronal T1W image; (E, F) ROI on dynamic MR mage and corresponding time intensity curve. (b) The nodule showed contrast enhancement and plateau pattern (type-II) on DCE-MRI. The final histopathologic diagnosis was follicular thyroid carcinoma. (A, B) Non-enhanced axial T1W and T2W MR image; (C) contrast enhanced axial T1W image; (D) contrast enhanced coronal T1W image; (E, F) ROI on dynamic MR mage and corresponding time-intensity curve. (c) The nodule showed contrast enhancement and delayed inflow pattern (type-III) on DCE-MRI. The final histopathologic diagnosis was papillary thyroid carcinoma. (A, B) Non-enhanced axial T1W and T2W MR image; (C) contrast enhanced axial T1W image; (D) contrast enhanced coronal T1W image; (E, F) ROI on dynamic MR mage and corresponding time-intensity curve.
3316
Y. Yuan et al. / European Journal of Radiology 81 (2012) 3313–3318
Table 2 Time-intensity curves of lesions. Number of cases
Benign Malignant
28 14
TIC pattern I
II
III
21 0
7 2
0 12
TIC, time-intensity curve.
Fig. 2. The distribution of Tpeak , ERmax , and Slopemax according to the final diagnosis (benign or malignant).
Twenty-nine nodules were solitary. Twenty cases were patients with multiple nodules, and the dominant nodules with the most clinical suspicious of malignancy were excised and evaluated. Preoperative FNAB was performed in one patient. TICs (axis coordinate, time; vertical, signal intensity) exhibited three patterns: (1) the rapid inflow and washout pattern (typeI) with TICs displayed a rapid increase slope and a final intensity lower than 90% of peak grade (Fig. 1a); (2) the plateau pattern (typeII) with a relatively prominent increase slope and a final intensity 90–100% of peak grade (Fig. 1b); or (3) the delayed inflow pattern (type-III) with a peak enhancement reached during the second half of the dynamic acquisition (Fig. 1c). All of the lesions that demonstrated the rapid inflow and washout pattern (type-I) were benign (n = 21). Among nine cases presenting the plateau pattern (type-II), seven cases were benign and the other two cases were malignant, which were both follicular carcinoma. All of the 12 cases with the delayed inflow pattern (type-III) were malignant (see Table 2). The results of ERmax , Slopemax and Tpeak are plotted in Fig. 2. The Tpeak of malignant lesions (mean ± SD, 50.50 ± 11.61) was statistically higher than benign ones (14.50 ± 3.66) (P < 0.0001). The mean ± SD Slopemax was 4.99 ± 3.14 for benign lesions, and 0.94 ± 0.93 for malignant tumors. Significant difference was also found between them (P = 0.0003 < 0.05) (see Table 3). No statistical difference was found between the ERmax of malignant lesions (130.37 ± 119.92) and those of benign lesions (78.47 ± 74.36) (P = 0.15 > 0.05) (see Table 3). The ROC curves are shown in Fig. 3. The AUC for ERmax , Slopemax and Tpeak in differentiating benign thyroid lesions from malignant ones were 0.63, 0.93and 1, respectively. The ERmax cut-off value of Table 3 Tpeak , Slopemax , ERmax value of lesions.
Benign Malignant P value*
Number of cases
Tpeak
ERmax
Slopemax
28 14 –
14.50 ± 3.66 50.50 ± 11.61 P < 0.0001
78.47 ± 74.36 130.37 ± 119.92 P = 0.15
4.99 ± 3.14 0.94 ± 0.93 P = 0.0003
* P value for the comparison of Tpeak , ERmax , and Slopemax values between benign and malignant lesions. P < 0.05 was considered statistically significant.
Fig. 3. Receiver operator characteristics (ROC) curve for Tpeak , ERmax , and Slopemax measurements in this study. The dotted red line represents the ROC curve of Tpeak , the solid blue line the ROC curve for Slopemax , and solid black line the ROC curve for ERmax . The area under the ROC curve (AUCs) for Tpeak , ERmax , and Slopemax are 1.00 (standard error [SE] 0; 95% confidence interval [CI] 1.000–1.000), 0.63 (SE 0.094; CI 0.449–0.816), and 0.93 (SE 0.040; CI 0.849–1.004).
73.86 (sensitivity, 71.4%; specificity, 64.3%), Slopemax cut-off value of 2.4126 (sensitivity, 92.9%; specificity, 82.1%) and Tpeak value of 28 (sensitivity, 100%; specificity, 100%) offered the best diagnostic performances. 4. Discussion MRI offers the advantages of higher contrast resolution and the lack of ionizing radiation. With development in MR technique and thriving of dynamic enhancement technology, it could be expected that MRI might contribute to improve the diagnostic ability. DCEMRI is an emerging tool for functional assessment of a specific target tissue [9,10]. Kusunoki et al. [3] and Tezelman et al. [11] have reported 100% sensitivity for DCE-MRI to diagnose thyroid disease. With the information provided by TIC, the diagnostic accuracy and specificity of DCE-MRI were also further improved [4,12]. Thyroid is an endocrine organ and consists of all sizes of acinus, which are filled with colloid with a main component of thyroid hormone and thyroglobulin (Tg) [13]. Even under normal condition, the vascular distribution is abundant in the thyroid gland in adapt to its function. Once pathological changes take place, corresponding changes may be detected in thyroglobulin level, vascular permeability and perfusion, which eventually may alter TIC pattern. Kusunoki et al. [3] found that in most of malignant thyroid diseases and a few benign diseases that had marked cell proliferative activity, the TICs displayed the plateau pattern, and almost all benign diseases and a few well differentiated carcinomas displayed the rapid washout pattern. It was concluded that Gd-DTPA may be well take in lesions that show marked cell proliferative activity with hypervascularity, and pooled in these lesions, which resulted to exhibit the plateau pattern. Tezelman et al. [11] assess the results from 33 patients with thyroid lesions and found that malignant lesions tended to display the plateau pattern on DCE-MRI and a correlation was found between them. Different opinions were also proposed. In a recently published study of 28 patients with solitary thyroid nodule, rapid enhancement and washout pattern was seen in most malignant cases [14]. However, results of another
Y. Yuan et al. / European Journal of Radiology 81 (2012) 3313–3318
study reported by Tunca et al. [4] has demonstrated that thyroid carcinoma presented delayed washout pattern, but no plateau or rapid washout pattern. In the current study, most (85.17%, 12/14) of the malignant tumors, that was all of the 12 cases of papillary carcinoma, demonstrated the delayed inflow pattern (type-III). The TICs of the other two cases of thyroid carcinoma exhibited a plateau pattern (type-II), and both of them were follicular carcinoma. While benign lesions mainly presented a rapid inflow and washout pattern (type-I) or plateau pattern (type-II). The pathological specificity of the TIC pattern in our study is interesting and should be worth of further study. Possible explanation for variations of curves and the results among the studies may be related to the following aspects. Firstly, increased vascular permeability of malignant lesions, due to the larger space between vascular endothelial cells and incomplete basilar membrane, provides the pathophysiological foundation of the rapid inflow and washout pattern. However, the nuclear-cytoplasmic ratio of thyroid carcinoma is high. Thyroid carcinoma or thyroid nodules with a high cell proliferation index tend to demonstrate plateau pattern on DCE-MRI [3]. The component of thyroid lesions matters as well. The component of benign thyroid nodule is complex, including colloid substance, necrosis, hemorrhage, fibrous tissue and calcification, which would alter the TIC pattern. Therefore, in our study, we choose to place multiple small ROIs voxel by voxel and picked up the representative ROI, instead of manually delineating the lesion and calculating the DCE parameters over the entire tumor, in order to avoid cystic areas and calcifications in the lesions, which would introduce bias. Furthermore, the microvessel density (MVD) count is the most reliable method for estimating tumor angiogenesis [15]. Studies have proved that the MVD in follicular is higher than that in papillary carcinoma [16]. In recent years, along with the studies on specific markers of lymphatic endothelium, lymphatic vessel density (LVD) has been adopted as a new biomarker [17,18]. Reduction in the amount of D2-40 positive lymphatic vessel was detected in follicular thyroid carcinoma when compared with papillary carcinoma. Therefore, the two cases of follicular carcinoma presented plateau pattern in our studies could be explained by the relatively high MVD and low LVD in follicular carcinoma. It is also in line with the bionomics that papillary carcinoma tends to have lymphatic metastasis while hematogenous metastasis is more common in follicular carcinoma. Technologic methods, such as the parameter and sequence of MR scans, the way of deriving TICs and placing ROIs, the contrast (dose, rate and delay time) may differed from one study to another, which explains these discrepancies as well. Dynamic MRI relies on the use of a fast imaging technique with high temporal resolution and provides quantitative estimation of physiologic parameters related to perfusion and/or permeability in vivo [19]. The protocol of dynamic MRI partly differed from each other in previous studies. Kusunoki et al. [3] had take DCE-MR scans for 26 patients with thyroid masses. The overall dynamic duration was 10 min, and scans were performed at 30 s intervals for 3, 4, 5, 7 and 10 min after the injection. In another study [11] to evaluate DCE-MR performance of 30 patients with thyroid lesions, dynamic scan was executed at 3, 7, 10, 14, 17, 21 min after contrast injection. Comparing to our study, these two studies employed a longer dynamic MR scanning, which may better reflect the washout pattern. In our study, DCE-MR scans were started 10 s before the injection and 100 sets of dynamic MR images were obtained every 3.5 s over 6 min, in order to guarantee that no data was missed during the dynamic procedure and the TIC patterns obtained were sufficient for data analysis. The more or less difference of imaging protocol might have impact on the TICs generated. The MRI protocol adopted in the current study was based on previous studies of thyroid cancer [4–6] and was only adjusted to enable better image quality and sound-to-noise ratio. All the mentioned components and features would alter the result
3317
of TIC. Further studies including larger number of patients should be conducted to explain the possibility of heterogeneity between studies. The other two biomarkers Tpeak and Slopemax generated from DCE-MRI was significantly higher and lower in malignant thyroid lesions compared with benign lesions, respectively, which was in line with the TIC pattern. No statistical difference of ERmax was found between the two groups, and the results of ROC curve and AUC value of ERmax (AUC, 0.63; CI, 0.449–0.816) confirmed a limited diagnostic value with a confidential interval including 0.5. More optimal diagnostic ability was achieved by Slopemax and Tpeak value. The AUCs for Slopemax and Tpeak in differentiate benign thyroid lesions from malignant ones were 0.93 and 1, respectively. Best diagnostic performances were obtained by adopting the Slopemax cut-off value of 2.4126 (sensitivity, 92.9%; specificity, 82.1%) and Tpeak value of 28 (sensitivity, 100%; specificity, 100%). The results of the ROC analysis were promising; however, it should be mentioned that the method for determining the cutoff value may also differed from one study to another, depending on the sequence of MR imaging, the ROI selection, and so forth. Furthermore, it is not clear whether the same threshold can be proposed for patients before and after thyroidectomy, and for patients with different circulating Tg concentrations. Therefore, the main limitation of the current study should be related to the relatively small number of malignant cases included in the study (n = 14). The number of patients within each subgroup had, to certain extent, limited the subgroup analysis based on different histological types and other patients’ characteristics. For instance, the limited amount of patients with follicular carcinoma (n = 2) deprived the statistically ability of explaining DCE-MRI results. Therefore, further studies are worth to be conducted in larger group of patients, which would further confirm the diagnostic ability of dynamic MR and evaluate cut-off level depending on the characteristics of patients, lesions and imaging techniques. Another limitation is related to the method we used to generate the representative ROI for the evaluation. Most of the previous studies chose to manually delineate the lesion and calculate the DCE parameters over the entire tumor. In our study, we placed multiple small ROIs voxel by voxel over the entire lesions and then picked up the representative ROI with the maximal enhancement. This method would rule out the possibility of including cystic or calcific components along with the solitary tumor, since the lesions analyzed in our study are not only solitary tumors but also those with cyst and calcification. The subjective bias might be reduced as well, as the number of ROIs acquired in the same lesion should be stable. The bias might be introduced during the procedure of selecting the representative ROIs. As a result, two readers independently selected the representative ROI and a third reviewer assessed discrepant items according to the majority opinion was used for analysis. FNAB, especially US-guided FNAB, has an essential role in the evaluation of patients with thyroid nodules to reduce the rate of unnecessary thyroid surgery and triage patients to appropriate surgery. However, there are still a low incidence of both falsepositive and false-negative diagnosis of FNA of thyroid nodules [20]. For example, it is more likely to get specimens inadequate for cytology analysis for nodules smaller than 1.5 cm. False-negative results may cause delay in the surgical treatment of patients with thyroid carcinoma and lead to the development of local invasion in some patients. On the other side, sampling error may be especially relevant for larger nodules even with US guidance [21]. Cystic or calcific changes in lesions can cause further diagnostic problems due to inadequate sampling. In our study, preoperative FNAB was performed in only one patient. This could be attibuted to the abovementioned limitations of FNAB, as well as the patients’ willing to refuse an invasive examination. The cost was also a consideration.
3318
Y. Yuan et al. / European Journal of Radiology 81 (2012) 3313–3318
However, we have to admit that it is a limitation of our study for not comparing the results of DCE-MRI with US guided-FNAB, which are being clinically recommended. 5. Conclusions Dynamic MRI could be clinically useful in the differential diagnosis of benign thyroid lesions from malignant lesions. The rapid inflow and washout pattern of time-intensity curve indicates a benign lesion, and most malignant lesions demonstrated the delayed inflow pattern. The Slopemax and Tpeak performed well in differentiating benign thyroid lesions from malignant ones. Therefore, further studies are worth to be conducted in larger group of patients, which would further confirm the diagnostic ability of dynamic MR and evaluate cut-off level depending on the characteristics of patients, lesions and imaging techniques. References [1] Frasoldati A, Valcavi R. Challenges in neck ultrasonography: lymphadenopathy and parathyroid glands. Endocrine Practice 2004;10(3):261–8. [2] Boland GW, Lee MJ, Mueller PR, Mayo-Smith W, Dawson SL, Simeone JF. Efficacy of sonographically guided biopsy of thyroid masses and cervical lymph nodes. American Journal of Roentgenology 1993;161(5):1053–6. [3] Kusunoki T, Murata K, Nishida S, Tomura T, Inoue M. Histopathological findings of human thyroid tumors and dynamic MRI. Auris, Nasus, Larynx 2002;29(4):357–60. [4] Tunca F, Giles Y, Salmaslioglu A, et al. The preoperative exclusion of thyroid carcinoma in multinodular goiter: dynamic contrast-enhanced magnetic resonance imaging versus ultrasonography-guided fine-needle aspiration biopsy. Surgery 2007;142(6):992–1002. [5] Noworolski SM, Fischbein NJ, Kaplan MJ, et al. Challenges in dynamic contrastenhanced MRI imaging of cervical lymph nodes to detect metastatic disease. Journal of Magnetic Resonance Imaging 2003;17(4):455–62. [6] Schueller-Weidekamm C, Kaserer K, Schueller G, et al. Can quantitative diffusion-weighted MR imaging differentiate benign and malignant cold thyroidnodules initial results in 25 patients. AJNR American Journal of Neuroradiology 2009;30:417–22.
[7] Unlu E, Albayram S, Temizoz O, Caloglu M, Tamer Y. The role of dynamic contrast-enhanced magnetic resonance imaging parameters in head-neck tumors. Rivista di Neuroradiologia 2006;19:245–56. [8] Yabuuchi H, Fukuya T, Tajima T, Hachitanda Y, Tomita K, Koga M. Salivary gland tumors: diagnostic value of gadolinium-enhanced dynamic MR imaging with histopathologic correlation. Radiology 2003;226(2): 345–54. [9] Choyke PL, Dwyer AJ, Knopp MV. Functional tumor imaging with dynamic contrast-enhanced magnetic resonance imaging. Journal of Magnetic Resonance Imaging 2003;17(5):509–20. [10] O’Connor JP, Jackson A, Asselin MC, Buckley DL, Parker GJ, Jayson GC. Quantitative imaging biomarkers in the clinical development of targeted therapeutics: current and future perspectives. Lancet Oncology 2008;9(8): 766–76. [11] Tezelman S, Giles Y, Tunca F, et al. Diagnostic value of dynamic contrast medium-enhanced magnetic resonance imaging in preoperative detection of thyroid carcinoma. Archives of Surgery 2007;142(11): 1036–41. [12] Kusunoki T, Murata K, Hosoi H, Nishida S, Tomura T, Inoue M. Malignancies of human thyroid tumors and dynamic magnetic resonance imaging (MRI). Auris, Nasus, Larynx 1998;25(4):419–24. [13] Wissig SL. The anatomy of secretion in the follicular cells of the thyroid gland. I. The fine structure of the gland in the normal rat. Journal of Biophysical and Biochemical Cytology 1960;7:419–32. [14] Gupta N, Norbu C, Goswami B, et al. Role of dynamic MRI in differentiating benign from malignant follicular thyroid nodule. Auris, Nasus, Larynx 2011;38(6):718–23. [15] Weidner N. Tumor angiogenesis: review of current applications in tumor prognostication. Seminars in Diagnostic Pathology 1993;10(4):302–13. [16] Ramsden JD. Angiogenesis in the thyroid gland. Journal of Endocrinology 2000;166(3):475–80. [17] Giorgadze TA, Baloch ZW, Pasha T, Zhang PJ, Livolsi VA. Lymphatic and blood vessel density in the follicular patterned lesions of thyroid. Modern Pathology 2005;18(11):1424–31. [18] de la Torre NG, Buley I, Wass JA, Turner HE. Angiogenesis and lymphangiogenesis in thyroid proliferative lesions: relationship to type and tumour behaviour. Endocrine Related Cancer 2006;13(3):931–44. [19] Oostendorp M, Post MJ, Backes WH. Vessel growth and function: depiction with contrast-enhanced MR imaging. Radiology 2009;251(2):317–35. [20] Yeh MW, Demircan O, Ituarte P, Clark OH. False-negative fine-needle aspiration cytology results delay treatment and adversely affect outcome in patients with thyroid carcinoma. Thyroid 2004;14(3):207–15. [21] Porterfield Jr JR, Grant CS, Dean DS, et al. Reliability of benign fine needle aspiration cytology of large thyroid nodules. Surgery 2008;144:963–8.