Application of whole-lesion intravoxel incoherent motion analysis using iZOOM DWI to differentiate malignant from benign thyroid nodules

Abstract

Background

Intravoxel incoherent motion (IVIM) parameters derived from iZOOM diffusion-weighted imaging (DWI) can be used to quantitatively identify malignant thyroid nodules. However, the criteria are not standardized.

Purpose

To determine IVIM parameter values derived from iZOOM DWI that can differentiate malignant from benign thyroid nodules.

Material and Methods

Forty-two patients with 46 pathologically confirmed thyroid nodules. All patients underwent preoperative examinations with conventional and iZOOM DWI. A three-dimensional region of interest was drawn on multiple slices to cover the entire nodule. IVIM parameters (D, pure diffusion; f, perfusion fraction; and D*, pseudodiffusion) were measured. The IVIM parameters of the malignant and benign thyroid nodules were compared using independent samples t-tests, a multiple logistic regression, and receiver operating characteristic (ROC) curve analysis.

Results

D, D*, and f exhibited good reproducibility. The D and f values in the malignant nodules (D = [0.72 ± 0.14] × 10⁻³ mm²/s, [29.94 ± 7.36] %) were significantly lower than those in benign nodules (D = [1.23 ± 0.35] × 10⁻³ mm²/s, [36.00 ± 8.35] %). The D value achieved the highest area under curve (0.939). The optimal cut-off value for D was 0.87 × 10⁻³ mm²/s (sensitivity = 95.83%, specificity = 90.91%, positive predictive value [PPV] = 92.00%, negative predictive value [NPV] = 95.24%, and accuracy rate = 93.48%). The area under ROC curve (AUC) of D combined with the f value did not significantly differ from that of D.

Conclusion

IVIM parameter D derived from iZOOM DWI may be helpful in differentiating malignant from benign thyroid nodules.

Keywords

Intravoxel incoherent motion magnetic resonance imaging reduced field of view thyroid nodules

Introduction

Thyroid cancer is the most common malignant disease of the endocrine system and its incidence is rapidly increasing (1,2). Because of the widespread use of ultrasonography (US), the detection rate of thyroid carcinomas has recently increased. US is the preferred modality for the assessment of thyroid nodules and the reported accuracy of US in predicting malignant thyroid nodules has a sensitivity in the range of 26–87% and a specificity in the range of 40–93% (3 –6). However, evaluations by US are operator-dependent and no single sonographic feature has a sufficient diagnostic value for thyroid cancer and metastatic lymph nodes. Nonetheless, the use of US has been further emphasized for the personalized management of patients with thyroid nodules. Fine-needle aspiration biopsy (FNAB) is regarded as the most accurate procedure to identify malignant nodules, but 15–20% of FNAB results have been found to be non-diagnostic or indeterminate because its accuracy may vary depending on many factors, including the biopsy technique and location, even when performed by experienced professionals (7,8). Radionuclide imaging is another important tool that provides excellent functional information; however, this technique has a poor spatial resolution and is associated with the administration of a substantial dose of radiation to the entire body and the thyroid gland in particular (9). Computed tomography perfusion (CTP) imaging and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) serve as potential imaging biomarkers for diagnosis. However, these techniques require rigorous data analysis processes, limiting their routine clinical application (10). Developing a non-invasive and reliable technique to differentiate benign from malignant lesions remains a considerable challenge.

MRI somewhat unifies the advantages of the other modalities because of its ability to objectively examine anatomy at a rather high spatial resolution and a high intrinsic soft tissue contrast while also providing functional tissue information without using radiation. Diffusion-weighted imaging (DWI) has been reported to be valuable in differentiating malignant from benign thyroid nodules (11 –13). The main challenge in applying DWI to thyroid imaging lies in the artifacts and distortion caused by magnetic susceptibility. To overcome this challenge, iZOOM DWI, which is a special protocol with a shorter echo train length that uses a two-dimensional (2D) spatially selective echo-planar radiofrequency (RF) pulse, has been developed. iZOOM DWI is less sensitive to inhomogeneity in magnetic fields and has been previously applied to imaging of the spinal cord, pancreas, and breast (14 –16). Compared with the regular DWI model, the intravoxel incoherent motion (IVIM) model has the ability to separate tissue diffusivity from microcapillary perfusion and is, therefore, promising for use in tumor diagnosis (17 –19). However, the performance of IVIM parameters derived from iZOOM DWI in differentiating thyroid nodules is unclear. Therefore, this study was performed to assess whether IVIM parameters derived from iZOOM DWI can distinguish between malignant and benign thyroid nodules.

Material and Methods

Patients

After receiving approval from the Ethics Committee, 59 consecutive patients with pathologically confirmed benign or malignant thyroid nodules who underwent preoperative MRI, including an iZOOM DWI sequence, between November 2015 and April 2017 were enrolled in this study after obtaining their signed informed consent. The recommendation for surgical treatment was based on the US and FNAB findings. The inclusion criteria were based on the American Thyroid Association Management Guidelines for adult patients with thyroid nodules and differentiated thyroid cancer (20). For sections of nodules suspected of being benign, the applied surgical treatment was based on patient preference. The exclusion criteria were as follows: patients who underwent prior treatment (radiation therapy or thyroid lobectomy) or an FNAB of the thyroid < 4 weeks before the MRI examination (n = 5); routine neck MRI and iZOOM DWI examinations of nodules < 1 cm (n = 6); any MRI-incompatible metallic device in the patients’ bodies; tremors that could cause motion artifacts during the MRI examination; and claustrophobia (n = 6). Finally, 42 patients (7 men, 35 women; mean age = 46 ± 13 years; age range = 13–66 years) with 46 thyroid nodules were included in this study.

MRI protocol

All patients underwent preoperative examinations (Philips 3.0T Ingenia, Philips Medical System, The Netherlands) with both conventional and iZOOM DWI using an eight-channel phased-array carotid coil. The room temperature was kept constant at 21 °C. First, T1-weighted/T2-weighted (T1W/T2W) turbo spin-echo sequences (field of view [FOV] = 22 × 22 × 29 cm; voxel size = 0.85 × 0.85 mm²/0.76 × 0.76 mm²; number of slices = 20; slice thickness = 4 mm; slice spacing = 0.6 mm; TR/TE = 525/36 ms and 3600/100 ms, respectively; turbo factor = 8 and 22; flip angle = 90°; and acquisition time = 6 min 34 s) were acquired to gain sufficient anatomical information. Second, iZOOM DWI using a 2D RF pulse was performed with the following parameters: TR/TE = 1351/69 ms; turbo factor = 29; EPI factor = 29; FOV = 160 × 47 mm²; voxel size, 1.50 × 1.50 mm; NSA = 4; 10 slices at a thickness of 5.0 mm with a 1.0 mm gap; and eight b-values (0, 20, 50, 100, 200, 400, 600, and 990 s/mm²) were used, and the acquisition time was 5 min 37 s.

Image postprocessing

The IVIM model equation described by Le Bihan et al. is Sb/S0 = (1−f) exp(−bD) + exp(−bD^∗) (17). All three parameters were derived directly from the curve fitting. No cut-off was used. In the curve fitting, both D and D* were set to be positive and < 0.01, while D* should be larger than D. f was in the range of 0–1. The IVIM parameters were calculated pixel-by-pixel.

All images were retrospectively and independently analyzed by two radiologists (with 20 and 11 years of experience, respectively) blinded to the pathological findings. The maximum diameters of the nodules were recorded. A three-dimensional (3D) region of interest (ROI) was manually drawn on multiple slices to cover the entire nodule, and the margins of the whole lesion were traced on all continuous sections of the DWI-obtained IVIM scans, including the cystic, necrotic, and hemorrhagic portions. The non-linear fitting of the biexponential model was performed using MATLAB (Math Works, MA, USA). The IVIM parameters (D, pure diffusion; f, perfusion fraction; and D*, pseudodiffusion) were measured, and the final value was obtained as the mean value calculated by the two radiologists.

Statistical analysis

The data were first subjected to a normal distribution test and homogeneity of variance testing using the Levene test. If the data were normally distributed, the continuous variables were expressed as the mean ± standard deviation. The intra-observer and inter-observer agreement regarding the IVIM-derived parameters were analyzed via the interclass correlation coefficient (ICC). The ICC model was a two-way mixed effects model. The mean major diameter and the mean values of D, D*, and f of the malignant and benign groups were compared using independent sample t-tests. A receiver operating characteristic (ROC) analysis was performed to determine an optimal cut-off value for differentiating benign nodules from malignant nodules. The optimal threshold was chosen according to the Youden index. The area under the ROC curve (AUC) was calculated. Furthermore, the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy rate were determined. The parameters found to be statistically significant in the univariate analysis were subsequently entered into multiple logistic regression models to gauge their independent predictive value in predicting thyroid nodules. The statistical analysis was performed using MedCalc Online, version 16.2 (Medcalc Software, Mariakerke, Belgium) and SPSS (18.0 for Windows, SPSS, Chicago, IL, USA). A P value < 0.05 indicated statistically significant differences.

Results

The MRI results revealed 46 thyroid nodules in 42 patients. The pathological findings confirmed 22 benign nodules and 24 malignant nodules with the following diagnoses: papillary thyroid cancer (n = 23); medullary thyroid carcinoma (n = 1); nodular goiter (n = 20); and thyroid adenoma (n = 2). Furthermore, 38 patients had solitary thyroid nodules and four patients had two nodules. No significant difference was found between the mean major diameter of the benign nodules and that of the malignant nodules (malignant nodules = 3.06 ± 1.22 cm; benign nodules = 3.18 ± 1.65 cm; P > 0.05). The patient demographics are presented in Table 1.

Table 1.

Pathological results and patient characteristics.

Variable	n (%)
Patients (n)	42
Female	35 (83.33)
Male	7 (16.67)
Mean age (years)*	46 ± 13
Nodule number	46
Single	38 (82.61)
Multiple	4 (8.70)
Nodule origin
Malignant	24 (52.17)
Papillary thyroid cancer	23 (50.00)
Medullary thyroid carcinoma	1 (2.17)
Benign	22 (47.83)
Nodular goiter	20 (43.48)
Thyroid adenoma	2 (4.35)
Nodule size (cm)
Benign*	3.18 ± 1.65
Malignant*	3.06 ± 1.22

*Values are expressed as mean ± standard deviations.

In the benign group, the values of the IVIM parameters were as follows: D = (1.23 ± 0.35) × 10⁻³ mm²/s, f = (36.00 ± 8.35) %, and D* = (10.91 ± 2.04) × 10⁻³ mm²/s. In the malignant group, the values of the IVIM parameters were as follows: D = (0.72 ± 0.14) × 10⁻³ mm²/s, f = (29.94 ± 7.36) %, and D* = (11.23 ± 2.20) × 10⁻³ mm²/s. Among the three IVIM parameters, the D and f values in the malignant nodules were significantly lower than those in the benign nodules (P < 0.0001, P = 0.012). However, there was no significant difference in the D* value between the malignant and benign nodules (P = 0.6069) (Figs. 1 –3 and Table 2).

Fig. 1.

Images of a 48-year-old woman with left lobe papillary thyroid cancer. (a) Axial T2W image. (b) DWI with b = 20 s/mm². (c) DWI with b = 990 s/mm². (d–f) IVIM color map (from left to right representing D, f, and D*). D = 0.79 × 10⁻³ mm²/s, f = 18.50%, and D* = 14.84 × 10⁻³ mm²/s.

Fig. 2.

Images of a 60-year-old woman with right lobe nodular goiter. (a) Axial T2W with selected partial inversion recovery (SPIR) image. (b) DWI with b = 20 s/mm². (c) DWI with b = 990 s/mm². (d–f) IVIM color map (from left to right representing D, f, and D*). D = 1.03 × 10⁻³ mm²/s, f = 43.00%, and D* = 8.56 × 10⁻³ mm²/s.

Fig. 3.

Box-and-whisker plots of the IVIM parameters (D, f, and D*) in the malignant and benign thyroid nodules.

Table 2.

Comparison of IVIM parameters between the benign and malignant groups.

IVIM parameters	Benign	Malignant	P
D (×10^–3 mm²/s)	1.23 ± 0.35	0.72 ± 0.14	< 0.0001
D* (×10^–3 mm²/s)	10.91 ± 2.04	11.23 ± 2.20	0.6069
f (%)	36.00 ± 8.35	29.94 ± 7.36	0.012

The data are expressed as mean ± standard deviation.

D, pure diffusion; f, perfusion fraction; D*, pseudodiffusion.

The respective optimal cut-off values based on the ROC curve analyses (including the respective sensitivity, specificity, PPV, NPV, and accuracy rate) are summarized in Table 3. The values were as follows: D = 0.87 × 10⁻³ mm²/s (the D values of the benign nodules were greater than this value: 95.83%, 90.91%, 92.00%, 95.24%, and 93.48%); D* = 11.91 × 10⁻³mm²/s (the D* values of the benign nodules were greater than this value: 50.00%, 68.18%, 63.16%, 55.56%, and 58.70%); and f = 29.20% (the f values of the benign nodules were greater than this value: 54.17%, 81.82%, 76.47%, 62.07%, and 67.39%). The respective ROC curves and AUC are shown in Fig. 4. D achieved the highest AUC (0.939; 95% confidence interval [CI] = 0.827–0.988) in differentiating the malignant from benign thyroid nodules. The multivariate regression analysis revealed that compared with the best individual parameter from the univariate analysis, the addition of other parameters in different combinations of the model did not result in any significant improvement in the diagnostic accuracy (AUC = 0.962, P = 0.475).

Fig. 4.

ROC curve analysis for discriminating malignant from benign thyroid nodules with D combined with f (yellow line), D (blue line), f (orange line), and D* (green line) values.

Table 3.

Diagnostic performance of IVIM parameters in distinguishing benign from malignant thyroid nodules.

IVIM parameters	Cut-off value	AUC (95% CI)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	Accuracy rate (%)
D	0.87 × 10⁻³mm²/s	0.939 (0.827–0.988)	95.83	90.91	92.00	95.24	93.48
D*	11.91 × 10⁻³mm²/s	0.553 (0.399–0.700)	50.00	68.18	63.16	55.56	58.70
f	29.20%	0.708 (0.556–0.833)	54.17	81.82	76.47	62.07	67.39

There was excellent intra-observer and inter-observer agreement in the measurements of D, D*, and f (Table 4).

Table 4.

Intra-observer and inter-observer agreement (ICC) in the measurements of the thyroid nodule IVIM parameters.

	Intra-observer ICC (95% CI)	Inter-observer ICC (95% CI)
Parameters
D	0.988 (0.978–0.994)	0.970 (0.946–0.984)
D*	0.905 (0.827–0.948)	0.894 (0.808–0.941)
F	0.937 (0.885–0.966)	0.951 (0.911–0.973)

Discussion

This retrospective study evaluated the diagnostic value of IVIM parameters extracted from iZOOM DWI images in differentiating malignant from benign thyroid nodules. This novel study measured thyroid nodules with IVIM parameters and ROC curves using 3D whole lesions. The IVIM parameters D and f in the malignant group were significantly lower than those in the benign group. D was the most powerful parameter based on the AUC (0.939).

The thyroid gland is a small organ located near the air–tissue boundary. There were considerable technical challenges in performing high-quality DWI of the head and neck using the full FOV (fFOV) single shot echo-planar imaging (EPI) technique, which is the most commonly used sequence for DWI acquisition. The image quality of EPI-DWI was frequently deteriorated due to susceptibility artifacts because the EPI sequence is prone to phase error accumulation. Moreover, the thyroid gland area was especially sensitive to magnetic inhomogeneity due to the presence of dental alloy or its complex structure with many boundaries, such as the trachea and bones. These susceptibility artifacts cause image distortion or signal loss, which more seriously affect the performance of the fFOV EPI-DWI technique and its applicability to quantitative investigation. Using a fast spin-echo sequence for DWI could effectively reduce these magnetic susceptibility artifacts and increase the signal-to-noise ratio (SNR). However, the scan time of this sequence is long due to multiple RF refocusing pulses and the RF heating restrictions with an increase in the specific absorption rate (21). Therefore, a technique that can minimize these artifacts and provide higher-quality images of thyroid glands is clinically required to replace the fFOV DWI method. iZOOM DWI is a neoteric sequence with 2D RF excitation that excites only the ROI and falls under reduced FOV DWI. This technique is effective in providing high-resolution images with fewer artifacts and less distortion and provides more reliable and repeatable results (22). Furthermore, this method has been proven to be a feasible quantitative imaging tool for investigating thyroid glands in healthy volunteers (23). Thus far, there are no reports about the application of iZOOM DWI technology to the diagnosis of benign and malignant thyroid nodules.

The 3D whole-lesion analysis method was used in the present study to evaluate entire thyroid nodules. In previous studies, most methods used measurements derived from single or multiple images with nonuniform results (11 –13,24). Subjective differences in the selection and delineation of the ROI might lead to reductions in intra-observer and inter-observer agreement. Recent research has suggested that volumetric analyses demonstrate better inter-observer reproducibility than single-section ROI analyses in rectal cancer (25). The intra-observer and inter-observer reproducibility of measuring the IVIM parameters in this study were also excellent. Therefore, the combination of 2D RF iZOOM DWI and the whole-lesion analysis method might help eliminate ROI-delineated bias and enhance the assessment of intra-lesion heterogeneity in thyroid nodules.

The conventional DWI method uses a mono-exponential model and cannot differentiate microcirculation-related diffusion from water diffusion in tissues. Using high b-values is important for obtaining real DW images and minimizing the perfusion effect. Some previous studies used a higher b-value (b = 1000 s/mm²) and achieved higher sensitivity and specificity (nearly 100%) in differentiating malignant from benign thyroid nodules (26,27). However, there is a risk of noise contamination and distorted image quality with subsequent, apparently unreliable dispersion coefficient (ADC) measurements. According to a recent meta-analysis, most studies (11/15) adopted the b-value (b ≤ 800 s/mm²) for differentiating benign from malignant thyroid nodules. The sensitivity of quantitative DWI and ADC was 0.90–0.91 and the specificity was 0.93–0.95. However, the threshold value of ADC for differentiating benign from malignant thyroid lesions showed great variability (0.36–2.56 × 10⁻³mm²/s) (28,29). Clinical research requires a new, more stable, and reproducible method for identifying benign and malignant thyroid nodules.

To compare with conventional DWI, biexponential model fitting was used in the IVIM analysis. The IVIM-derived parameters included the pure diffusion coefficient (D), perfusion-related incoherent microcirculation (D*, pseudodiffusion coefficient), and perfusion fraction (f). Multiple b-values were used to sample the diffusion signals because diffusion (D)- and microcirculation (D* and f)-related diffusion could be separated, allowing for the measurable microcirculation effect to be obtained. Another advantage of IVIM analyses is that if the distribution of the b-value is reasonable, the biomarkers are not affected by the b-value. Using the IVIM model and sufficiently low b-values (< 200 s/mm²), the microcirculation- or perfusion-related effects could be separated from pure tissue diffusion (D); then, the perfusion characteristics (D*) and their volume fraction (f) could be derived (30,31). A previous study demonstrated that IVIM measurements in the thyroid gland of healthy volunteers were feasible using the proposed protocol (32). In the present study, IVIM technology was used to determine the accuracy and reproducibility of identifying benign and malignant thyroid nodules. The following eight b-values were used: five b-values (0, 20, 50, 100, and 200 s/mm²) that have been recommended within the perfusion-sensitive range and three higher b-values (400, 600, and 990 s/mm²) due to their stability and reproducibility. The malignant nodules were still apparent because they exhibited diffusion-limited signals at higher b-values, whereas the diffusion-limited signals in the benign nodules decreased with increasing b-values. The D value achieved the highest AUC (0.939). The optimal cut-off value for D was 0.87 × 10⁻³ mm²/s (sensitivity = 95.83%, specificity = 90.91%, and accuracy rate = 93.48%). The D value was more valuable than the D* and f values in differentiating the malignant from benign thyroid nodules. This phenomenon revealed that the water diffusion in the thyroid nodules was more weighted than the microcirculation-related diffusion. This finding is consistent with the pathological characteristics of thyroid nodules in which major modifications can be observed in follicular-derived neoplasms (33). The specificity was slightly lower than that reported in previous studies (26,27,34), which may be related to the 3D whole-lesion measurement method that includes all heterogeneous components, especially the cystic component, in the malignant nodules, which could increase the D values and improve the false-positive rate. The multiple regression analysis showed that the AUC of D combined with the f value could reach 0.962 (P = 0.475). Although the AUC did not significantly differ from that of the best univariate parameter (D), the f value, which was related to perfusion, was still worthy of attention.

This study had some limitations. First, this study was retrospective and included only resected nodules with pathological findings, which could have led to inherent biases in the patient selection. In addition, the patient sample size was not sufficiently large. Second, nodules < 1 cm were not included in this research because of the poor resolution of DW images for small nodules. Third, the correlation between the ADC values and parameter (D) derived from IVIM was unexplored; however, we aim to evaluate this correlation in future studies. Fourth, these data were derived using a specific vendor sequence, but other major vendors have offered a similar sequence that can be applied to thyroid examinations and need further clinical application. Finally, this study discussed only IVIM, which is a method of functional imaging. The most recent study using diffusion tensor imaging and diffusion kurtosis imaging of the head and neck provides a reference for future thyroid research (35,36).

In conclusion, whole-lesion IVIM parameters (D) derived from iZOOM DWI may help differentiate malignant from benign thyroid nodules at 3T.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Jemal

Bray

Center

, et al. Global cancer statistics. CA Cancer J Clin 2011; 61:69–90.

Lim

Devesa

Sosa

, et al. Trends in thyroid cancer incidence and mortality in the United States, 1974–2013. JAMA 2017; 317:1338–1348.

Remonti

Kramer

Leitao

, et al. Thyroid ultrasound features and risk of carcinoma: a systematic review and meta-analysis of observational studies. Thyroid 2015; 25:538–550.

Kim

Son

, et al. Preoperative prediction of the extrathyroidal extension of papillary thyroid carcinoma with ultrasonography versus MRI: a retrospective cohort study. Int J Surg 2014; 12:544–548.

Shin

Baek

Chung

, et al. Ultrasonography diagnosis and imaging-based management of thyroid nodules: revised Korean Society of Thyroid Radiology consensus statement and recommendations. Korean J Radiol 2016; 17:370–395.

Hoang

Lee

, et al. US features of thyroid malignancy: pearls and pitfalls. Radiographics 2007; 27:847–860; discussion 861–845.

Iannuccilli

Cronan

Monchik

JM.

Risk for malignancy of thyroid nodules as assessed by sonographic criteria: the need for biopsy.

J Ultrasound Med 2004; 23:1455–1464.

Wojtczak

Sutkowski

Bolanowski

, et al. The prognostic value of fine-needle aspiration biopsy of the thyroid gland - analysis of results of 1078 patients. Neuro Endocrinol Lett 2012; 33:511–516.

Lundh

Norden

Nilsson

, et al. Reduced iodide transport (stunning) and DNA synthesis in thyrocytes exposed to low absorbed doses from 131I in vitro. J Nucl Med 2007; 48:481–486.

10.

Ben-David

Sadeghi

Rezaei

, et al. Semiquantitative and quantitative analyses of dynamic contrast-enhanced magnetic resonance imaging of thyroid nodules. J Comput Assist Tomogr 2015; 39:855–859.

11.

Erdem

Muammer

, et al. Diffusion-weighted images differentiate benign from malignant thyroid nodules. J Magn Reson Imaging 2010; 31:94–100.

12.

Abdel-Rahman

Abowarda

Abdel-Aal

SM.

Diffusion-weighted MRI and apparent diffusion coefficient in differentiation of benign from malignant solitary thyroid nodule.

Egyp J Radiol Nucl Med 2016; 47:1385–1390.

13.

Yue

Shen

, et al. Diagnostic value of diffusion-weighted MR imaging in thyroid disease: application in differentiating benign from malignant disease. BMC Med Imaging 2013; 13:23.

14.

Saritas

Cunningham

Lee

, et al. DWI of the spinal cord with reduced FOV single-shot EPI. Magn Reson Med 2008; 60:468–473.

15.

Zaharchuk

Saritas

Andre

, et al. Reduced field-of-view diffusion imaging of the human spinal cord: comparison with conventional single-shot echo-planar imaging. AJNR Am J Neuroradiol 2011; 32:813–820.

16.

Dong

, et al. Study of the reduced field-of-view diffusion-weighted imaging of the breast. Clin Breast Cancer 2014; 14:265–271.

17.

Le Bihan

Breton

Lallemand

, et al. Separation of diffusion and perfusion in intravoxel incoherent motion MR imaging. Radiology 1988; 168:497–505.

18.

Zhang

Liang

, et al. Intravoxel incoherent motion MRI for the differentiation of benign, intermediate, and malignant solid soft-tissue tumors. J Magn Reson Imaging 2017; 46:1611–1618.

19.

Surov

Meyer

Hohn

, et al. Correlations between intravoxel incoherent motion (IVIM) parameters and histological findings in rectal cancer: preliminary results. Oncotarget 2017; 8:21974–21983.

20.

Haugen

Alexander

Bible

, et al. 2015 American thyroid association management guidelines for adult patients with thyroid nodules and differentiated thyroid cancer: the American Thyroid Association Guidelines Task force on thyroid nodules and differentiated thyroid cancer. Thyroid 2016; 26:1–133.

21.

Mikayama

Yabuuchi

Sonoda

, et al. Comparison of intravoxel incoherent motion diffusion-weighted imaging between turbo spin-echo and echo-planar imaging of the head and neck. Eur Radiol 2018; 28:316–324.

22.

Taviani

Nagala

Priest

, et al. 3T diffusion-weighted MRI of the thyroid gland with reduced distortion: preliminary results. Br J Radiol 2013; 86:20130022.

23.

Hatzoglou

Banerjee

, et al. Repeatability investigation of reduced field-of-view diffusion-weighted magnetic resonance imaging on thyroid glands. J Comput Assist Tomogr 2015; 39:334–339.

24.

Noda

Kanematsu

Goshima

, et al. MRI of the thyroid for differential diagnosis of benign thyroid nodules and papillary carcinomas. Am J Roentgenol 2015; 204:W332–335.

25.

Nougaret

Vargas

Lakhman

, et al. Intravoxel incoherent motion-derived histogram metrics for assessment of response after combined chemotherapy and radiation therapy in rectal cancer: initial experience and comparison between single-section and volumetric analyses. Radiology 2016; 280:446–454.

26.

Elshafey

Elattar

Mlees

, et al. Role of quantitative diffusion-weighted MRI and 1H MR spectroscopy in distinguishing between benign and malignant thyroid nodules. Egyp J Radiol Nucl Med 2014; 45:89–96.

27.

Mutlu

Sivrioglu

Sonmez

, et al. Role of apparent diffusion coefficient values and diffusion-weighted magnetic resonance imaging in differentiation between benign and malignant thyroid nodules. Clin Imaging 2012; 36:1–7.

28.

Chen

Bao

, et al. Diffusion-weighted MRI in differentiating malignant from benign thyroid nodules: a meta-analysis. BMJ Open 2016; 6:e008413.

29.

Chen

, et al. On the utility of quantitative diffusion-weighted MR imaging as a tool in differentiation between malignant and benign thyroid nodules. Acad Radiol 2014; 21:355–363.

30.

Iima

Le Bihan

Clinical intravoxel incoherent motion and diffusion MR imaging: past, present, and future.

Radiology 2016; 278:13–32.

31.

Lim

Jee

Jung

, et al. Intravoxel incoherent motion diffusion-weighted MR imaging for differentiation of benign and malignant musculoskeletal tumors at 3 T. Br J Radiol 2018; 91:20170636.

32.

Becker

Wurnig

Finkenstaedt

, et al. Non-parametric intravoxel incoherent motion analysis of the thyroid gland. Heliyon 2017; 3:e00239.

33.

Lam

- y. Pathology of endocrine tumors update: world health organization new classification 2017—other thyroid tumors. AJSP: Reviews & Reports 2017; 22:209–216.

34.

Razek

Sadek

Kombar

, et al. Role of apparent diffusion coefficient values in differentiation between malignant and benign solitary thyroid nodules. Am J Neuroradiol 2008; 29:563–568.

35.

Abdel Razek

AAK.

Routine and advanced diffusion imaging modules of the salivary glands.

Neuroimaging Clin N Am 2018; 28:245–254.

36.

Abdel Razek

AAK.

Characterization of salivary gland tumours with diffusion tensor imaging.

Dentomaxillofac Radiol 2018; 47:20170343.