Diffusion-weighted MRI in the evaluation of the thyroid nodule: Comparison between integrated-shimming EPI and conventional 3D-shimming EPI techniques

This study aimed to evaluate whether a prototype echo planar imaging sequence with integrated-shimming (iShim-EPI) can improve image quality in the thyroid gland in comparison to 3D-volume shimming echo planar imaging (3D-Shim-EPI), and to compare ADC values derived from iShim-EPI with those of 3D-Shim-EPI. Twenty-one patients with thyroid disease were enrolled and underwent axial DWIs with iShim-EPI and 3D-Shim-EPI using a 3 Tesla magnetic resonance scanner in this prospective study. Both sets of DWI images were evaluated by two independent observers who identified susceptibility and ghost artifacts and evaluated the images' capacity to detect thyroid nodules using quantitative scores. The ADC values of the thyroid nodules and the normal thyroid gland were measured two times within a 4-week period. The reproducibility was evaluated using the intraclass correlation coefficient (ICC) and Bland-Altman plots. There were significant differences in the image quality scores for susceptibility (2.81 ± 0.37 vs. 1.93 ± 0.29, p < 0.001), ghost artifacts (2.95 ± 0.15 vs. 1.93 ± 0.29, p < 0.001) and the detectability of thyroid nodules (3.00 ± 0.00 vs. 2.55 ± 0.75, p = 0.008) between the iShim-EPI and 3D-Shim-EPI techniques, except for the ADC values of the thyroid nodules (1.607 ± 0.466×10−3 mm2/s vs. 1.561 ± 0.483 × 10−3 mm2/s, p = 0.184) and contralateral normal thyroid gland (1.295 ± 0.340 × 10−3 mm2/s vs.1.279 ± 0.411 × 10−3 mm2/s, p = 0.777). Both techniques demonstrated excellent agreement between the ADC values using the ICC (range, 0.963 to 0.999) and Bland-Altman plots. The iShim-EPI technique demonstrated significantly higher image quality compared with the conventional 3D-Shim-EPI technique, with no significant differences in the ADC values.


INTRODUCTION
The thyroid nodule is a common disorder of the thyroid gland [1]. The prevalence is highest in females. The overall incidence of thyroid nodules in the general population is 4-7%, of which 5-7% possess the potential for malignancy [2,3]. Although a small proportion of thyroid nodules are malignant, early and accurate diagnosis of the malignant thyroid is crucial for effective treatment of these patients.
Several imaging modalities, including ultra sonography, computed tomography, radionuclide scinti Clinical Research Paper graphy and positron emission tomography, have been used to distinguish malignant thyroid nodules from benign ones. However, these imaging techniques can have several drawbacks, including ionizing radiation, poor accuracy and subsequent limited utility [4]. With the advancement of imaging technologies, magnetic resonance imaging (MRI) is becoming a useful tool to evaluate thyroid gland disease due to its noninvasiveness and better image contrast compared with other imaging techniques. The routine MRI (including T1Weighted, T2Weighted and contrast enhanced T1Weighted imaging) has been used to assess the extent of the disease; however, it remains difficult to accurately predict the nature of thyroid nodules using these structural images [5,6].
Diffusionweighted imaging (DWI) is a noninvasive method for evaluating Brownian motion of microscopic water diffusion in tissues in vivo, and this technique has been widely used in routine examinations. A variety of studies have reported that the apparent diffusion coefficient (ADC) values derived from diffusionweighted images can be used to discriminate between benign and malignant thyroid nodules [1,[7][8][9]. However, image distortions caused by susceptibility changes and ghost artifacts, for example, severely degrade the image quality of DWI in the neck region, especially poor at 3 Tesla (T) field strength.
For DWI, singleshot echo planar imaging is the most commonly used method due to its relative insensitivity to motioninduced phase errors, its high signaltonoise ratio (SNR) and short acquisition time [10]. However, EPI images are normally obscured by blurring along the phaseencoding direction because of T2* decay and are sensitive to offresonance effects [11], particularly in the neck region. Distortions in EPI (Δd(r)) are unavoidable and are proportional to the phase field of view (FOV PE ), the echo spacing (Δt PE ) and the local off-resonance (ΔB 0 ) [12]. Therefore, the distortions can be expressed as Δd(r) ~ FOV PE * Δt PE * ΔB 0 . Several techniques have been developed to reduce EPI distortions. For instance, parallel imaging and reduced FOV methods with twodimensional radio frequencyexcitation (2DRFexcitation) (e.g. the ZOOMit technique) have been used to reduce the FOV PE , and the readout-segmented EPI decreases the Δt PE [11,13,14]. Recently, a novel technique called integrated shimming echo planar imaging (iShimEPI), combining integrated slicespecific dynamic shimming and pixel wise unwrapping distortion correction, has been proposed to reduce distortions and signal voids caused by local B 0 inhomogeneity [15]. Previous studies [12,16] have demonstrated how iShimEPI outperformed 3DShim EPI on diffusion imaging through distortion reduction and lesion detection in the neck region.
In this study, we hypothesized that iShimEPI could be used for DWI in the thyroid gland, which is prone to severe artifacts in the neck region. To the best of our knowledge, no such study for the thyroid gland has been reported in the literature. Therefore, the goals of this study were to evaluate whether iShimEPI can improve the image quality of the thyroid gland DWI in comparison to 3DShimEPI, and to compare the ADC values, acquired from iShimEPI with those of 3DShimEPI.

RESULTS
All 21 patients included in this study underwent MRI scans. However, three patients were excluded from the quantitative analysis because of poor image quality (scores were = 1). The histologic assessment revealed 8 patients had papillary thyroid carcinoma, 2 patients had medullary thyroid carcinoma, 6 patients had nodular goiter and 2 patients had thyroid adenoma.

Qualitative comparisons of image quality
The analysis results of image quality for the iShim EPI and 3DShimEPI techniques are shown in Table 1. Figure 1 shows example images with iShimEPI and 3DShimEPI images from a patient with thyroid nodules. For both Observers 1 and 2, the average scores for iShim EPI were significantly higher than those with 3DShim EPI images for susceptibility artifacts (2.81 ± 0.37 vs. 1.93 ± 0.29, p < 0.001), ghost artifacts (2.95 ± 0.15 vs. 1.93 ± 0.29, p < 0.001), and the detectability of thyroid nodules (3.00 ± 0.00 vs. 2.55 ± 0.75, p = 0.008) (Figure 2), respectively.

Quantitative comparisons of ADC values
There was no significant difference between the iShimEPI and 3DShimEPI sequences for the ADC values of the thyroid nodules (1.607 ± 0.466 × 10 -3 mm 2 /s vs. 1.561 ± 0.483 × 10 -3 mm 2 /s, p = 0.184), as well as the contralateral normal thyroid gland (1.295 ± 0.340 × 10 -3 mm 2 /s vs. 1.279 ± 0.411 × 10 -3 mm 2 /s, p = 0.777). The reproducibility of the ADC measurements is summarized in Table 2. In addition, an excellent agreement was demonstrated for the ADC values of the thyroid nodules and contralateral normal tissues employing the iShimEPI and 3DShimEPI sequences, with ICCs ranging from 0.963 to 0.999. The BlandAltman plots also showed excellent reproducibility of the ADC measurements using both techniques for the evaluation of the thyroid nodules ( Figure 3) and contralateral normal thyroid gland ( Figure 4).

DISCUSSION
In this study, we demonstrated a much higher image quality of thyroid DWI using iShimEPI compared with the conventional 3DShimEPI technique. Image quality of iShimEPI with respect to susceptibility artifacts, ghost artifacts and the detectability of thyroid nodules was superior to that using the 3DShimEPI technique. However, there was no significant difference in the ADC values obtained using both techniques. In addition, an excellent reproducibility was demonstrated for the ADC measurements using the iShimEPI and 3DShimEPI techniques for evaluating thyroid nodules and normal tissue.
As one of the most commonly used functional MRI, DWI has emerged as an important method that can be noninvasively evaluate the diffusion of microscopic water molecules through tissues. DWI was initially used to evaluate brain disease, such as stroke, which offered better sensitivity and specificity for the evaluation of ischemia [17,18]. Furthermore, DWI has been extensively used characteristics of the thyroid nodule were observed on these images (white arrows). The iShimEPI DWI image had b = 500 s/mm 2 (B) and ADC map (E). 3DShimEPI DWI image with b = 500 s/mm 2 (C) and ADC map (F). Susceptibility (red arrows) and ghost artifacts are shown on the 3DShimEPI images, whereas improved image quality was observed using the iShimEPI technique. to detect and assess some diseases in many other organs in the past several years, including, for example, liver carcinoma, pancreatic cancer, renal carcinoma, prostate cancer, rectal cancer and cervical cancer [19][20][21][22][23][24][25]. DWI can also be used to differentiate benign tumors from malignant ones using ADC values. Of note, 3DShim EPI has conventionally been used to evaluate disease in clinical practice because of both its resistance to ghost artifacts and fast acquisition time. Several studies have been reported to use this technique for assessing thyroid lesions [7-9, 11, 26-29]. However, it is very challenging to perform conventional DWI in the neck region where the local field inhomogeneity cannot be completely compensated by shimming and therefore can lead to signal voids and image distortions. In this study, the 3DShim EPI and iShimEPI techniques were used on thyroid nodule DWI and the results compared. The iShimEPI technique uses a slicespecific dynamic shimming scheme rather than 3D volume shimming of the entire stack of slices. This appears to remarkably improve the magnetic field homogeneity of each slice. Additionally, the pixel vise unwrapping distortion correction method was used to remove any residual distortion due to the second or higher order field inhomogeneity. Thus, artifacts were significantly reduced in the iShimEPI images. To our knowledge, the application of iShimEPI in the evaluation of thyroid nodules has thus far not been explored.
In our study, the image quality scores of the iShim EPI images were found to be higher than those of the 3DShimEPI images in respect to susceptibility artifacts, ghost artifacts and the detectability of thyroid nodules. The image quality scores benefitted from the slicespecific dynamic shimming scheme combined with the pixelvise unwrapping distortion method. Gatidis et al. showed that the utility of iShimEPI for DWI contributes to improved image quality in the neck region at 3 T compared with other EPIbased techniques (3DShimEPI, Zoomed EPI and readoutsegmented EPI) [12]. Zhang et al. demonstrated that iShimEPI can improve lesion detection in wholebody DWI at 3 T [16]. Similar results were consistently demonstrated in our study.
The ADC values of the thyroid nodules and contralateral normal thyroid gland were evaluated quantitatively for both techniques. No significant differences in the ADC measurements using the 3DShim EPI and iShimEPI techniques were observed (both had p values > 0.05). However, the average ADC values of the thyroid nodules and contralateral normal thyroid gland on iShimEPI were higher than those on 3DShimEPI. The reason may be due to better shimming using iShimEPI compared with 3DShimEPI, thus less susceptibility and distortion were observed on the iShimEPI images, which would slightly vary the ADC values. Additionally, we used the shortest repetition time/echo time (TR/TE) for both sequences respectively, while kept the other parameters consistent. Several studies have shown that ADC values depend on various imaging parameters, such as magnetic field strength, TR, TE, the b value and gradient mode, which might explain why ADC values are slightly different between the two DWI sequences [30][31][32]. We also assessed the reproducibility of ADC measurements on lesions and normal tissues. Figures 3 and 4 show the excellent reproducibility of the ADC values using Bland Altman plots, and narrow intervals of agreement compared with the mean were observed for these measurements. Additionally, we also noted an excellent agreement of these measurements using the ICC values. The consistency of ADC values in both techniques and improved image quality revealed that we may use the new iShimEPI technique to replace conventional 3DShimEPI technique for routine thyroid DWI examination.
The present study has several limitations. First, the patient population was relatively small. A larger patient population is needed to verify our findings and for further study. Second, iShimEPI can decrease but not thoroughly eliminate susceptibility and distortion artifacts. Therefore, a more advanced technique within the sphere of MRI technology is needed to overcome these artifacts. Third, techniques. ** P < 0.005. www.oncotarget.com iShimEPI and 3DShim EPI techniques were performed with their shortest TR/TE respectively. As discussed above, different TR/TE might influence the calculation of the ADC value, however, no significant differences were observed in ADC between both techniques in the present study. But further study with strictly identical parameters is conducted. Finally, the benign and malignant properties of thyroid nodules were not differentiated by the ADC   values. Since this was not the main purpose of the present study, we will evaluate this feature with a larger sample size in later studies.
In conclusion, iShimEPI can decrease reduce susceptibility and ghost artifacts, improve the overall image quality and the detectability of thyroid nodules compared with conventional 3DShimEPI. Notably, iShimEPI DWI may serve in clinical practice as a promising technique for decreasing artifact and assessing thyroid disease.

Subjects
Between October 2015 and July 2016, 21 patients (age, 22-84 years; male/female, 4/17) with thyroid nodules were consecutively enrolled in this prospective study. The inclusion criterion was patients who were found to have thyroid nodules diagnosed by ultrasound. Thyroid nodules of diameter less than 5 millimeters, with ill defined borders and of poor image quality were excluded from the study [29]. This study was approved by the local institutional review board, and written informed consent was obtained from each patient.

Image acquisition
Magnetic resonance imaging was performed on a 3 T whole body MRI system (MAGNETOM Skyra, Siemens Healthcare, Erlangen, Germany) using a 20channel phasedarray head/neck coil. The thyroid gland examination protocols included an axial T1weighted turbo spinecho (T1W TSE), coronal T2weighted TSE and axial DWI sequences using both a prototype iShim EPI and 3DShimEPI techniques with a acquisition time of 1minute 17 seconds and 1minute 16 seconds, respectively. T1W and T2W images were used to locate thyroid nodules and position the DWI sequences. The parameters of these protocols are summarized in Table 3.
The total scan time of each subject was approximately seven minutes.

Image analysis
All morphological images and DWI images were transferred to a workstation (syngoMMWP, Siemens Healthcare, Germany) for analysis. iShimEPI and 3DShimEPI DWI images were evaluated by two independent observers (Observer 1 and Observer 2 with four and six years of experience in thyroid imaging, respectively) who were blind to the subjects' clinical information, with T1W and T2W images serving as a reference to identify susceptibility artifacts, ghost artifacts (1, marked; 2, median; 3, no artifact), and the detectability of thyroid nodules (1, poor; 2, median; 3, good) using threescale scores. Freehand circular regions of interest (ROIs) were drawn in the nodule and normal regions of the thyroid gland on ADC maps by the first Observer, cautiously avoiding vessels, and areas of necrosis and hemorrhage. The mean value of the ADC was recorded for each ROI. To test the intraobserver variability, the observer measured the two sets of ADC images twice with an intervening 4week period to avoid any recall bias.

Statistical analysis
All the analyses were performed using SPSS (version 16.0, SPSS Inc., Chicago, IL, USA) and MedCalc (version 13.0.0.0, MedCalc Softaware, Mariakerke, Belgium) softwares for Microsoft Windows. Quantitative data were described as mean ± standard deviation. Significant differences in image quality scores, and ADC values between iShimEPI and 3DShimEPI were assessed using Wilcoxon signedrank tests. The intra observer variability was evaluated using the intraclass correlation coefficient (ICC) and BlandAltman plots for ADC values [33]. ICC values < 0.4, 0.4 -0.75 and > 0.75 were regarded as representing poor agreement, good agreement and excellent agreement, respectively [34].