Effectiveness of diffusion-weighted magnetic resonance imaging (DW-MRI) in the differentiation of thyroid nodules

Abstract

Background

The aim was to investigate which of two different b values (b 500 s/mm² and b 800 s/mm²) are more effective in the differentiation of benign-malignant nodules using Diffusion-Weighted Magnetic Resonance Imaging (DW-MRI).

Materials and methods

Patients presenting with a preoperative diagnosis of nodular goiter or multinodular goiter were included in this study. These patients underwent neck MRI examinations, and their cases were analyzed retrospectively. A total of 26 patients were included in the study. A total of 46 nodules meeting the study criteria were examined. Measurements were performed on Apparent Diffusion Coefficient (ADC) maps of patients at two different b values (b 500 s/mm² and b 800 s/mm²), and the results were compared with histopathological findings.

Results

Out of a total of 46 nodules, 37 were identified as benign, and 9 as malignant based on histopathological analysis. The mean ADC value at b 500 was lower in malignant nodules (1259.65 ± 328.13) compared to benign nodules (19037.48 ± 472.74). Similarly, the mean ADC value at b 800 was lower in malignant nodules (1081.72 ± 200.23) compared to benign nodules (1610.44 ± 418.06). When a cut-off value of 1.1 × 10^− 3 was accepted for the differentiation of pathology, the sensitivity for distinguishing pathology with ADC values at b 500 was 83.3%, with a specificity of 90.0%, and for ADC values at b 800, the sensitivity was 71.4%, with a specificity of 89.7%.

Conclusion

DW-MRI without the need for contrast agent administration is a useful method in the differentiation of benign-malignant thyroid nodules.

Introduction

Thyroid nodules are quite common in the population, detected by palpation in 4–8% of adult patients and by ultrasound examination in 10–41% of cases. Autopsy reports reveal a prevalence of around 50% [1]. Current guidelines suggest that 7–15% of thyroid nodules may be malignant [2]. Differentiating between benign and malignant nodules is crucial to prevent inadequate follow-up and treatment in patients at high risk of thyroid malignancy, and to avoid unnecessary treatment and surgery in low-risk patients [3].

In the diagnosis of thyroid nodules, a variety of methods are employed, including patient history, physical examination findings, laboratory tests, thyroid scintigraphy, Ultrasonography (US), and Fine-Needle Aspiration Biopsy (FNAB) [1,2,3,4]. Despite these methods, US alone cannot definitively categorize a nodule as benign or malignant [1,2,3,4]. A definitive diagnosis often requires histopathological evaluation, but FNAB’s effectiveness can be limited and sometimes yields inconclusive results [4, 5]. This limitation underscores the importance of the physician’s skill and the cytopathologist’s experience in FNAB procedures [5]. Therefore, it is crucial to utilize other imaging modalities to detect or at least suspect thyroid malignancies noninvasively.

To address these challenges, recent advancements in DW-MRI have shown promise. Specifically, DW-MRI has been effectively used across various body parts, including the head and neck region, to enhance diagnostic accuracy [6,7,8]. This technique, which does not require the administration of a contrast agent, offers the advantage of superior soft tissue contrast resolution and the ability to perform multiplanar imaging. It is based on the detection of water molecules’ physiological motion at a microscopic level, with the ADC value providing critical diagnostic information [9]. Despite its benefits, the search for a universally accepted cut-off value or optimal b value for thyroid nodule evaluation continues [10, 11]. In our study, we explore the efficacy of using two distinct b values in diffusion-weighted imaging to differentiate between benign and malignant thyroid nodules.

Materials and methods

Procedure

Before the study commenced, ethical approval was obtained from the Ethics Committee of Celal Bayar University Faculty of Medicine. All participants underwent neck MRI and diffusion-weighted MRI examinations using a 1.5 Tesla MR scanner (Signa HDx, General Electric, Milwaukee, Wisconsin, USA) equipped with a head-neck coil. The MRI protocol included a variety of sequences designed to provide comprehensive diagnostic information:

• Sagittal T2-weighted fast spin echo (TR: 3160ms, TE: 102ms, FOV: 27 cm, matrix: 320 × 256, slice thickness: 6 mm, slice gap: 1 mm),

• Axial T2-weighted fast spin echo (TR: 3280ms, TE: 85ms, FOV: 24 cm, matrix: 320 × 256, slice thickness: 6 mm, slice gap: 1 mm),

• Axial T1-weighted fast spin echo (TR: 400ms, TE: 89.3ms, FOV: 24 cm, matrix: 320 × 256, slice thickness: 6 mm, slice gap: 1.5 mm),

• Diffusion-weighted imaging echo-planar imaging sequences with b-values of 500 s/mm² and 800 s/mm² (TR: 5000ms, TE: 89.3ms, FOV: 24 cm, matrix: 92 × 128, slice thickness: 6 mm).

ADC maps were automatically generated by the 1.5 Tesla MR scanner, facilitating the evaluation of all nodules across T1-weighted, T2-weighted, diffusion-weighted, and ADC maps. Efforts were made to identify and exclude cystic portions of the nodules from the Regions of Interest (ROI) area during T2-weighted and T1-weighted imaging to enhance measurement accuracy. Measurements were conducted on ADC maps at b 500 and b 800 values for each nodule, repeated three times to calculate an average value. Additionally, diffusion-weighted imaging was utilized to assess normal-looking parenchymal areas without nodules at two different b values (b = 500, 800 s/mm²), providing a baseline for comparison. The data obtained were then compared with the patients’ postoperative histopathological results, establishing a correlation between imaging findings and actual histopathology.

Sample

A total of 26 patients, comprising 17 females (65.4%) and 9 males (34.6%), were included in the study. The patients’ ages ranged from 20 to 77 years, with an average age of 48.62 ± 12.12 years.

Method

Between January 6, 2010, and June 5, 2011, 49 patients with a preoperative diagnosis of nodular goiter or multinodular goiter underwent neck MRI examinations and were referred to the Radiology department for imaging, identifying a total of 69 nodules. However, exclusions were made based on specific criteria: nodules smaller than 1 cm in 10 patients, thyroiditis diagnosis in 6 patients, non-evaluable diffusion images due to motion artifacts in 5 patients, a cystic nodule in 1 patient, and a hemorrhagic nodule pathology in another patient. Following these exclusions, the study ultimately included 46 nodules from 26 patients. The distribution of nodules among these patients was as follows: 18 patients had 1 nodule each, 3 patients had 2 nodules each, 2 patients had 3 nodules each, 2 patients had 5 nodules each, and 1 patient had 6 nodules.

Data analysis

The data from the study were analyzed using Statistical Package for the Social Sciences (SPSS) version 23.0. Descriptive statistical methods, including frequency, minimum, maximum, mean, median, and Odds Ratio (OR), were employed to analyze the collected data. To compare groups, categorical variables were examined using the Chi-square test and independent sample t test was utilized for comparing means. Furthermore, ROC curve analysis was performed to identify the optimal cut-off value. The results were considered statistically significant at a p-value of less than 0.05, with a confidence interval set at 95%.

Findings

The postoperative histopathology results revealed that 19.6% (N = 9) of the nodules were malignant, while 80.4% (N = 37) were benign”.

Of the malignant nodules, 8 were identified as papillary carcinoma, and 1 was classified as anaplastic thyroid cancer. The distribution of lesions according to histopathological results is presented in Table 1. The average size of the nodules was determined to be 21.78 ± 13.87 mm, ranging from 10.00 to 70.00 mm. The average size of malignant nodules was 19.22 ± 13.19, while that of benign nodules was 22.41 ± 14.14, with no significant difference in size between benign and malignant nodules (p > 0.05). Malignant nodules had a significantly lower mean b 500 ADC value of 1.26 × 10^-3 mm²/s compared to the benign nodules’ 1.91 × 10^-3 mm²/s (p < 0.001) (Fig. 1). Similarly, malignant nodules had a significantly lower mean b 800 ADC value of 1.08 × 10^-3 mm²/s compared to the benign nodules 1.61 × 10^-3 mm²/s (p < 0.001) (Fig. 2). The b 500 and b 800 ADC values of benign and malignant nodules are shown in Table 2. The mean parenchymal ADC values were found to be 1.57 × 10^-3 for b 500 ADC and 1.39 × 10^-3 for b 800 ADC in benign nodules. In malignant nodules, the mean b 500 ADC was 1.49 × 10^-3 and the mean b 800 ADC was 1.23 × 10^-3. There was no significant difference in parenchymal ADC values between benign and malignant nodules (p > 0.05). In a notable case, despite the histopathology indicating papillary carcinoma, the measured b 500 and b 800 ADC values of 1.9 × 10^-3 mm²/s and 1.26 × 10^-3 mm²/s, respectively, aligned with those typically seen in benign nodules. In another patient, although the histopathology indicating benign nodule, the measured b 500 and b 800 ADC values of 0.94 × 10^-3 mm²/s and 0.93 × 10^-3 mm²/s, respectively, aligned with those typically seen in malignant nodules. The mean ADC values for benign nodules, malignant nodules, and normal parenchyma were compared, and the results are presented in Figs. 3 and 4. Consequently, ROC curve analysis was performed to assess the efficacy of b 500 and b 800 ADC values in distinguishing malignancy, aiming to establish an optimal cut-off value (Figs. 5 and 6). The ROC area (AUC) of the ADC500 values was found to be 0.880 ± 0.065 (95% CI: 0.753–1.006). The prediction point was determined as 1.1 × 10^-3. At this cutoff value, the sensitivity of ADC500 is 83.3%, specificity is 90.0%, the positive predictive value (PPV) is 55.6%, and the negative predictive value (NPV) is 97.3%. The ROC area (AUC) of the ADC800 values was found to be 0.916 ± 0.041 (95% CI: 0.835–0.996). The prediction point was determined as 1.1 × 10^-3. At this cutoff value, the sensitivity of ADC800 is 71.4%, specificity is 89.7%, the positive predictive value (PPV) is 55.6%, and the negative predictive value (NPV) is 94.6%. These ADC500 and ADC800 values indicate that at the optimal cutoff of 1.1 × 10^-3, ADC500 and ADC800 demonstrates a high specificity, making it a reliable marker for diagnostic purposes. The ROC analysis’s thus supports the robustness of ADC500 as a diagnostic tool, with the optimal cutoff point providing a good balance between sensitivity and specificity. The b 500 value appears to be more advantageous than the b 800, given its higher sensitivity and specificity values.

Fig. 1

In a 77-year-old female patient diagnosed with anaplastic tumor, (A) Axial T2A image shows a solid nodule with hyperintense capsular invasion and vascular invasion in the left lobe, compressing the trachea. Pleural fluid is also observed in the left hemithorax in the images. (B) In the b 500 value ADC map, the ADC value of the nodule was measured as 1.0 × 10^− 3, (C) in the b 800 value ADC map, the ADC value of the nodule was measured as 1.0 × 10^− 3

Fig. 2

In a 47-year-old female patient with a benign nodule (A), a nodule is observed in the left lobe of the thyroid gland, compressing the trachea. (B) On the ADC map, the nodule was hyperintense and the ADC value of the nodule was measured as 1.8 × 10^− 3 at b 500 value. (C) The ADC value of the nodule was measured as 1.3 × 10^− 3 on the b 800 value ADC map

Table 1 Pathology of thyroid nodule

Table 2 Comparison between the ADC values of benign and malignant thyroid nodules

Fig. 3

Boxplot of ADC values in benign nodules, malignant nodules, and normal thyroid parenchyma

Fig. 4

Boxplot of ADC values in benign nodules, malignant nodules, and normal thyroid parenchyma

Fig. 5

ROC Curve for ADC500 value in malignant-benign nodule analysis. The ROC area of b 500 ADC values was found to be 0.880 ± 0.065 (0.753–1.006). The prediction point was determined as 1.1 × 10^-3. At this value, the sensitivity of b 500 ADC value is 83.3%; specificity 90.0%; The positive predictive value was 55.6% and the negative predictive value was 97.3%

Fig. 6

ROC curve for ADC800 value in malignant-benign nodule analysis. The ROC area of b 800 ADC values was found to be 0.916 ± 0.041 (0.835–0.996). The prediction point was determined as 1.1 × 10^-3. At this value, the sensitivity of the b 800 ADC value is 71.4%; specificity 89.7%; The positive predictive value was 55.6% and the negative predictive value was 94.6%

Discussion

In our research, we evaluated the efficacy of DW-MRI in differentiating between benign and malignant thyroid nodules. The results demonstrated that the ADC values at both b 500 and b 800 were significantly lower in malignant nodules compared to benign ones. Interestingly, no significant difference was observed in ADC measurements between these two b values in the normal thyroid parenchyma.

Except for one study [12], all other research has reported lower ADC values for malignant nodules when compared to benign ones, a finding consistent with studies involving various tissues in the literature [13,14,15,16,17,18]. Generally, malignant tumors exhibit hypercellularity, which leads to a reduction in the extracellular matrix and, consequently, a decrease in ADC values [6,7,8]. In the context of the thyroid gland, this characteristic correlates with the presence of large, oval, irregularly nucleated malignant thyroid nodules, accompanied by multiple micronucleoli, a thin chromatin structure, and intranuclear pseudoinclusions [13, 14]. Similarly, our study identified lower ADC values for malignant nodules in comparison to benign ones. Razek et al. observed a broad range of ADC values for benign nodules, attributing this variability to diverse components within the nodules, such as colloid, microcystic necrosis, hemorrhage, fibrous tissue, and calcification [13]. In our research, by aiming to minimize the measurement of cystic components within nodules, we excluded cystic and hemorrhagic nodules from our analysis, thereby achieving a narrower range of ADC values.

In the existing literature, ADC measurements have been conducted using a variety of b values [12,13,14,15, 17, 18]. Schuller-Weiderkamm et al. utilized a b value of 800, whereas Razek et al. employed b values of 250 and 500; Bozgeyik et al. conducted measurements with b values of 100, 200, and 300; Erdem et al. utilized a b value of 1000; Aghaghazvini et al. implemented b values of 50, 500, and 1000; and Zhu et al. explored b values ranging from 0 to 1000 and from 0 to 2000 [12,13,14,15, 17, 18]. The selection of the b value is a critical factor influencing both image quality and the accuracy of ADC measurements. It has been observed that ADC values obtained at lower b values may be affected by the microcapillary perfusion of water molecules in the blood, leading to artificially higher ADC readings [14]. Conversely, higher b values mitigate the influence of perfusion on ADC measurements but can diminish the Signal-to-Noise Ratio (SNR), potentially introducing errors into the measurements [8,9,10]. In previous studies, Aghaghazvini et al. stated in their comparison that the b 1000 value had higher sensitivity and specificity, while Bozgeyik et al. stated that the b 300 value had higher sensitivity and specificity [11,12,13,14,15]. In a study conducted using high b values [14], similar diagnostic performance was found between b 1000 and b 2000 values. Consequently, the diverse range of b values employed across studies has precluded the establishment of an optimal cutoff ADC value. A comparative analysis of the sensitivity and specificity values for ADC500 and ADC800 reveals that ADC500 exhibits superior performance metrics. Specifically, ADC500 demonstrates a higher sensitivity (83.3% compared to 71.4%) and a slightly elevated specificity (90.0% compared to 89.7%) at the identical cutoff point of 1.1 × 10^− 3.

The enhanced sensitivity and specificity values of ADC500 indicate its greater utility in diagnostic applications. Consequently, the ROC analysis supports the robustness of both ADC500 and ADC800 as diagnostic tools. However, ADC500, with its higher sensitivity and specificity values, seems to offer a more reliable diagnostic performance compared to ADC800. Therefore, ADC500 is be considered the more useful marker for diagnostic applications.

In recent years, studies employing magnetic field strengths and devices with varied b values have consistently reported lower ADC values in malignant nodules [17, 18]. However, the consistency of these findings has been called into question due to variations in magnetic field strengths, b values, and DW-MRI parameters. As a result, establishing a universally applicable cutoff ADC value presents significant challenges [10, 11].

In our evaluation, two nodules demonstrated discrepancies between their histopathological results and DW-MRI values. For instance, a benign nodule exhibited ADC values akin to those typically seen in malignant nodules, recording b values of 500 and 800 at 0.94 × 10^− 3 mm²/s and 0.93 × 10^− 3 mm²/s, respectively. Upon pathological examination, extensive fibrotic and calcified areas were identified within the nodule. These findings suggest that the reduced extracellular fluid and the consequent restriction in diffusion could have influenced the ADC measurements. Additionally, the presence of calcifications may lead to inaccuracies in ADC value calculations. Hence, a pre-MRI evaluation of nodules for calcification, fibrosis, and other internal characteristics could mitigate the risk of misinterpretation.

In a separate case involving a malignant nodule, the ADC values were notably higher, with measurements of 1.9 × 10^− 3 mm²/s for b 500 and 1.26 × 10^− 3 mm²/s for b 800. Pathological analysis confirmed the diagnosis of a macrofollicular variant of papillary carcinoma. This variant is characterized by an increased colloid volume within the follicles, which elevates the extracellular water content, thus enhancing water mobility and diffusion capacity, and resulting in higher ADC values. This observation is consistent with the interpretations of Weidekamm et al. regarding malignant nodules [12]. However, the rarity of this variant in our study, with only a single patient represented, precluded statistical analysis. Consequently, further research involving a larger cohort of patients is essential to deepen our understanding of the variability in ADC values among malignant nodules.

In our study, measurements obtained from normal-looking parenchymal areas devoid of nodules did not show any significant differences across the two b values. This finding is in line with previous research, which has also reported no discernible differences in ADC values between the parenchyma of healthy individuals and those with nodules, suggesting that the pathology does not impact the normal parenchyma.

Our research methodology involved comparing ADC values against patients’ postoperative histopathological findings, providing a direct correlation with the definitive diagnostic outcome [14]. Contrastingly, some studies have opted to compare ADC values with the results from FNAB [12, 14, 15]. However, the reliability of FNAB can sometimes be compromised due to reports of “inadequate” sample collection. Furthermore, FNAB alone is insufficient for the evaluation of follicular neoplasms, as it cannot definitively distinguish between follicular adenoma and follicular carcinoma. Accurate diagnosis in such cases necessitates a thorough examination of the entire nodule, with specific attention to vascular and capsular invasion, to determine the nature of the neoplasm. This highlights the limitations of FNAB and underscores the importance of comprehensive histopathological examination for accurate diagnosis [5,6,7,8,9,10,11,12,13,14,15,16,17,18,19].

Our study acknowledges several limitations. Primarily, the resolution offered by MRI technology falls short when it comes to nodules smaller than one centimeter in size. Due to this limitation, such nodules cannot be distinctly differentiated from the surrounding parenchyma, rendering ADC measurement unfeasible. Furthermore, despite efforts to exclude cystic portions from the ROI, the potential inclusion of cystic areas within small nodules might have inadvertently resulted in elevated ADC values.

Another significant constraint is the limited number and diversity of malignant nodules examined in our study. This limitation restricts the generalizability and applicability of our findings across the broader spectrum of thyroid nodules with varying etiologies.

To address these challenges and enhance the robustness of future research, studies designed with suitable parameters that encompass a more extensive assortment of nodules, inclusive of diverse thyroid nodule etiologies, are imperative. Such studies will not only contribute to overcoming the limitations identified but also aid in refining the diagnostic accuracy and reliability of ADC measurements for thyroid nodules.

Conclusion

Our study identified significant findings consistent with prior research regarding the utility of DW-MRI in assessing thyroid nodules larger than 10 mm. The efficacy of DW-MRI as a diagnostic tool lies in its ability to differentiate between benign and malignant thyroid nodules. This effectiveness stems from several key advantages: DW-MRI is non-invasive, provides rapid results, and eliminates the necessity for contrast agents. These attributes not only enhance patient comfort but also reduce the risk of potential contrast-related complications. As we advance, further research involving a greater number of nodules and the optimization of b values is imperative. Such studies will refine our understanding and potentially position DW-MRI as a viable alternative to traditional methods in the evaluation of thyroid nodules, offering a promising avenue for both clinicians and patients in the management of thyroid conditions.

Declarations.