Main points

  • The multivariate prediction model demonstrated satisfactory diagnostic performance with an AUC of 0.946 (0.896–0.996).

  • Combined DWISIR and ADCmin thresholds model demonstrated a high specificity with an unnecessary biopsy rate of 0%.

  • The multivariate prediction model and the combined threshold models are better than the ACR-TIRADS.

Introduction

Thyroid nodules manifest in up to 50–60% of the general population as detected by high-resolution ultrasound, while only around 10% of these nodules are malignant [1,2,3]. Since 2014, the overall incidence of thyroid cancer incidence rate has decreased, but the incidence and mortality rate of tumors larger than 4 cm are still rising [4]. Thyroid nodules larger than 4 cm are important for surgical decision-making in adult patients with thyroid nodules, according to the 2017 Thyroid Cancer Staging Manual of the American Joint Committee on Cancer (AJCC) [5]. Ultrasonography (US) is currently the main imaging technique for evaluating thyroid nodules [6,7,8,9]. US characteristics of thyroid nodules, such as the American College of Radiology Thyroid Imaging Reporting and Data System (ACR-TIRADS), have been employed for risk stratification [10]. However, the interobserver agreement on the TI-RADS remains only fair to moderate [11, 12] and these methods mainly focus on thyroid nodules smaller than 4 cm. Although biopsy is regarded as the gold standard for the preoperative diagnosis of thyroid cancer, it has reduced sensitivity when applied to thyroid nodules larger than 4 cm [13,14,15].

Diffusion-weighted imaging (DWI), a non-contrast magnetic resonance imaging (MRI) technology, detects water molecule random mobility and offers information on tissue microstructure and cell density. The apparent diffusion coefficient (ADC) map can be further calculated from DWI to quantify the diffusion characteristic of tissues [16, 17]. DWI was used to diagnose benign and malignant tumors [18,19,20,21]. Although earlier research has demonstrated its effectiveness of DWI in distinguishing between benign and malignant thyroid nodules, those studies have included nodules of all sizes [22, 23]. For larger lesions, DWI has a superior diagnostic value, but there were limited researcher on the diagnostic value of DWI for thyroid nodules larger than 4 cm. Meanwhile, accurate pre-operative assessment of thyroid nodules is crucial for subsequent treatment. Accordingly, it is important to distinguish between benign and malignant nodules larger than 4 cm before surgery [24].

Consequently, this study aimed to evaluate approaches using quantitative DWI, and compare them with ACR-TIRADS to differentiate between benign and malignant thyroid nodules larger than 4 cm preoperatively.

Materials and methods

Patients and study design

The study followed the Declaration of Helsinki (revised 2013). The Institutional Ethics Committee of Minhang Hospital affiliated with Fudan University approved this observational, retrospective study (approval number: 2021–008-01 K) with a waiver of informed consent.

We reviewed consecutive patients with thyroid nodules who had pathology results at our institution between 2017 and 2022. The inclusion criteria included: 1) lesion diameter larger than 4 cm; 2) patients who underwent preoperative thyroid MRI; 3) complete pathology of postoperative specimens. The exclusion criteria included: 1) incomplete clinical and imaging data; 2) poor image quality; 3) lack of contrast enhancement on MRI. Figure 1 displays the study flowchart.

Fig. 1
figure 1

The study flowchart. Abbreviations: MRI, Magnetic resonance imaging; DWI, diffusion weighted imaging; ADC, apparent diffusion coefficient; SIR, signal intensity rate; ADCmin, minimum apparent diffusion coefficient value; ACR-TIRADS, American College of Radiology Thyroid Imaging Reporting and Data System; TR, TI-RADS; AUC, area under the receiver operating characteristic curve

Full size image

Finally, 82 lesions from 78 patients (32 males and 50 females; age: 50.26 ± 16.20 years; age range: 15–78 years) met the inclusion criteria. Lesions were categorized into benign (n = 62) and malignant (n = 20) groups according to the postoperative pathology.

MRI acquisition

The 1.5 T MRI scanner (Excite HD; GE Healthcare, Waukesha, WI, USA) used for all MRI examinations was set up with an 8-channel customized neck surface coil (Chenguang Medical Technology Ltd, Shanghai, China). The scan covered the thoracic inlet to the base of the cranium were covered by the scan. The MRI sequences used (CE-T1WI) included axial and coronal fat-suppressed T2-weighted imaging (T2WI), axial T1-weighted imaging (T1WI), single-shot spin-echo echo-planar imaging (SS-SE-EPI) DWI at b values of 0 and 800 s/mm2, and axial multiphasic contrast-enhanced T1WI comprised the MRI sequences used (CE-T1WI). A gadolinium contrast agent (Magnevist; Bayer Healthcare, Berlin, Germany) was injected for the CE-T1WI acquisition at a dose of 0.2 mL/kg and a rate of 3 mL/s, followed immediately by 20 mL of physiological saline flushing. Following the injection of the contrast agent, six phases were recorded at intervals of 30, 60, 120, 180, 240, and 300 s intervals while the patients were asked to hold their breath. Table S1 lists detailed acquisition parameters.

Image analysis

ADC maps were automatically created from DWI images (b = 0 and 800 s/mm2) on the console using mono-exponential fitting. Quantitative DWI parameters were measured by two MRI diagnosticians who were blind to the lesion pathology (a chief physician with eight years of experience and a resident with one year each in thyroid MRI diagnosis), using picture archiving and communication system (PACS) and Advantage Workstation 4.5 (GE Healthcare, Waukesha, WI, USA). The section of the whole solid leision portion of the lesion with maximum transverse diameter (excluding cystic, hemorrhage, necrosis, calcium, and vascular structures) was selected to delineate the first region of interest (ROI 1). The following quantitative features in the ROI 1 were measured: 1) mean DWI signal intensity (DWISI); 2) mean ADC signal intensity (ADCSI); 3) mean ADC value (ADCmean) and minimum ADC value (ADCmin). Another ROI with an 8–10 mm2 area is also outlined as a relatively homogeneous solid part without cystic, hemorrhage, necrosis, calcium and vascular structures in the lesion and contralateral to the normal thyroid tissue. The following quantitative features of the ROI 2 were measured in the lesion: 1) DWI signal intensity standard deviation (DWISD) and ADC signal intensity standard deviation (ADCSD) and ADC value standard deviation (ADCVSD); 2) mean DWI signal intensity and mean ADC signal intensity of contralateral normal thyroid tissue (DWINSI and ADCNSI). The following formulas, DWISIR = DWISI / DWINSI and ADCSIR = ADCSI / ADCNSI, were used to calculate the DWI signal intensity rate (DWISIR) and ADC signal intensity rate (ADCSIR). DWI images (b = 800 s/mm2) and ADC map generated from DWI images (b = 0 and 800 s/mm2) were used for quantitative parameters extraction. All measurements were performed twice and averaged. Figure 2a demonstrates representative images of ROI delineation.

Fig. 2
figure 2

Representative DWI quantitative measurements. Lesions as indicated by blue arrows. a shows an example diagram of the ROI sketch. The black line outlines ROI 1, which is the whole solid portion of the slice with maximum transverse diameter for the lesion. The red line outlines ROI 2 of 8-10 mm2, the red dashed line is the interior of the lesion, and the red solid line is the contralateral-normal thyroid tissue. b and c show a patient with thyroid adenoma. b shows a DWI image with DWISD and DWISIR of 22 and 1.07. c shows ADC images with ADCSD, ADCSIR, ADCmin, ADCmean and ADCVSD of 110, 0.97, 1.15, 152 and 171. d and e show a patient with papillary thyroid cancer. d is a DWI image with DWISD and DWSIR of 24.7 and 1.56. e is an ADC image with ADCSD, ADCSIR, ADCmin, ADCmean and ADCVSD of 83, 0.74, 0.65, respectively, 1.04 and 69.59, respectively. Abbreviations: DWI, diffusion weighted imaging; ADC, apparent diffusion coefficient; SI, signal intensity; SD, standard deviation; ADCmin, minimum apparent diffusion coefficient value; ADCmean, mean apparent diffusion coefficient value; ADCVSD, standard deviation of apparent diffusion coefficient value

Full size image

ACR-TIRADS

Two US experts retrospectively reviewed US images of thyroid lesions, reaching a consensus without knowledge of the lesion pathology. All lesions with category ≥ 4 were considered malignant according to ACR-TIRADS.

Statistical analysis

SPSS statistical software (version 26.0; IBM Corp, Armonk, NY, USA) and MedCalc (version 20.100; MedCalc Software, Ostend, Belgium) were used for all statistical analyses, and P values < 0.05 were deemed statistically significant. Quantitative MRI parameters and categorical variables of the malignant and benign groups were compared using independent t-tests, Chi-square tests, and Fisher's exact tests, respectively. Interobserver agreement was assessed using the intraclass correlation coefficient (ICC).

The malignancy prediction model was built using independent factors that were found using univariate and multivariate logistic stepwise regression. By optimizing the Youden's index, receiver operating characteristic (ROC) curve analysis was used to determine the ideal threshold values for the pertinent parameters. Combined thresholds approaches were established based on malignancy-related parameters (Supplementary Method). Individual parameters and models were evaluated using ROC curves, with the area under the ROC curve (AUC) compared by the DeLong test. Unnecessary biopsy rate was defined as the percentage of benign lesions for those requiring biopsy. The diagnostic performance measures for each model, including as accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and unnecessary biopsy rate were calculated, comparing the ACR-TIRADS result was compared.

Results

Clinicopathological characteristics

Table 1 lists the clinicopathological features of thyroid nodules. Except for the location (P = 0.015), there was no difference in the distribution of other features in the benign and malignant thyroid nodules. Table S2 demonstrates the pathological types of thyroid nodules.

Table 1 Clinicopathologic characteristics
Full size table

Diagnostic performance of quantitative parameters

Figure 2 shows representative DWI images and the ROI delineation. Moreover, Table 2 demonstrates the results of the univariate and multivariate logistic regression analyses in predicting malignant thyroid nodules. Malignant nodules displayed significantly greater DWISD (P = 0.002) and DWISIR (P = 0.007) than benign nodules. Additionally, malignant nodules had significantly lower ADCSD (P = 0.005), ADCSIR (P = 0.008), ADCmin (P < 0.001), and ADCmean (P < 0.001) than benign nodules. The ICCs of DWISD, DWISIR, ADCSD, ADCSIR, ADCmin and ADCmean were 0.776, 0.758, 0.720, 0.923, 0.789, 0.783 and 0.743, respectively. ADCmin was the best-performing parameter with an AUC of 0.933 (0.874—0.992). Figure 3a and Table 3 represent the ROC curves and diagnostic performance metrics at the optimal threshold of relevant individual parameters, respectively.

Table 2 Comparisons of quantitative DWI parameters to identify malignant thyroid nodules
Full size table
Fig. 3
figure 3

The ROC curves. a is ROC curves of meaningful single parameters. b is multivariate prediction model, combined thresholds models and ACR-TIRADS for malignant and benign thyroid nodules. Abbreviations: DWI, diffusion weighted imaging; ADC, apparent diffusion coefficient; SD, standard deviation; SIR, signal intensity rate; ADCmin, minimum value of apparent diffusion coefficient; ADCmean, mean apparent diffusion coefficient value; ADCVSD, standard deviation of apparent diffusion coefficient value

Full size image
Table 3 The diagnostic performance of meaningful single DWI parameters to identify malignant thyroid nodules
Full size table

The optimal threshold values were 1.13 × 10–3 mm2/s for ADCmin, 1.25 for ADCSIR, and 1.20 for DWISIR, showing their distribution in Fig. 4. For benign and malignant nodules, DWISIR, ADCSIR, and ADCmin overlapped; however, malignant nodule ADCmin was comparatively low.

Fig. 4
figure 4

The Violin chart of DWISIR, ADCSIR and ADCmin. Abbreviations: DWI, diffusion weighted imaging; ADC, apparent diffusion coefficient; SIR, signal intensity rate; ADCmin, minimum apparent diffusion coefficient value

Full size image

Comparison of the diagnostic performance with ACR-TIRADS

Table 4 and Fig. 3b depict the diagnostic performance of the multivariate prediction model, combined threshold model, and ACR-TIRADS for malignant and benign thyroid nodules. The multivariate prediction model had the best diagnostic performance with an AUC of 0.946 (0.896–0.996) at a cutoff value of 0.198, which was higher than the AUC achieved by the combined threshold model (DWISIR and ADCmin), with insignificance difference (P = 0.500). The AUC of combined threshold model (DWISIR and ADC SIR) and ACR-TIRADS were 0.777 (0.648–0.907) and 0.722 (0.588–0.857), respectively. Figure 5 reveals the grouped scatter plots of the two combined threshold models, and Table S3 summarizes the Delong test results for AUC comparison among different models.

Table 4 The diagnostic performance of models based on quantitative DWI parameters to identify malignant thyroid nodules
Full size table
Fig. 5
figure 5

The grouped scatter plots of the two combined thresholds models. The blue dots are benign lesions, and the red dots are malignant lesions. a is the scatter plot of DWISIR and ADCmin of the lesions. b is the scatter plot of DWISIR and ADCmin of the lesions. Abbreviations: DWI, diffusion weighted imaging; ADC, apparent diffusion coefficient; SIR, signal intensity rate; ADCmin, minimum apparent diffusion coefficient value

Full size image

The sensitivity (90.0%) and NPV (96.6%) were the highest in the multivariate prediction model. The results showed three false negative lesions, all follicular thyroid carcinoma, and seven false positive lesions: three nodular goiters, three adenomatous nodular goiters, and one adenoma. The best specificity and PPV (both 100%) were achieved by the combined threshold model (DWISIR and ADCmin), where five false negative lesions were all follicular thyroid carcinomas. The combined DWISIR and ADCmin had the lowest unnecessary biopsy rate with no false positive cases. The accuracy, sensitivity, specificity, PPV and NPV of the combined thresholds model (DWISIR and ADCSIR) were 81.7%, 70%, 85.5%, 60.9% and 89.2%, respectively. Compared with ACR-TIRADS, the quantitative DWI parameter-based models significantly improved differentiating benign and malignant thyroid nodules.

Discussion

This study proposed diagnostic models based on quantitative DWI parameters without enhancement to differentiate between benign and malignant thyroid nodules larger than 4 cm. The combined threshold model (DWISIR and ADCmin) demonstrated satisfactory diagnostic efficacy with significantly reduced unnecessary biopsy rate.

ADC map, derived from DWI, measures water diffusion in tissue and provides a new imaging biomarker for the diagnosis of benign and malignant tumours [25]; it has proven to be effective in predicting tumor malignancy [26]. Malignant thyroid nodules had much lower ADC values than the benign thyroid nodules, according to numerous studies [22, 27, 28], but they were for nodules smaller than 4 cm. We measured ADCmin, ADCmean and ADCSD, and found that the former two were associated with malignancy. In a meta-analysis of 2137 thyroid nodules, ADCmean was demonstrated to be a useful tool for differentiating between benign and malignant thyroid tumors and should be used in routine preoperative clinical testing. The ADCmean was 1.88 × 10–3 mm2/s in the benign nodules and 1.15 × 10–3 mm2/s in the malignant [22]; herein, for lesions larger than 4 cm, the ADCmean value was 1.74 × 10–3 mm2/s in benign nodules and 1.08 × 10–3 mm2/s in malignant nodules. ADCmin was found to be an independent predictor in our study. One study of benign and malignant lymph node metastases in the breast indicated that ADCmean had better diagnostic efficacy than ADCmin [29]. However, in a study of benign and malignant prostate tumors, ADCmin was found to be superior to ADCmean [30], which was similar to our results. Due to the multicollinearity between ADCmin and ADCmean in the combined threshold model, we chose ADCmin as it had a greater AUC value in the univariate analysis.

We also investigated the signal intensity-related parameters of DWI and corresponding ADC images on PACS. Leila et al. [31] reported that static MRI measurements like signal intensity and heterogeneity were unuseful in distinguishing between benign and malignant lesions. However, Wang et al. [32] showed that DWISIR was lower in benign nodules than in malignant ones, corroborating our findings. Our study found that DWISIR was another independent predictor of malignant nodules, with malignant nodules showing a higher DWISIR. which corroborated with our findings. DWISIR can complement ADCmin, and using the combined threshold model (DWISD and ADCmin) reduced the number of false positive cases by four and resulted in the highest specificity (100%). According to ACR-TIRADS, the biopsy is recommended for lesions with TR ≥ 4 and diameter > 1.5 cm. Therefore, all nodules with TR ≥ 4 in this study required biopsy to determine their benignity and malignancy, and the unnecessary biopsy rate was up to 56.3%. The combined threshold model (DWISIR and ADCmin) had the lowest unnecessary biopsy rate. DWI may be performed before preparation for biopsy and assisted in determining the need for biopsy and surgery based on quantitative DWI. Quantitative DWI may become a method of thyroid nodules larger than 4 cm surveillance to aid clinicians in their medical decisions.

In addition to ADCmin, we also measured ADCSIR, which is easily available on PACS. ADCSIR was lower in malignant nodules than in benign nodules significantly, which may be because malignant lesions have dense parenchymal cells, narrow cell spaces, and relatively limited extracellular water molecule diffusion. While the combined threshold model (DWISIR and ADCSIR), while it was inferior to the combined threshold model (DWISIR and ADCmin), it still resulted in nine fewer false positive cases than ACR-TIRADS alone.

This study found that individual parameters performed poorly in distinguishing the benignity of follicular thyroid neoplasm. The difference between follicular carcinoma and benign follicular neoplasm is the invasion of the envelope invasion, which can only be detected by postoperative pathology. Our measurements are limited to the substantial lesion component, and this pathological level of distinction is difficult to detect on images.

Although this study provided important insights, there were several limitations. First, selection bias is unavoidable in a retrospective observational study. Second, the sample size was relatively small, necessitating additional validation. Third, reviewing static ultrasound images and reports may differ from real-life clinical practice, introducing some bias. Finally, the study only used a b-value of 800 s/mm2. Therefore, more research must be conducted with various b-values to find the best value.

In conclusion, quantitative DWI parameters can separate benign from malignant thyroid nodules larger than 4 cm. The multivariate prediction and combined threshold model (DWISIR and ADCmin) demonstrated satisfactory diagnostic performance. Our results suggested that quantitative DWI parameters can assess benign and malignant nodules with sizes larger than 4 cm and assist clinicians in pre-operative decision-making.