Thyroid Multimodal Ultrasound Evaluation—Impact on Presurgical Diagnosis of Intermediate Cytology Cases

: Fine needle aspiration (FNA) is recommended as the ﬁnal evaluation in diagnosing thyroid nodules. Intermediate cytology is observed in about 15%–25% of the FNA results and has divergent recommendations: follow-up or surgery, either hemi- or total thyroidectomy. The present study aimed to assess the beneﬁt of multimodal ultrasound (US) evaluation in clarifying the attitude in cases with intermediate cytology on FNA. Sixty-four successive cases with Bethesda III or IV cytology results were evaluated using two-dimensional B-mode US (2B), qualitative strain elastography, as well as planar and volumetric (3D) color Doppler using both a linear multifrequency probe and a linear volumetric probe (Hitachi Preirus Machine, Hitachi Inc. Japan). The analyzed nodules were all categorized as low, intermediate, or high risk, based on the following US criteria: taller-than-wide shape, marked hypoechogenicity, irregular borders, inhomogeneity, microcalciﬁcations, or the presence of suspicious adenopathy. Elastographic criteria included sti ﬀ lesions, while volumetric Doppler criteria included moderate and marked perinodular vascularization, as suspicious for malignancy. The gold standard for the analysis was the result of the pathology report after thyroidectomy. Our results showed that the prevalence of cancer was 25% (16 cases). In ﬁve cases, borderline follicular neoplasia was identiﬁed, group which requires watchful waiting. These lesions were also considered as malignant in the ﬁnal analysis, with a total number of 21 neoplasm cases. Sixteen of the 21 malignant nodules presented with high sti ﬀ ness, while 15 / 21 had increased perinodular vascularization. Eight of the 21 cases displayed both important sti ﬀ ness and perinodular vascularization. Cancer prevalence increased both with severity of sti ﬀ ness (9.0%–15.0%–66.6%–80.0%) and intensity of vascularization in the adjacent perinodular parenchyma (18.2%–27.7%–35.29%–50.0%). Combining the use of grayscale US, elastography, and 3D Doppler in the evaluation of intermediate cytology cases showed a sensitivity of 85.7%, a speciﬁcity of 88.3%, and an accuracy of 90.3% in detecting thyroid cancer cases. We can conclude that the identiﬁcation of highly suspicious US characteristics observed in 2B, qualitative elastography, and volumetric Doppler increases the risk of malignancy.


Introduction
The prevalence of thyroid nodular disease is significant [1], with a consistent increase concerning differentiated carcinomas [2], regardless of the severity or size, and a steady prevalence of other forms of thyroid cancer [3]. The current recommendations propose risk stratification by means of anamnesis, clinical evaluation, and ultrasound characterization [4][5][6]. Positive history, some clinical features, and specific ultrasound (US) patterns are considered high risk.
The conventional diagnostic algorithm recommends fine needle aspiration (FNA) biopsy for the further management of cases. It is recognized as the gold standard for diagnosing thyroid neoplasia [4,6]. Currently, the Bethesda reporting system is recommended [7]. However, it is generally accepted that FNA has unclear results in Bethesda categories III and IV, and there is no consensus about the management of these particular cases: whether to follow-up, refer to surgery, or complete the evaluation with molecular genetic tests [4,6]. Even in experienced centers [8], inconclusive results represent up to 30% of all cytology results. FNA has shown valuable diagnostic value and a favorable overall sensitivity between 70% and 85% [9,10]. The risk of malignancy in the intermediate cytological category covers a wide range, with a mean of 16%, ranging between 6% and 48% [11][12][13]. Our center has a previously reported rate of 28.8% [14] in this particular cytological category.
Multimodal US is being used with increasing frequency, with the purpose of achieving an improved accuracy in the pre-surgical risk prediction [25][26][27].
In cases of intermediate cytology, there are results showing that the sonographic characteristics should influence the further management of these cases [28] by considering molecular diagnostic tests or surgery [29].
Thus, the aim of this study was to determine the diagnostic value of multiparametric ultrasound evaluation in the assessment of thyroid lesions with indeterminate cytology results, with respect to post-FNA attitude.

Subjects
This prospective study was performed in our Endocrinology US Unit ("Dr. D Medical Center", Timisoara, Romania) from January until December 2019. During this recruitment period, 218 patients presenting with solid thyroid nodules displaying suspicious US features were selected for FNA. The cytology analysis identified intermediate results (Bethesda categories III and IV) in 64 of the examined specimens, which were included in the final analysis. In all these cases, total thyroidectomy was recommended and performed by experienced surgeons. Thus, all had a pathology report, which is considered the gold standard for the diagnosis and analysis of US results. Prior to FNA, all nodules included in the study were evaluated using combined advanced US technologies, by an experienced operator with more than ten years of experience in both B-mode (2B) US and strain elastography (SE), and two years of experience in volumetric (3D) Doppler. The study was approved by the Ethics Committee in our centre ("Dr. D Medical Center" CECS nr. 17/09.12.2018) and was performed in accordance with the ethical guidelines of the Helsinki Declaration. All patients gave their written informed consent, which was obtained prior to inclusion.

Ultrasound Techniques
All B-mode US evaluation, SE, and 3D color Doppler measurements were performed in the same session, with a Hitachi Preirus device (Hitachi Medical Corporation, Tokyo, Japan). A multifrequency 5-18 MHz linear probe (Hitachi L64) was used for conventional US and SE, and a linear volumetric multifrequency 5-13 MHz probe (Hitachi VL54) for 3D Doppler assessment. The following criteria were considered: echogenicity, margins, shape and homogeneity, the presence of intranodular calcifications, and the evaluation of laterocervical lymph nodes. SE examination, also called real-time elastography (RTE), was applied using gentle external pressure by a handheld probe, verified on the pressure scale using optimal images and loops. This resulted in a color map image called an elastogram, overlaying the B-mode image [21]. The elastogram was interpreted using a scale of 1 to 4 (soft to hard), according to Asteria et al. [30]. Figure 1 shows the stiffness categories applied to the nodules examined in the study, in blue-red-green standard color map code.
Appl. Sci. 2020, 10, x 3 of 13 All B-mode US evaluation, SE, and 3D color Doppler measurements were performed in the same session, with a Hitachi Preirus device (Hitachi Medical Corporation, Tokyo, Japan). A multifrequency 5-18 MHz linear probe (Hitachi L64) was used for conventional US and SE, and a linear volumetric multifrequency 5-13 MHz probe (Hitachi VL54) for 3D Doppler assessment. The following criteria were considered: echogenicity, margins, shape and homogeneity, the presence of intranodular calcifications, and the evaluation of laterocervical lymph nodes. SE examination, also called real-time elastography (RTE), was applied using gentle external pressure by a handheld probe, verified on the pressure scale using optimal images and loops. This resulted in a color map image called an elastogram, overlaying the B-mode image [21]. The elastogram was interpreted using a scale of 1 to 4 (soft to hard), according to Asteria et al. [30]. Figure 1 shows the stiffness categories applied to the nodules examined in the study, in blue-red-green standard color map code.   Volumetric Doppler examination displays a three-dimensional image of the scanned area, placing the lesion in the middle of the region of interest (ROI), after a rotational 180° scan inside the probe. A thorough evaluation of the nodule's vascularization is completed by the display of spreading vessels from the neighboring parenchyma into the lesion. Figure 2 shows the possible 3D Doppler pattern that we observed. Volumetric Doppler examination displays a three-dimensional image of the scanned area, placing the lesion in the middle of the region of interest (ROI), after a rotational 180 • scan inside the probe. A thorough evaluation of the nodule's vascularization is completed by the display of spreading vessels from the neighboring parenchyma into the lesion. Figure 2 shows the possible 3D Doppler pattern that we observed.

FNA
US-guided FNA was performed in all cases by an endocrinologist in our department, who has more than 20 years of experience. The patient was positioned in the supine position and a pillow was placed beneath their shoulders, in order to get the best possible hyperextension of the neck. Firstly, a thorough disinfection of the skin was performed, followed by local anesthesia using topical application of 2 g of 5% prilocaine/lidocaine anesthetic cream on the skin at the side of the nodule. The transducer probe had a sterile covering to protect the patient when locating the lesion. Very fine sterile needles (25 gauge) attached to 10 cc syringes were used. After locating the nodule on US, the needle was introduced parallel to the probe, using the anterior technique. US guidance was utilized to visualize the needle and confirm that it had entered the nodule. "Coring" movements were executed with the needle within the lesion for the cellular material to be gathered. At least two punctures per nodule were made, with the obtained material spread on glass slides and fixed in 95% ethyl alcohol (minimum 5 slides per nodule). The quality of the obtained material was not assessed at collection. An experienced cytologist in the field reviewed the specimens and reported the findings according to the Bethesda System for Reporting Thyroid Cytopathology. All samples were reliably assigned to one of the five categories. The indeterminate categories were considered Bethesda III and IV. Findings in this group were included in the study.

Surgical Intervention and Pathological Examination
Total thyroidectomy was recommended in all cases, and was performed by experienced surgeons in our team. In a few cases, suspicious lymph nodes were detected and therefore their dissection was also necessary. The pathologic diagnosis was given by specialists in thyroid pathology from the Pathology Department. In some cases, a more complete evaluation was performed, including immunohistochemical stain for Ki-67 protein, cytokeratin 19 (CK-19), Hector Battifora mesothelial antigen (HBME-1), and thyroid transcription factor-1 (TTF-1).

Statistical Analysis
Data were analyzed using SPSS statistical software version 17 (SPSS Inc. Chicago, IL, USA). Descriptive analysis was used for the bio data, cytology results, and all the US-evaluated parameters. Cross-tabulation was used to study the relationship between two or more variables. Correlational analysis was applied in order to evaluate association as well as relationship between them, and hence correctly identify dependent and independent variables. The Student's t-test assessed the differences between groups for continuous variables with normal distribution. Chi-squared and ANOVA tests were applied for inferential statistics. Qualitative variables were displayed as numbers and percentages. Sensitivity (Se-true positive/real positive cases), specificity (Sp-true negative/real negative cases), positive predictive value (PPV-true positive/all positive cases), negative predictive value (NPV-true negative/all negative cases) and accuracy (sum of true positives and true negatives/total cases) were calculated for all the evaluated US parameters.

Results
The final evaluation included 64 patients, 52 were females and 12 were males. The mean age in the group was 32.5 years (within 23-56 years old). Consecutive to surgery, the pathology report confirmed the malignancy diagnosis in 16 cases (25%). Papillary carcinoma was found in the majority of these cases (15/16). Follicular carcinoma was diagnosed only in one case (6.25% of the confirmed malignancies). Forty-three of the nodules were classified as certainly benign (67.18 %). About half of them, 21/43 (32.28% of the total examined cases), were follicular adenomas, 8 (12.5%) were hyperplasic nodules, 5 (7.81%) were Hurthle cell adenomas, 2 (3.1%) were granulomatous thyroiditis, and 6 (9.37%) were autoimmune thyroiditis. Five cases (7.81%) were classified as "follicular neoplasia with uncertain risk" or "borderline neoplasia". This category included three cases described as having some invasion of the capsule without involving its entire width, and two cases with tumoral cell islands detected within the capsule, and associated disruption of the collagen matrix. According to current recommendations [31][32][33][34] we considered them as malignant cases because of the suggestion of watchful waiting due to their unclear evolution.
The size of the examined nodules ranged within 11.5-32.5 mm with a mean of 14.7 mm. It was observed that benign lesions were bigger compared to the malignant ones (mean diameter of 21.1 mm for benign cases and 12.3 mm for the malignancies).
The general characteristics evaluated for each nodule are illustrated in Table 1. Considering qualitative SE scores 1 and 2 as predictive for benignity and scores 3 and 4 as suspicious of malignancy, stiffness evaluation identified 42 low-risk lesions (22 corresponding to score 1 and 20 to score 2) and 22 high-risk nodules (12 in score 3 and 10 in score 4). Three of the five cases with a final diagnosis of malignancy that were classified as low risk on elastography assessment were borderline neoplasia-a category which was considered in this study together with the confirmed cancers, given their uncertain evolution. As for stiff lesions, most of the score 4 nodules (80%) and an important proportion of score 3 cases (66.6%) proved malignant. Qualitative SE alone proved to be of good diagnostic value for detecting cancer due to its 76.2% sensitivity, 86% specificity, PPV of 72.7%, NPV of 88.1%, and an accuracy of 82.81% (54/64 cases). Compared to soft nodules, stiff lesions presented a considerably greater risk of malignancy (p = 0.020). The distribution of cases according to stiffness scores is reported in Table 2. Figure 3 shows that the area under the receiver operating characteristic (ROC) curve was very good for SE-only information (0.83, CI = 0.721−0.916). This outlines the great value that SE brings to the risk assessment, even when considered on its own. . At 3D Doppler evaluation, 11 nodules presented with no vascularization in the thyroid tissue surrounding the nodule, 15 nodules had low perinodular vascularization, and 36 nodules had moderate and marked perinodular vascularization. The vast majority of malignant nodules showed significant vascularization in the surrounding tissue. Taking into consideration that lesions with low and absent perinodular vascularization are probably benign and those with moderate to high perinodular vascularization are suggestive of malignancy, volumetric Doppler showed 71.43% sensitivity, 90.7% specificity, a PPV of 78.9%, NPV of 86.7%, and an accuracy of 84.37% in detecting thyroid cancer. The cancer prevalence values in the described 3D Doppler categories are shown in Table 3. When 3D Doppler is considered alone, the area under ROC curve displayed in Figure 4 also shows a very good value (0.85, CI = 0.747−0.932).  At 3D Doppler evaluation, 11 nodules presented with no vascularization in the thyroid tissue surrounding the nodule, 15 nodules had low perinodular vascularization, and 36 nodules had moderate and marked perinodular vascularization. The vast majority of malignant nodules showed significant vascularization in the surrounding tissue. Taking into consideration that lesions with low and absent perinodular vascularization are probably benign and those with moderate to high perinodular vascularization are suggestive of malignancy, volumetric Doppler showed 71.43% sensitivity, 90.7% specificity, a PPV of 78.9%, NPV of 86.7%, and an accuracy of 84.37% in detecting thyroid cancer. The cancer prevalence values in the described 3D Doppler categories are shown in Table 3. When 3D Doppler is considered alone, the area under ROC curve displayed in Figure 4 also shows a very good value (0.85, CI = 0.747−0.932). Most of the cases that proved malignant on the pathology report (15/21) did show important perinodular 3D vascularization. Although 3D Doppler seems to be a good evaluator by itself, it is more relevant to include it in a more complex algorithm for detecting cancerous lesions. Most of the cases that proved malignant on the pathology report (15/21) did show important perinodular 3D vascularization. Although 3D Doppler seems to be a good evaluator by itself, it is more relevant to include it in a more complex algorithm for detecting cancerous lesions.
The risk of malignancy predicted by conventional US alone was compared to the risk upgrade given by the elastographic information, 3D Doppler information, and both volumetric Doppler and elastographic information. The compared sensitivity, specificity, PPV, NPV, and accuracy are presented below in Table 4. We considered iso/hyper or mildly hypoechoic nodules without any suspicious sign as low risk; while those that were intensely hypoechoic as mild risk. In cases where more than two suspicious signs were present, we considered the nodule to belong to the high-risk class. When taking into account all information given by 2B ultrasound as well as both elastography and volumetric Doppler, we considered the most severe ultrasound characteristics: conventional, strain, or perinodular vascularization. There was only an upgrade type of risk evaluation; we did not downgrade the risk category in the presence of low stiffness or absence of perinodular vascularization. Table 4. Diagnostic qualities of different models of risk assessment in the studied cases (%).

Se
Sp PPV NPV Acc 2B 57.14 67. 44   There are some other signs in grayscale US that associate a sensitivity of 100%, such as intranodular microcalcifications and the presence of lymph nodes with altered central hilum. However, the low prevalence of these signs does not ensure a good specificity.  The risk of malignancy predicted by conventional US alone was compared to the risk upgrade given by the elastographic information, 3D Doppler information, and both volumetric Doppler and elastographic information. The compared sensitivity, specificity, PPV, NPV, and accuracy are presented below in Table 4. We considered iso/hyper or mildly hypoechoic nodules without any suspicious sign as low risk; while those that were intensely hypoechoic as mild risk. In cases where more than two suspicious signs were present, we considered the nodule to belong to the high-risk class. When taking into account all information given by 2B ultrasound as well as both elastography and volumetric Doppler, we considered the most severe ultrasound characteristics: conventional, strain, or perinodular vascularization. There was only an upgrade type of risk evaluation; we did not downgrade the risk category in the presence of low stiffness or absence of perinodular vascularization. There are some other signs in grayscale US that associate a sensitivity of 100%, such as intranodular microcalcifications and the presence of lymph nodes with altered central hilum. However, the low prevalence of these signs does not ensure a good specificity.

Discussion
Thyroid nodules present an increasing prevalence, hence, in order to avoid unnecessary surgeries, the target of clinicians when investigating thyroid nodules is to accurately identify the ones that are at risk. In this regard, FNA is still considered the most appropriate instrument for the characterization of thyroid lesions [4,6], in spite of its notable diagnostic limitation, consisting of indeterminate results. Different anamnestic and clinical conditions were investigated as possible independent predictors of malignancy in intermediate cytology cases-male gender was found to increase the odds for thyroid cancer by 1.5 times [35], as well as the growth and dimensions of the nodule. The predictive value of the size of the nodule is controversial [36]. Some data report that nodules larger than 4 cm increase the odds of thyroid cancer by two [35], others suggest that the lowest risk is suggested by nodules measuring 2.5 cm, and any increase or decrease in size by 1 cm will raise the risk by 40% to 50% [37]. Because of the conflicting data regarding gender, age, number of nodules, and size, we did not take these parameters into consideration.
In intermediate cytology cases, several studies have described that high-risk ultrasound features, regardless of the model of risk assessment [14,[38][39][40], associate with high risk for thyroid cancer.
Conventional ultrasound evaluation remains the first approach in the examination of thyroid nodular pathology. Solid consistency, irregular margins, the presence of intranodular point/scattered calcifications, intense hypoechogenicity, and/or a taller-than-wide shape are the generally accepted parameters that define high-risk nodules [4,5,20].
Even if some particular ultrasound characteristics such as calcifications and metastatic lymph nodes [4] are highly suggestive of malignancy, their prevalence in nodular thyroid disease is rare. On the other hand, solid pattern or hypoechogenicity are present in both benign and malignant cases, thus US assessment does not offer good diagnostic quality for detecting malignancy when only one parameter is considered [20].
Diagnostic performance of conventional US-only evaluation has been studied before, with excellent results for nodules classified in Bethesda categories I, II, V, and VI, but poorer outcomes for intermediate categories III and IV, underlining the importance of additional evaluation for these cases [47]. We observed an enhancement of risk predictions by using combined conventional ultrasound characteristics and volumetric Doppler: 68.75% versus 64.06%; classical grayscale characteristics and degree of stiffness: 85.93% versus 64.06%; and all conventional, volumetric Doppler and elastography score: 90.32% versus 64.06%. Similar data were proposed by previous research with integrated elastographic information [14,42,44]. By combining sonographic patterns, elastography scores, and cytological details in cases of intermediate cytology, we achieved an area under the ROC curve of 0.87 [48].
There are convincing data indicating that incorporating clinical risk assessment and complex ultrasound characteristics can allow a personalized management of cases with intermediate cytology [28]. Those with very low suspicion would be evaluated with ultrasound only, those with low/intermediate suspicion would have a repeat FNAC, and those with high suspicion would undergo diagnostic surgery.
Both elastography and volumetric Doppler show very good NPV (92.68% and 82.85%, respectively), reconfirming and endorsing the recommendation of active follow-up in cases with Bethesda III and IV FNA that associate with low-risk US pattern, as outlined in previous research [49,50].
Only one follicular neoplasia case was diagnosed among the cases investigated in our study, so a conclusion cannot be derived. However, similar diagnostic performance is expected for combined advanced US techniques.

Conclusions
Our results suggest that incorporating volumetric Doppler results and stiffness information alongside the classical sonographic features in the risk evaluation of intermediate cytology cases adds diagnostic confidence and narrows the number of unclear cases, with respect to treatment recommendations. Funding: This research received no external funding.

Conflicts of Interest:
The authors declare no conflicts of interest.