US continues to be the predominant imaging modality for the assessment of metastatic cervical adenopathy in PTC patients and serves as the major tool for both preoperative evaluation and follow-up examination (10). Currently, several US features of LNs are predictive of malignancy, including hypo-echogenicity, absence of hilum, cystic changes, round shape, increased vascularity, and microcalcifications ( Figure 1 ) (10, 11). However, the use of US alone is insufficient for the assessment of cervical LNM (12, 13). Previous reports have indicated that the efficacy of conventional US is disputed due to its inconsistent sensitivity, which ranges from 0.33 to 0.7 (14). Contrast-enhanced US (CEUS) is a highly effective method for examining tissue vascularity compared to conventional ultrasound (15). The application of this technique has demonstrated great efficacy in accurately distinguishing between malignant and benign LNs in PTC (16–18). Among the preoperative sonographic features of CEUS, heterogeneous enhancement, nonenhancement, and centripetal perfusion are independent ultrasonographic characteristics of malignant LNs, while homogeneous enhancement, ring enhancement, and centrifugal perfusion are those of benign LNs ( Figure 1 ) (19, 20). In 2023, two publications showed that significant progress was made in the field of CEUS-based LN imaging, especially in facilitating the precise identification of siLNs (20, 21). In the first article, Zhang et al. (20) included 64 participants with suspected PTC (76 LNs in total) and compared the diagnostic performance of CEUS and US for lymphatic metastasis. The sensitivity, specificity, and accuracy of the CEUS group were 16%, 10%, and 3% greater than those of the US group, respectively. Through further careful exploration of the association between diagnostic performance and LN size and location, they discovered that CEUS exhibited superior diagnostic accuracy when the LNs were measured 1 cm or smaller and were located in the central neck region. The second study specifically investigated the value of CEUS in the identification of siLNs. Of note, siLNs were defined as those with short-axis diameters less than or equal to 8 mm rather than 1 cm. This research performed a comprehensive evaluation of these LNs in the vascular and postvascular phases separately. This evaluation indicated that the presence of perfusion defects during the vascular phase exhibited a significantly high level of specificity, reaching 96%. Additionally, the absence of iso-enhancement in the postvascular phase demonstrated a high level of accuracy, particularly 93%, as well as a negative predictive value of 100%. Finally, it was confirmed that the combined model of US and postvascular phase features yielded better results than did the model that relies solely on US features for the purpose of diagnosing potentially malignant small lateral cervical LNs (AUC 0.94 vs. 0.73; P <.001) (21). Two reports utilized the same contrast agent, Sonazoid, owing to its advantageous characteristic of providing a distinct postvascular phase (Kupffer phase) (22). In contrast, a distinctive difference of them was the method used to inject Sonazoid; namely, the former directly injected the agent into the middle lobe of the thyroid, while the latter was an intravenous bolus injection. The benefit of intrathyroidal injection lies in the agent being absorbed by sentinel LNs, similar to sentinel LN mapping, resulting in more precise and targeted detection of metastatic LNs. Moreover, both articles, interestingly, were published on the same day.

Even so, a subsequent investigation revealed the suboptimal performance of CEUS in terms of diagnostic sensitivity for LNM within the diameter ranges of 0–5 mm and 5–10 mm, with sensitivities reaching only 61.0% and 66.6%, respectively (23). One potential explanation could be attributed to the use of another contrast agent, SonoVue, in this study. A detailed comparison of the difference between SonoVue and Sonazoid can be found in another review (24). At times, the effectiveness of CEUS is limited due to certain inherent sonographical defects. One difficulty is differentiating between the subtle enhancements in tiny metastatic foci and the surrounding lymphatic parenchyma, which is primarily associated with the indolent characteristic of PTC (17). Another explanation is that homogenous and iso-enhancement can be observed in both benign and metastatic lymph nodes (25).

Following US, CT is recommended by the ATA2015 guideline to serve as a secondary option for imaging the status of cervical LNs (9, 10). When utilizing conventional CT, it is also challenging to diagnose LNs that measure less than 1 cm. Therefore, Liu et al. (26) developed a parameter model based on dual-source dual-energy CT at a small field of view for the detection of small LNMs, which showed superior diagnostic accuracy compared to that of the CT image model. When combined with US, Zheng et al. (27) also constructed a CT-based deep learning model for the diagnostic assessment of suspicious lateral LNs. As reported, its diagnostic performance is comparable to that of FNA and experienced radiologists. In addition, dual-energy CT parameters could complement US in identifying siLNs (28). The main obstacles associated with CT images mostly stem from the limited diagnostic value of known metastatic signs. In this regard, a recent report has assessed the diagnostic efficacy of several contrast-enhanced CT features from the K-TIRADS guidelines within siLNs. The results demonstrated that the removal of “heterogeneous enhancement” from the K-TIRADS LN classification led to a higher specificity and positive predictive value than before (29). Additionally, Lee et al. (30) investigated the predictive accuracy of CT for detecting cervical lymphatic micrometastasis in PTMC patients who had opted for active surveillance (AS). They found that CT displayed a better diagnostic performance in detecting cervical LNM than US and that CT provided additional benefit after the evaluation of cervical LNs by US. Regardless of the individual effectiveness of CT or US, most reports on the use of dual imaging, including both CT and US, have shown that the detection rate has increased markedly (31). Nonetheless, in this study, the combination of US and CT did not improve diagnostic accuracy compared with the use of CT alone, which is distinct from what was stated previously. The likely reasons for this difference are the individuality of the LN status and the unique patient selection criteria for AS. Hence, it should be noted that the conclusions were applicable only to this specific scenario. In summary, further studies are warranted to identify additional dimensions in which CT has superior efficacy; otherwise, CT will be used as an auxiliary approach for US most of the time.

Guidelines strongly recommend FNA for the diagnosis of siLNs, of which fine needle aspiration cytology (FNAC) and fine needle aspiration thyroglobulin (FNA-Tg) are two popular techniques (10, 32). A recent meta-analysis, which included twenty-one studies and pooled 1662 malignant and 1279 benign LNs from 2712 patients, revealed that FNAC had a greater level of specificity than FNA-Tg, while FNA-Tg showed greater sensitivity than FNAC (33). Moreover, the study concluded that the combination of FNAC and FNA-Tg, whether used preoperatively or postoperatively, could achieve better diagnostic performance than either method alone. In mainstream guidelines from seven international societies, including the American College of Radiology (ACR), ATA, American Association of Clinical Endocrinologists (AACE), ETA, KTA, National Comprehensive Cancer Network (NCCN), and SRU, only those of the ATA, KTA, SRU, and ETA provided the detailed US characteristics and LN size that can be used to determine the optimal timing for conducting FNA of suspicious LNs ( Table 1 ) (10, 32, 34–38). According to KTA, the minimum achievable size for FNA is 3 mm. However, some experts believe that FNAC is indicated only for LNs larger than 5 mm to obtain an adequate number of cells for a more reliable diagnosis, which is consistent with the recommendations of other guidelines except for KTA (39). Clinically, when the diameter of the LN is less than 10 mm, it is challenging to puncture the center of the lesion; thus, the procedure is best completed by skilled radiologists and cytologists to increase the success rate (40, 41). Hydrodissection is a promising technique used to separate target lesions from nearby essential structures for the ablation of cervical lesions. At present, hydrodissection is being applied in FNA procedures. Cheng et al. (42) evaluated the safety and effectiveness of hydrodissection-assisted small LN biopsy by calculating the separation success rate of hydrodissection, the technical success rate, and the histopathologic success rate. According to their results, all of these related indicators achieved an astonishing 100%. Therefore, they established that hydrodissection enabled the safe and efficient performance of US-guided biopsy of subcentimeter cervical LNs near important arteries. Chen et al. (41) reviewed the medical records of 996 suspicious lateral LNs aspirated 2 to 3 times and classified them into two groups based on their respective sizes (large group and small group). They observed that the incidence of malignancy in larger LNs was marginally greater than that in smaller LNs, whereas LN size was not an independent predictor of malignancy according to multiple regression analysis. In contrast, there are several paradoxical results regarding the predictive role of LN size, whereas other studies have identified it as a predictor (36, 43, 44). Therefore, we should understand that being tiny may not necessarily indicate a reduced likelihood of malignancy. Notably, patients undergoing FNAC have a 0.5% cumulative incidence of needle tract implantation (45). It suggested that mindlessly increasing aspiration times or expanding puncture indications was not advised, despite this issue often being overlooked in previous clinical practice.

Ever since Pacini et al. first presented a method in which FNA was combined with thyroglobulin washout, this method has proven to have outstanding specificity and selectivity in clinical practice (46–48). Currently, FNA-Tg has the ability to detect and identify indeterminate LNs with a diameter as small as 5 mm (49). However, the primary concern in the current FNA-Tg technique revolves around the determination of a universally applicable and optimal cutoff value (47). Recent studies have revealed a collection of Tg thresholds and their corresponding indicators, encompassing sensitivity, specificity, and accuracy. For instance, Zhu et al. (50) analyzed a total of 22 studies and performed a subgroup analysis of patients according to the cutoff value; the results confirmed that the threshold of 1 ng/mL had the maximum sensitivity, while 40 ng/mL had the highest specificity. The measurement of TgAb via FNA is also a feasible way to detect the presence of metastatic LNs (i.e., FNA-TgAb). Nonetheless, compared with FNA-Tg, FNA-TgAb exhibited poorer performance in identifying LNs with a diameter less than 10 mm (51). Indeed, it is well recommended to utilize FNA-Tg at any FNA facility, particularly when LNs are low in volume (50). Furthermore, the latest studies have suggested several points that need to be considered during clinical application: (1) FNA-Tg should be interpreted with caution only if the serum Tg concentration exceeds 10 ng/ml (47); (2) the absolute Tg content may be a potential alternative that could provide a more objective expression of Tg washout (52); and (3) FNA-TG washout may be particularly valuable for diagnosing cystic or highly vascular LNs, as these types of LNs may yield negative FNAC results (53). In addition to medical imaging, FNA serves as the most commonly used approach because of its considerable sensitivity and specificity. Along with its associated techniques, such as FNAC and FNA-Tg, FNA has significant potential for making further advancements in diagnosing siLNs (54).

Facing the difficulties in identifying and dissecting small LNs with indeterminate imaging indications of metastasis, sentinel lymph node biopsy (SLNB) has attracted increasing amounts of attention from oncologists and clinicians because of its high detection rate and minimal invasiveness. In addition to being initially used in melanoma and breast cancer, SLNB was also applied to detect indeterminate metastatic LNs in the TC and acts as an alternative to replace cervical LND (55). As the main tumor drainage LNs to which LNM first occurs, sentinel lymph nodes were marked with the help of specific LN dyes. The stained LNs were removed from the suspicious compartment, while the unstained LNs were reserved. Selective removal of stained sentinel LNs effectively eliminates the uncertainty of random sampling, and the subsequent frozen pathology reliably verifies whether LNM occurs at this level. In this way, SLNB provides an accurate reflection of the status of the remaining LNs, particularly when some of them appear as siLNs (56). In the event that the SLNB yields a negative result, siLNs within the remaining portion of LNs can be confirmed to be benign, regardless of whether some siLNs are included in the sampled portion. Afterward, radical whole-level LND of the suspicious compartment and underlying surgical complications were avoided; otherwise, the opposite was true ( Figure 2 ). Table 2 presents a comparative analysis of SLNB studies that employed different modalities over the past five years. It was reported that an astonishing 100% detection rate of SLNB could be achieved, while the false-negative rate was comparable to approximately 20%, or even as high as 40% (55, 57). Combined with imaging examination, Liu et al. (58) recently proposed a new sentinel LN technique that uses conventional US and CEUS to assign every targeted sentinel LN a score (range from 0–5). The results demonstrated that the sensitivity and specificity could reach 97.1% and 93.9%, respectively, at a cutoff of 4 or higher.

Several related reports have been published presenting controversial results regarding the benefit of SLNB. On the one hand, SLNB was shown to have a low identification rate and only moderate sensitivity in the study by Albers et al. (57), who suggested that the concept of the sentinel LN as the first LN is not fully applicable to TC due to the interconnected communication of lymphatic vessels within and outside the thyroid gland. Therefore, the reports conclude that SLNB is inappropriate as an adjunct to surgery within the framework of contemporary PTC treatment concepts. On the other hand, some studies employing the same methods as those of Albers et al. have suggested the opposite findings, where SLNB was still considered to be a reliable method with excellent diagnostic accuracy, as least in the early stage of PTC (cT1–2, cN0) (59, 60). Past studies have focused mainly on the utility of SLNB in central compartment LNs, whereas SLNB has also begun to emerge as an effective diagnostic procedure for lateral compartment LNs in papillary and medullary TC (61, 62). Furthermore, one publication noted that the administration of activated charcoal suspension could aid in the visual confirmation of small and non–palpable LNs during lateral neck dissection in patients with recurrent differentiated TC (63, 64).

Overall, SLNB demonstrates promise in improving the decision–making of siLNs, but the issues of a high false–negative rate and deficiency in tracer techniques persist (55). Nevertheless, continual upgrades are ongoing. For example, the innovative application of 68Ga-tilmanocept PET/CT has solved the longstanding “shine-through” problem in SLNB, which refers to the accumulation of peritumorally injected radiotracers in the tumor bed and has produced a large hotspot in imaging (65). Moreover, successful cervical SLN identification using 68Ga-tilmanocept PET/CT, especially in lateral compartments, has been demonstrated in a proof-of-concept study (66).

Despite the various efficient techniques described above, there remain patients for whom the diagnosis of cervical small metastases using these methods proves hard. When a series of preoperative examinations of patients with small LNMs are clinically suspicious but not yet confirmed, the decisions of whether and to what extent to perform the dissection are complicated. In this scenario, after the possibility of local complications, surgical risks, and patient preference are balanced, intraoperative frozen section (FS) may be considered a last-resort diagnostic option. FS referred to a thorough regional LN dissection of the compartment suspected of metastasis based on anatomical or surgical boundaries, and then the specimens were sent to the pathology department for rapid testing (67, 68). All LNs within the compartment, including those smaller than 1 cm, are sent for frozen pathology examination, which helps alleviate the diagnostic challenges associated with the low volume of small LNs. Even if all the submitted LN specimens were less than 1 cm, the FS demonstrated excellent validity and consistency (69). Finally, according to the final FS report, the decision and scope of LND could be determined (the detailed instructions can be found in Section 4) (70).

At first glance, FS appeared to bear a striking resemblance to SLNB. There were, in fact, two noticeable differences between them: (1) the former requires whole-level dissection, whereas the latter solely involves the sampling of LNs stained with lymphatic channel dye, which are a portion of LNs within the entire compartment; (2) the purpose of the former is to determine whether extended cervical LND should be performed, whereas the latter is aimed at determining whether to remove the remaining LNs within the sampled level. According to current studies, FS seems to have a greater sensitivity and lower specificity in comparison to SLNB (67, 68, 71). A recent study indicated that FS could be replaced by preoperative FNAC combined with FNA-Tg because of its better diagnostic ability (67). Nevertheless, the observed outcome seemed to exhibit a degree of inconsistency with our actual experience derived from clinical practice. Collectively, FS pathology of the entire compartment was deemed to be a somewhat aggressive but pretty reliable diagnostic method. The term “aggressive” refers to underlying surgical complications such as nerve injury and lymphatic leakage (72), while “reliable” refers to the ability of FS to identify all LNs that require cleaning because patients who are negative in frozen sections or even positive in permanent sections are proven to be unnecessary for LND (73).

CK19 is a type of cytokeratin with a relatively small molecular weight that is commonly expressed in both normal epithelial cells and epithelial-origin tumors (74). Obvious differences in expression can be observed between differentiated TC and benign thyroid lesions (75, 76).

At the mRNA level, a new biomolecular technique called one-step nucleic acid amplification (OSNA) can be used to quantify the number of CK19 mRNA copies (77, 78). Several studies have assessed the intraoperative diagnostic accuracy of OSNA for detecting LNMs in PTC patients (79–82). Although they incorporated LNs with varying diameters ranging from 3 mm to 17 mm, the vast majority of LNs were composed of LNs less than 10 mm in length, and OSNA showed a high level of accuracy for diagnosing indeterminate LNs. A meta-analysis incorporated several other studies and provided a pooled sensitivity of 88% and specificity of 90% (83). This result was derived from a comprehensive analysis of all LNs, regardless of size. Additionally, a positive linear relationship between CK19 expression and LN size was proven in a prospective study (81).

At the protein level, cytokeratin fragment 21-1 (CYFRA 21-1) is a proteolytic portion of CK-19 and has been proposed to be a dependable biomarker owing to its high expression in the extracellular fluid (84, 85). Lee et al. (86) reported that washout CYFRA 21-1 measurements in indeterminate LNs served as a valuable diagnostic approach, by which 91% of LNs with discordant FNAC and FNA-Tg results were accurately confirmed. Moreover, a new technique termed immunochromatographic strip significantly decreased the detection time of CYFRA 21-1, which has dramatically increased the application prospects of the approach (87). Despite its great potential, the effectiveness of CYFRA 21-1 measurements still lacks additional validation in patients with siLNs.