Development and validation of an ultrasound-based radiomics nomogram for predicting metastatic central cervical lymph nodes in clinically lymph node-negative patients with papillary thyroid microcarcinoma

Introduction

Papillary thyroid microcarcinoma (PTMC) refers to thyroid papillary carcinoma of which its maximum diameter is less than or equal to 10 mm (1), with high incidence due to the rapid development of fine needle aspiration biopsy (FNAB) and extensive surgeries (2). Although PTMC is considered as an indolent growth tumor, the rate of cervical central lymph node metastasis (CLNM) in PTMC patients is 20% to 60% approximately (3-5), with great differences. At present, there is a common consensus in medicine that the standard surgical approach for patients with clinically positive central lymph nodes (CLNs) entails total thyroidectomy and therapeutic central lymph node dissection (CLND) (6,7). However, the utility of routine prophylactic central neck dissection (pCND) in patients with clinically lymph node (LN)-negative (cN0) PTMC is still subject to ongoing debate. Meanwhile, the rate of CLNM in cN0 PTMC patients is 30% to 50% approximately (8-10). Zhao et al. reported that prophylactic cervical cleaning for cN0 in PTMC patients might improve survival, local recurrence and the level of thyroglobulin (Tg) after operation, while others pointed out that it did not lead to expected survival benefits such as reducing recurrence in the long run but increased the risk of temporary complications, such as hypocalcemia (11-14). Jin et al. found that the five-year recurrence-free survival rates were of no significant difference between the pCND and non-pCND group (P=0.976) (15). The American Thyroid Association (ATA) guidelines suggest that in patients with PTMC, the fundamental goals of treatment are to increase overall survival quality, reduce disease-related complications, and carry out disease staging and risk stratification precisely, while lowering the therapy-associated morbidity and unnecessary treatments, and an active surveillance management approach to papillary microcarcinoma is optional (16). The excessive medical treatments in thyroid nodules are associated with longer hospitalization time and additional costs, which are brought into focus. Thus, if the status of LNs is diagnosed and predicted accurately before operation, unnecessary prophylactic surgery treatments and related syndromes can be eliminated as much as possible.

Two-dimensional (2D) ultrasound (US), as the non-invasive auxiliary examination, has been widely used to evaluate thyroid lesions and cervical LNs, to complete risk stratification, through differentiating the shape, margin, echogenicity and so on. If a thyroid nodule shows solid component, hypoechogenicity (compared to the thyroid gland), marked hypoechogenicity (echo intensity is lower than hypoechogenicity), microlobulated or irregular margins, microcalcifications, and taller-than-wide shape in US examination, it is highly possible that this nodule is associated with malignancy (17). Meanwhile, preoperative ultrasonic testing demonstrates great diagnostic performance for metastatic lateral cervical LN of papillary thyroid carcinoma (PTC) (18). However, the sensitivity of cervical US in ascertaining central cervical LN status is limited on account of the interference of neck tissue and dependence of the operators’ skills (19).

Radiomics analysis, a method by using computer algorithms to extract hidden quantitative features of tumor from medical images, has been applied to a large number of disciplines, and it is most well developed in oncology. Radiomics could make up for the deficiency of traditional 2D US, improve diagnostic accuracy and achieve prediction to assist clinical decision-making, without invasive treatment, which is activated by the idea that tumor images contain underlying pathophysiology information (20-22). Recently, the radiomics analysis based on US images of thyroid nodules has been proved to be favorable for predicting the CLNM of PTC (23).

Currently, a search of the existing literature revealed no studies on the use of a US-based radiomics nomogram to predict CLNM for cN0 patients with PTMC. Therefore, the purpose of this study is to develop and validate a nomogram that combines radiomics signature with ultrasonic features, so as to contribute to the individualized preoperative prediction of CLNM to facilitate designing the first-rank treatment strategy for patients. We present this article in accordance with the TRIPOD reporting checklist (available at https://gs.amegroups.com/article/view/10.21037/gs-2025-aw-512/rc).

Methods

Patients

This retrospective study was approved by the Institutional Ethics Committee of The First Affiliated Hospital of Nanjing Medical University (No. 2022-SR-512) and written informed consent was collected from all the patients. All study procedures complied with the relevant institutional guidelines and regulations. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

A total of 266 patients, ranging in age from 24 to 73 years and each presenting with a suspicious thyroid malignant nodule, were enrolled in this study [Thyroid Imaging Reporting and Data System (TIRADS) categories 4a–5, developed by Kwak et al. (24)]. These patients underwent preoperative 2D US, FNAB (cytology and BRAF^V600E analysis) and surgery from November 2021 to October 2024 in an institution. Using a stratified randomization approach, patients were split into training and internal validation sets (70% vs. 30%). An independent external validation set was formed by enrolling 62 consecutive patients from a different institution who met the same inclusion criteria. Across all sets, we categorized patients into two groups according to the pathological findings of the central cervical LNs obtained during neck dissection. In training set, 72 patients with non-metastatic LNs and 72 patients with metastatic LNs. In internal validation set, 30 patients with non-metastatic LNs and 30 patients with metastatic LNs. In external validation set, 33 patients with non-metastatic LNs and 29 patients with metastatic LNs (Figure 1).

Figure 1 Flowchart of PTMC enrollment and scheme for analysis. 2D, two-dimensional; PTMC, papillary thyroid microcarcinoma.

The inclusion criteria were as follows: (I) availability of complete clinical and ultrasonographic data; (II) a postoperative pathological diagnosis of PTMC; and (III) performance of central cervical LN dissection.

The exclusion criteria were as follows: (I) failure to complete 2D US examination (76 patients were excluded); (II) failure to complete BRAF^V600E mutation analysis (BRAF^V600E mutation is linked to poor prognosis of thyroid cancer) (62 patients were excluded); (III) preoperative clinical LN positive patients (palpation and imaging examination [US/computed tomography (CT)/magnetic resonance (MR)] showed suspicious malignant LNs) (105 patients were excluded).

US analysis

The Samsung XR80A US machine, with a 3–12 MHz linear probe was applied to 2D US examination. The shape (<1, =1, >1), composition (solid, mixed), echogenicity (very hypoechoic, hypoechoic, isoechoic/hyperechoic), margins (smooth, lobulated/irregular, ill-defined), calcifications (none, microcalcification, macrocalcifications) and other qualitative parameters of thyroid nodules were collected in the transverse plane containing as many malignant features as possible, using TIRADS to stratify patients by risk. According to clinical practice and quality control system, two blinded radiologists, each with more than a decade of experience, independently interpreted all preoperative US features. Any discrepancies between their assessments were resolved by a third senior radiologist, whose decision was final.

Radiomics analysis

2D US of thyroid lesion was manually segmented by two US experts using ITK-SNAP software (http://www.itksnap.org) to generate a 2D contour map (Figure 2). Then the whole tumor image was exported for further radiomics analysis using PyRadiomic open-source software package. We used the inter-group correlation coefficient (ICC) to evaluate the agreement between two observers for the US features. A random sample of 20 lesions from the training cohort was used to evaluate inter-observer agreement in region of interest (ROI) delineation. Two radiologists independently outlined the ROIs, with the second blinded to the first’s results and performing the task one month later. Features with an ICC below 0.8, indicating suboptimal reliability, were discarded. Then, the least absolute shrinkage and selection operator (LASSO) logistic regression with 10-fold cross-validation, implemented by the glmnet package in R software, was applied to select the most significant features to predict the status of CLNs from the training data set (Figure 3). A radiomics score (Rad-score) was computed for each patient via a linear combination of the definitive features and their respective coefficients. The Rad-score is calculating according to the weight of selected features.

Figure 2 Draw ROI and features extraction. Area covered in red denotes the segmented ROI. ROI, region of interest.

Figure 3 Feature selection and radiomics signature construction via LASSO regression. The optimal regularization parameter (λ) was identified through 10-fold cross-validation using the minimum deviance criterion. This λ value was then applied to select the non-zero radiomics features and construct the final radiomics signature (Rad-score). LASSO, least absolute shrinkage and selection operator.

Construction of three models

Model 1 clinical and ultrasonic findings

Univariate analysis was first performed to assess the differences in clinical and ultrasonic features between non-metastatic and metastatic groups. Subsequently, features with a P value of less than 0.10 in the univariate analysis were included in a multivariate logistic regression model using the enter method to construct model 1 in the training set (P<0.10 could prevent the premature exclusion of potentially important variables). During the model fitting process, all the specified variables are incorporated into the regression equation at once and simultaneously. No variable will be automatically excluded or filtered based on its statistical significance.

Model 2 radiomics signature

Model 2 was defined solely by the Rad-score in the training set, which was calculated from the weighted sum of the selected radiomics features.

Model 3 combined clinical, ultrasonographic, and radiomics features

The significant US image features, clinical characteristics, and the Rad-score were incorporated into a multivariate logistic regression analysis, thus forming model 3.

Models performance and validation

The difference of three models was compared by area under the curve (AUC) in receiver operating characteristic (ROC) curve. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were also calculated. Internal and external validation of three established models were performed using independent validation sets (internal and external datasets). A nomogram using the combined model (model 3) was also plotted, in order to provide the clinicians a feasible tool to predict metastatic CLN from PTMC. The discriminative ability of the nomogram was quantified by the AUC of ROC analysis. The calibration of the nomogram was evaluated by plotting the predicted probabilities against the observed outcomes. The clinical net benefit of the three models was compared using decision curve analysis (DCA).

Statistical analysis

All data underwent strict preprocessing before analysis, including handling missing values and detecting outliers. Descriptive statistics for categorical variables were reported as numbers, with group comparisons made by Pearson’s χ² test or Fisher’s exact test, where applicable. Normally distributed continuous variables were summarized as mean ± standard deviation and compared with the independent samples t-test. Significant variables from the univariate analysis were considered as candidates for the multivariate logistic regression model, which was aimed at identifying independent risk factors for CLNM. ROC curves and AUC were compared to evaluate the diagnostic efficacy of diverse models. The statistical software packages SPSS 24.0, MedCalc 19.0.4, and R 4.4.2 (R Project for Statistical Computing, Vienna, Austria, www.r-project.org) were employed for all analyses, with statistical significance set at P<0.05 (two-tailed).

Results

Patient characteristics

A comparison of clinical and US features between non-metastatic and metastatic CLN groups, stratified across the training and validation cohorts, is summarized in Table 1; 144 patients were enrolled in training set, 60 patients were enrolled in internal validation set, and 62 patients were enrolled in external validation set.

Model 1 clinical and ultrasonic features

Model 1, incorporating gender (male or female), extrathyroidal extension (whether the nodule invaded the capsule), echogenic foci (none, microcalcification, macrocalcifications), and TIRADS category (4a–5) (Table 2), achieved an AUC of 0.873 [95% confidence interval (CI): 0.808–0.923] in the training set. Its diagnostic efficacy was further characterized by a sensitivity of 79.2%, specificity of 81.9%, PPV of 81.4%, NPV of 79.7%, and a Youden index of 0.61 (Figure 4A).

Figure 4 ROC curves for the three predictive models. (A) Performance in the training cohort. (B) Performance in the internal validation cohort. (C) Performance in the external validation cohort. The pink dot on each curve indicates the optimal cutoff point based on the Youden index. ROC, receiver operating characteristic.

Model 2 radiomics signature

A total of 9 radiomics features were finally selected to calculate the Rad-score through the LASSO logistic regression analysis (Figure 3 and Table 2). Radiomics score = 0.500+0.015×original_shape2D_Elongation-0.053×original_firstorder_Kurtosis+0.027×original_firstorder_RobustMeanAbsoluteDeviation+0.011×original_glrlm_ShortRunHighGrayLevelEmphasis+0.033×wavelet-HL_glszm_GrayLevelNonUniformity+0.037×wavelet-HL_ngtdm_Contrast+0.009×wavelet-HH_glcm_ClusterTendency+0.031×wavelet-HH_gldm_SmallDependenceLowGrayLevelEmphasis-0.004×wavelet-LL_firstorder_Kurtosis. (Original and wavelet represent different filters, shape2D includes descriptions of the 2D size and shape of the ROI, firstorder describe the distribution of voxel intensities through commonly used and basic metrics, glrlm quantifies gray level runs, glszm quantifies gray level zones in an image, ngtdm quantifies the difference between a gray value and the average gray value of its neighbours within distance, glcm represents the number of times the combination of levels and occur in two pixels in the image, and gldm quantifies gray level dependencies in an image).

Model 2 demonstrated modest diagnostic performance with an AUC of 0.725 (95% CI: 0.644–0.796). Its sensitivity was 62.5%, specificity 75.0%, PPV 71.4%, NPV 66.7%, and the Youden index was 0.38 (Figure 4A).

Model 3 combined clinical, ultrasonographic, and radiomics features

Model 3, comprising three US features, one clinical characteristic, and the Rad-score (Table 2), yielded an optimal AUC of 0.892 (95% CI: 0.829–0.937), a Youden index of 0.65, and sensitivity, specificity, PPV, and NPV of 80.6%, 84.7%, 84.1%, and 81.3%, respectively (Figure 4A).

Validation of three models

Using the optimal cutoff value determined by the maximum Youden index in the training set, we evaluated all models in the validation cohorts. In the internal validation set, the AUCs for model 1, model 2, and model 3 were 0.755 (95% CI: 0.627–0.857), 0.588 (95% CI: 0.453–0.713), and 0.772 (95% CI: 0.646–0.871), respectively (Figure 4B). Model 3 demonstrated a statistically significant superiority over both model 1 (P=0.042) and model 2 (P=0.03). This performance was sustained in the external validation set, where model 3 achieved an AUC of 0.842 (95% CI: 0.727–0.922), significantly outperforming model 2 (AUC =0.753, P=0.02) and showing comparable performance to model 1 (AUC =0.841, P=0.045) (Figure 4C). The comprehensive diagnostic metrics, including sensitivity and specificity, for both validation sets are detailed in Table 3.

Development and validation of a US-based radiomics nomogram

The combined predictors of Model 3 were formulated into a nomogram for clinical use (Figure 5A). The nomogram demonstrated excellent calibration across both the training and validation sets, indicating a high agreement between predicted and observed outcomes (Figure 5B-5D) (training set: mean absolute error =0.025). Its discriminatory ability was excellent, with an AUC of 0.892 (95% CI: 0.829–0.937) in the training set. The superior clinical utility of the nomogram (model 3) was further evidenced by DCA, which showed a consistently higher net benefit than that of the other two models across most reasonable threshold probabilities (Figure 6). The combined model not only possesses outstanding diagnostic efficacy, but also its prediction probabilities can be directly converted into the basis for optimizing clinical decisions, demonstrating a high potential for transformation. This model could assist doctors in avoiding more severe injuries for low-risk patients and ensure that high-risk patients receive timely and adequate treatment by providing quantitative individual risk estimates to help patients understand the pros and cons.

Figure 5 Nomogram and calibration curve for the three research sets. (A) A radiomics nomogram was established incorporating conventional US images and radiomics features in the training set. (B) Calibration curve of radiomics nomogram is shown in the training set. (C) Calibration curve of radiomics nomogram is shown in the internal validation set. (D) Calibration curve of radiomics nomogram is shown in the external validation set. TIRADS, Thyroid Imaging Reporting and Data System; US, ultrasound.

Figure 6 Decision curve analysis and ROC curve for three predictive models. (A) Decision curve analysis for the three established models in the internal validation set. (B) Decision curve analysis for the three established models in the external validation set. (C) The ROC curve of nomogram in the training set. AUC, area under the curve; ROC, receiver operating characteristic.

Discussion

In this study, we developed and validated a nomogram incorporating the radiomics signature, clinical and US features to provide a noninvasive and personalized preoperative risk assessment for CLNM in cN0 PTMC patients. The combined model 3 proved to be a more robust predictor than models using either radiomics or US imaging features alone, showing a better AUC compared with previous studies, considering different populations (25,26). Li et al. found that the radiomics analysis achieved a high sensitivity of 0.90 (25). One study reported that a radiomics model based on dual-plane US images (combining longitudinal and transverse sections) demonstrated significantly higher specificity (0.85) compared to models using a single plane (0.64 and 0.56, respectively) (26). These results indicated that radiomics analysis could serve as an effective and standardized complementary to predict CLN status for patients with PTMC accurately. If the nomogram based on US radiomics is applied clinically, it can contribute to avoiding unnecessary prophylactic central cervical LN dissection. In our study, suspicious ultrasonic features including echogenic foci, extra extension, and TIRADS category were confirmed as prognostic features in PTMC. Likewise, evidence from previous studies has connected certain preoperative US features, including microcalcification, with CLNM (25,27,28). A previous study found that rapid growth of cancer cell accompanied by hyperplasia of blood vessels and fibrous, resulted in calcium salt deposition, that is, microcalcification (29). Thus, if microcalcification is found in the thyroid nodules, the central cervical LN should be scanned more carefully. Yan et al. found that external invasion was an independent risk factor for cN0 PTC patients (30). Once tumor cell invades the thyroid capsule, it is possible to disseminate the regional LNs along the rich lymphatic tissue, which is associated with a high risk of recurrence and poorer survival (16,31). Previous studies showed higher TIRADS score was markedly independently associated with LNM in the cervical central region, which meant a higher degree of malignancy (32,33). Nevertheless, the accurate identification of these US characteristics is highly contingent on the operator’s expertise, which can be a limiting factor in diagnostic reproducibility.

Recently, radiomics analysis has been wildly used in diagnosing malignant lesions and predicting the status of cervical LNM and could be a very useful marker. A nomogram based on US radiomics signature reached AUC of 0.946 and C-index of 0.947 for prediction of cervical LNM in PTC, which was significantly associated with the lateral LNM in both cohorts (training and validation sets) (23). To perform radiomics analysis, Zhou et al. utilized quantitative imaging biomarkers extracted from dual-energy CT-derived iodine maps, achieving a favorable diagnostic value for metastatic cervical LNs in PTC (34). They found that metastatic LNs were characterized by significantly greater tissue heterogeneity than their non-metastatic counterparts. The thyroid malignant nodule with metastatic central cervical LN postoperative pathology confirmed indicated lower uniformity, which was in accordance with the perspective of Cao et al. (35,36), reflecting the degree of intratumoral homogeneity (22). A nomogram established by Jiang et al., which utilized shear-wave elastography (SWE) radiomics for predicting metastatic cervical LN in PTC, achieved favorable predictive performance (21). SWE is an elastographic technique used to evaluate the tissue hardness quantitatively. Multi-modality US images could reveal diverse information about the tissue, which could be applied in more fields in the future.

Our integrated model, which combined radiomic, clinical, and ultrasonographic features, demonstrated robust and generalizable discriminatory power for predicting CLN status in cN0 PTMC patients, with AUCs of 0.892 (training), 0.772 (internal validation), and 0.842 (external validation), further affirming its excellent consistency and reproducibility. Our analysis revealed that female had a higher likelihood of CLNM in PTMC. And several previous studies showed that female sex was an independent risk factor for skip metastases (37,38). Thyroid cancer exhibits a strong female predilection, demonstrating a remarkably consistent sex distribution of 3:1 across most geographic regions and demographic groups (39). Therefore, female patients with PTMC should be managed more cautiously.

This study is subject to several limitations. Firstly, the retrospective nature of our work introduces the potential for selection bias. Furthermore, the cohort consisted of a relatively small number of patients who underwent surgery at two tertiary centers, which may limit generalizability. External validation in a larger, multi-center population is needed to confirm our findings. Secondly, this study did not establish the histopathologic correlates of the identified radiomics signatures, which represents a valuable direction for future research to elucidate their underlying biological basis. Thirdly, this study limited the same US machine from different institutions, which indicated that more types of US machines needed to be applied to collect images. Lastly, the long-term impact of CLNM on patient outcomes, including disease-specific survival and recurrence, requires further investigation through prospective studies.

Conclusions

The nomogram that integrates radiomics analysis with clinical and US features significantly improves preoperative identification of metastatic central cervical LNs in cN0 PTMC patients, offering a robust tool to refine treatment planning.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://gs.amegroups.com/article/view/10.21037/gs-2025-aw-512/rc

Data Sharing Statement: Available at https://gs.amegroups.com/article/view/10.21037/gs-2025-aw-512/dss

Peer Review File: Available at https://gs.amegroups.com/article/view/10.21037/gs-2025-aw-512/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://gs.amegroups.com/article/view/10.21037/gs-2025-aw-512/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This retrospective study was approved by the Institutional Ethics Committee of The First Affiliated Hospital of Nanjing Medical University (No. 2022-SR-512) and written informed consent was collected from all the patients.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Brito JP, Hay ID. Management of Papillary Thyroid Microcarcinoma. Endocrinol Metab Clin North Am 2019;48:199-213. [Crossref] [PubMed]
2. Li M, Dal Maso L, Pizzato M, et al. Evolving epidemiological patterns of thyroid cancer and estimates of overdiagnosis in 2013-17 in 63 countries worldwide: a population-based study. Lancet Diabetes Endocrinol 2024;12:824-36. [Crossref] [PubMed]
3. Tagliabue M, Giugliano G, Mariani MC, et al. Prevalence of Central Compartment Lymph Node Metastases in Papillary Thyroid Micro-Carcinoma: A Retrospective Evaluation of Predictive Preoperative Features. Cancers (Basel) 2021;13:6028. [Crossref] [PubMed]
4. Ito Y, Uruno T, Nakano K, et al. An observation trial without surgical treatment in patients with papillary microcarcinoma of the thyroid. Thyroid 2003;13:381-7. [Crossref] [PubMed]
5. Kutler DI, Crummey AD, Kuhel WI. Routine central compartment lymph node dissection for patients with papillary thyroid carcinoma. Head Neck 2012;34:260-3. [Crossref] [PubMed]
6. Podnos YD, Smith D, Wagman LD, et al. The implication of lymph node metastasis on survival in patients with well-differentiated thyroid cancer. Am Surg 2005;71:731-4. [Crossref] [PubMed]
7. American Thyroid Association Surgery Working Group, American Association of Endocrine Surgeons, American Academy of Otolaryngology-Head and Neck Surgery, et al. Consensus statement on the terminology and classification of central neck dissection for thyroid cancer. Thyroid 2009;19:1153-8. [Crossref] [PubMed]
8. Liu LS, Liang J, Li JH, et al. The incidence and risk factors for central lymph node metastasis in cN0 papillary thyroid microcarcinoma: a meta-analysis. Eur Arch Otorhinolaryngol 2017;274:1327-38. [Crossref] [PubMed]
9. Liu J, Fan XF, Yang M, et al. Analysis of the risk factors for central lymph-node metastasis of cN0 papillary thyroid microcarcinoma: A retrospective study. Asian J Surg 2022;45:1525-9. [Crossref] [PubMed]
10. Xiang Y, Lin K, Dong S, et al. Prediction of central lymph node metastasis in 392 patients with cervical lymph node-negative papillary thyroid carcinoma in Eastern China. Oncol Lett 2015;10:2559-64. [Crossref] [PubMed]
11. Zhao W, You L, Hou X, et al. The Effect of Prophylactic Central Neck Dissection on Locoregional Recurrence in Papillary Thyroid Cancer After Total Thyroidectomy: A Systematic Review and Meta-Analysis : pCND for the Locoregional Recurrence of Papillary Thyroid Cancer. Ann Surg Oncol 2017;24:2189-98. [Crossref] [PubMed]
12. McHenry CR, Stulberg JJ. Prophylactic central compartment neck dissection for papillary thyroid cancer. Surg Clin North Am 2014;94:529-40. [Crossref] [PubMed]
13. Chisholm EJ, Kulinskaya E, Tolley NS. Systematic review and meta-analysis of the adverse effects of thyroidectomy combined with central neck dissection as compared with thyroidectomy alone. Laryngoscope 2009;119:1135-9. [Crossref] [PubMed]
14. Hughes DT, White ML, Miller BS, et al. Influence of prophylactic central lymph node dissection on postoperative thyroglobulin levels and radioiodine treatment in papillary thyroid cancer. Surgery 2010;148:1100-6; discussion 1006-7. [Crossref] [PubMed]
15. Jin SH, Kim IS, Ji YB, et al. Efficacy of prophylactic central neck dissection in hemithyroidectomy for papillary thyroid carcinoma. Eur Arch Otorhinolaryngol 2020;277:873-9. [Crossref] [PubMed]
16. Ringel MD, Sosa JA, Baloch Z, et al. 2025 American Thyroid Association Management Guidelines for Adult Patients with Differentiated Thyroid Cancer. Thyroid 2025;35:841-985. [Crossref] [PubMed]
17. Kwak JY, Han KH, Yoon JH, et al. Thyroid imaging reporting and data system for US features of nodules: a step in establishing better stratification of cancer risk. Radiology 2011;260:892-9. [Crossref] [PubMed]
18. Zhao H, Li H. Meta-analysis of ultrasound for cervical lymph nodes in papillary thyroid cancer: Diagnosis of central and lateral compartment nodal metastases. Eur J Radiol 2019;112:14-21. [Crossref] [PubMed]
19. Kim E, Park JS, Son KR, et al. Preoperative diagnosis of cervical metastatic lymph nodes in papillary thyroid carcinoma: comparison of ultrasound, computed tomography, and combined ultrasound with computed tomography. Thyroid 2008;18:411-8. [Crossref] [PubMed]
20. Bi WL, Hosny A, Schabath MB, et al. Artificial intelligence in cancer imaging: Clinical challenges and applications. CA Cancer J Clin 2019;69:127-57. [Crossref] [PubMed]
21. Jiang M, Li C, Tang S, et al. Nomogram Based on Shear-Wave Elastography Radiomics Can Improve Preoperative Cervical Lymph Node Staging for Papillary Thyroid Carcinoma. Thyroid 2020;30:885-97. [Crossref] [PubMed]
22. Gillies RJ, Kinahan PE, Hricak H. Radiomics: Images Are More than Pictures, They Are Data. Radiology 2016;278:563-77. [Crossref] [PubMed]
23. Tong Y, Li J, Huang Y, et al. Ultrasound-Based Radiomic Nomogram for Predicting Lateral Cervical Lymph Node Metastasis in Papillary Thyroid Carcinoma. Acad Radiol 2021;28:1675-84. [Crossref] [PubMed]
24. Kwak JY, Jung I, Baek JH, et al. Image reporting and characterization system for ultrasound features of thyroid nodules: multicentric Korean retrospective study. Korean J Radiol 2013;14:110-7. [Crossref] [PubMed]
25. Li F, Pan D, He Y, et al. Using ultrasound features and radiomics analysis to predict lymph node metastasis in patients with thyroid cancer. BMC Surg 2020;20:315. [Crossref] [PubMed]
26. Wen Q, Wang Z, Traverso A, et al. A radiomics nomogram for the ultrasound-based evaluation of central cervical lymph node metastasis in papillary thyroid carcinoma. Front Endocrinol (Lausanne) 2022;13:1064434. [Crossref] [PubMed]
27. Gao X, Luo W, He L, et al. Predictors and a Prediction Model for Central Cervical Lymph Node Metastasis in Papillary Thyroid Carcinoma (cN0). Front Endocrinol (Lausanne) 2021;12:789310. [Crossref] [PubMed]
28. Iannuccilli JD, Cronan JJ, Monchik JM. Risk for malignancy of thyroid nodules as assessed by sonographic criteria: the need for biopsy. J Ultrasound Med 2004;23:1455-64. [Crossref] [PubMed]
29. Liu C, Xiao C, Chen J, et al. Risk factor analysis for predicting cervical lymph node metastasis in papillary thyroid carcinoma: a study of 966 patients. BMC Cancer 2019;19:622. [Crossref] [PubMed]
30. Yan H, Zhou X, Jin H, et al. A Study on Central Lymph Node Metastasis in 543 cN0 Papillary Thyroid Carcinoma Patients. Int J Endocrinol 2016;2016:1878194. [Crossref] [PubMed]
31. Amin MB, Greene FL, Edge SB, et al. The Eighth Edition AJCC Cancer Staging Manual: Continuing to build a bridge from a population-based to a more "personalized" approach to cancer staging. CA Cancer J Clin 2017;67:93-9.
32. Yang JR, Song Y, Chang SJ, et al. Prediction of central compartment nodal metastases in papillary thyroid cancer using TI-RADS score, blood flow, and multifocality. Acta Radiol 2022;63:1374-80. [Crossref] [PubMed]
33. Shen H, Lv G, Li T, et al. Construction and Validation of a Predictive Nomogram Based on Ultrasound for Lymph Node Metastasis of Papillary Thyroid Carcinoma in the Cervical Central Region. Ultrasound Q 2023;39:47-52. [Crossref] [PubMed]
34. Zhou Y, Su GY, Hu H, et al. Radiomics analysis of dual-energy CT-derived iodine maps for diagnosing metastatic cervical lymph nodes in patients with papillary thyroid cancer. Eur Radiol 2020;30:6251-62. [Crossref] [PubMed]
35. Cao Y, Zhong X, Diao W, et al. Radiomics in Differentiated Thyroid Cancer and Nodules: Explorations, Application, and Limitations. Cancers (Basel) 2021;13:2436. [Crossref] [PubMed]
36. Liu T, Zhou S, Yu J, et al. Prediction of Lymph Node Metastasis in Patients With Papillary Thyroid Carcinoma: A Radiomics Method Based on Preoperative Ultrasound Images. Technol Cancer Res Treat 2019;18:1533033819831713. [Crossref] [PubMed]
37. Zhao H, Huang T, Li H. Risk factors for skip metastasis and lateral lymph node metastasis of papillary thyroid cancer. Surgery 2019;166:55-60. [Crossref] [PubMed]
38. Sapuppo G, Tavarelli M, Russo M, et al. Lymph node location is a risk factor for papillary thyroid cancer-related death. J Endocrinol Invest 2018;41:1349-53. [Crossref] [PubMed]
39. Kilfoy BA, Zheng T, Holford TR, et al. International patterns and trends in thyroid cancer incidence, 1973-2002. Cancer Causes Control 2009;20:525-31. [Crossref] [PubMed]