To address the high computational complexity of the Transformer in the segmentation of ultrasound thyroid nodules and the loss of image details or omission of key spatial information caused by traditional image sampling techniques when dealing with high-resolution, complex texture or uneven density two-dimensional ultrasound images, this paper proposes a thyroid nodule segmentation method that integrates the receiving weighted key-value (RWKV) architecture and spherical geometry feature (SGF) sampling technology. This method effectively captures the details of adjacent regions through two-dimensional offset prediction and pixel-level sampling position adjustment, achieving precise segmentation. Additionally, this study introduces a patch attention module (PAM) to optimize the decoder feature map using a regional cross-attention mechanism, enabling it to focus more precisely on the high-resolution features of the encoder. Experiments on the thyroid nodule segmentation dataset (TN3K) and the digital database for thyroid images (DDTI) show that the proposed method achieves dice similarity coefficients (DSC) of 87.24% and 80.79% respectively, outperforming existing models while maintaining a lower computational complexity. This approach may provide an efficient solution for the precise segmentation of thyroid nodules.
Thyroid Nodule/diagnostic imaging* ; Humans ; Ultrasonography/methods* ; Algorithms ; Image Processing, Computer-Assisted/methods* ; Thyroid Gland/diagnostic imaging*

Thyroid nodules are a common endocrine disorder, and their early detection and accurate diagnosis are crucial for the prevention of thyroid cancer. However, the highly heterogeneous morphology and boundaries of thyroid nodules pose significant challenges to their precise identification and classification. Traditional diagnostic approaches rely heavily on physicians' experience, which increases the risk of misdiagnosis and missed diagnoses. With the rapid advancement of computer-aided diagnosis (CAD) technologies, applying deep learning algorithms to the analysis of thyroid nodule ultrasound images has shown great potential. This paper reviews the latest research progress on deep learning-based CAD methods for thyroid nodules, with a focus on their applications in image preprocessing, segmentation and classification. The advantages and limitations of current techniques are analyzed, and potential future directions are discussed. This review aims to highlight the potential of deep learning in thyroid nodule diagnosis and to provide a foundation for selecting feasible pathways for future clinical applications.
Humans ; Thyroid Nodule/diagnostic imaging* ; Deep Learning ; Ultrasonography/methods* ; Diagnosis, Computer-Assisted/methods* ; Algorithms ; Thyroid Neoplasms/diagnostic imaging* ; Image Processing, Computer-Assisted/methods*

Objective To analyze the cellular and histopathological characteristics of underdiagnosed thyroid nodules of Chinese thyroid imaging reporting and data system(C-TIRADS) categories 3 and 4a,thus improving the understanding of these lesions. Methods The data of ultrasound and fine needle aspiration cytology were collected from 683 nodules diagnosed based on pathological evidence in 549 patients undergoing thyroid surgery.The cellular and histopathological characteristics of C-TIRADS 3 and 4a nodules were analyzed. Results Two hundred and sixty-eight nodules were classified as C-TIRADS category 3,including 236 benign nodules,12 low-risk ones,and 20 (7.46%) malignant ones.Two hundred and twenty-one nodules were classified as C-TIRADS category 4a,including 133 benign nodules,7 low-risk ones,and 81 (36.65%) malignant ones.The malignancy rates differed between C-TIRADS 3 and 4a nodules (χ²=58.93,P<0.001),and both were higher than the recommended malignancy rate in the guidelines for malignancy risk stratification of thyroid nodules (C-TIRADS) (both P<0.001).According to the pathological evidence,the underdiagnosed C-TIRADS 3/4a nodules were mainly papillary thyroid carcinoma,especially in patients with Hashimoto thyroiditis.There was not a consistent one-to-one match between each ultrasound result and each cytological classification of low-risk thyroid nodules.Conclusions When the malignant features in preoprative ultrasound imaging are atypical or absent,papillary thyroid carcinoma (especially with Hashimoto thyroiditis),follicular carcinoma,and medullary carcinoma are likely to be underdiagnosed as C-TIRADS 3 or 4a nodules.Therefore,efforts should be made to fully understand the cellular and pathological characteristics of these lesions.
Humans ; Thyroid Nodule/diagnostic imaging* ; Female ; Male ; Middle Aged ; Adult ; Ultrasonography ; Biopsy, Fine-Needle ; Aged ; Young Adult ; Thyroid Neoplasms/diagnostic imaging* ; Adolescent

Objective To explore the research progress and hotspots of intelligent ultrasound imaging in the diagnosis of thyroid nodules and clarify the research directions via the bibliometric method.Methods The relevant research articles on intelligent ultrasound imaging in the diagnosis of thyroid nodules were retrieved from the Web of Science Core Collection,covering the period from January 2004 to August 2024.Python was used to analyze the number of annual publications.VOSviewer was used to create the co-occurrence network of authors and the keyword density map.CiteSpace was used to demonstrate the dual-map overlays of the journals,as well as the bursts and clustering of co-citations and keywords.Results A total of 1 179 articles were included.The annual number of publications increased steadily.The involved journals demonstrated high quality,and the publications showed a trend of cross-research.Chinese researchers were the core research force in this field.Haugen et al.'s study on the guidelines for thyroid nodules had the most citations.The clustering of co-citations and keywords indicated studies in multiple fields.Thyroid nodules,cancer,and deep learning were the representative keywords in this field.Conclusions The continuous enrichment of research topics promotes the rapid development of intelligent ultrasound imaging for thyroid nodules.Intelligent diagnosis methods based on deep learning can provide diagnostic suggestions,while there are still challenges such as interpretation.One of the research directions is the deep combination of intelligent diagnosis algorithms and medical knowledge.
Thyroid Nodule/diagnostic imaging* ; Humans ; Ultrasonography ; Bibliometrics

Objective To investigate the influencing factors of Bethesda Ⅲ results in fine-needle aspiration biopsy of thyroid nodules.Methods A total of 300 thyroid nodules with cytological diagnosis results were analyzed retrospectively,including 100 Bethesda Ⅲ nodules and 50 nodules of Bethesda Ⅱ,Ⅳ,Ⅴ,and Ⅵ categories,respectively.Univariate analysis and Logistic regression analysis were performed on the clinical data of patients and the ultrasound signs of thyroid nodules to clarify the factors influencing the diagnosis of Bethesda Ⅲ nodules.Results Univariate analysis showed that Bethesda Ⅲ nodules were mostly adjacent to the capsule(P<0.001),with no blood flow in the color Doppler assessment(P=0.011)and lack of blood supply(P=0.033)and maximum diameter ≤0.9 cm(P=0.038)as revealed by the contrast-enhanced ultrasound.Logistic regression showed that the position close to the capsule(OR=5.110,95%CI=2.153-12.130,P<0.001)and color Doppler without blood flow signal(OR=3.015,95%CI=1.094-8.311,P=0.033)were independent risk factors for the diagnosis of Bethesda Ⅲ nodules.Conclusions The puncture difficulty caused by the dangerous position of thyroid nodules close to the capsule and the aspiration difficulty caused by the absence of blood flow signal in color Doppler are the main factors influencing the diagnosis of Bethesda Ⅲ nodules.Therefore,corresponding avoidance measures should be taken during the aspiration process to reduce the diagnosis results of Bethesda Ⅲ nodules.
Humans ; Thyroid Nodule/diagnostic imaging* ; Thyroid Neoplasms/diagnosis* ; Biopsy, Fine-Needle/methods* ; Retrospective Studies ; Ultrasonography/methods*

Objective To investigate the influencing factors and establish a model predicting the performance of needle visualization in fine-needle aspiration (FNA) of thyroid nodules. Methods This study prospectively included 175 patients who underwent FNA of thyroid nodules in the Department of Ultrasound in China-Japan Friendship Hospital and compared the display of the needle tips in the examination of 199 thyroid nodules before and after the application of needle visualization.We recorded the location,the positional relationship with thyroid capsule,ultrasonic characteristics,and the distribution of the soft tissue strip structure at the puncture site of the nodules with unclear needle tips display before using needle visualization.Furthermore,according to the thyroid imaging reporting and data system proposed by the American College of Radiology,we graded the risk of the nodules.Lasso-Logistic regression was employed to screen out the factors influencing the performance of needle visualization and establish a nomogram for prediction. Results The needle tips were not clearly displayed in the examination of 135 (67.8%) and 53 (26.6%) nodules before and after the application of needle visualization,respectively,which showed a significant difference (P<0.001).Based on the positional relationship between the nodule and capsule,anteroposterior/transverse diameter (A/T) ratio,blood supply,and the distribution of subcutaneous strip structure at the puncture site,a nomogram was established to predict the probability of unclear display of the needle tips after application of needle visualization.The C-index of the prediction model was 0.75 (95%CI=0.67-0.84) and the area under the receiver operating characteristic curve was 0.72.The calibration curve confirmed the appreciable reliability of the prediction model,with the C-index of 0.70 in internal validation. Conclusions Needle visualization can improve the display of the needle tip in ultrasound-guided FNA of thyroid nodules.The nomogram established based on ultrasound features such as the positional relationship between the nodule and capsule,A/T ratio,blood supply,and the distribution of subcutaneous strip structure at the puncture site can predict whether needle visualization is suitable for the examination of nodules.
Humans ; Thyroid Nodule/diagnostic imaging* ; Biopsy, Fine-Needle/methods* ; Reproducibility of Results ; Ultrasonography ; Retrospective Studies ; Thyroid Neoplasms

Objective To evaluate the performance of micro-flow imaging(MFI)in the differential diagnosis of benign and malignant thyroid nodules. Methods Totally 50 patients with thyroid nodules examined by conventional ultrasound,MFI,and contrast-enhanced ultrasound and confirmed by histological or cytological pathology in the First Medical Center of Chinese PLA General Hospital from May to December in 2020 were enrolled in the study.The clinical data and ultrasound images were retrospectively analyzed.A binary logistic regression model was established to evaluate the performance of the model in predicting benign and malignant thyroid nodules. Results Logistic regression showed that composition and "S-W-C" sign were independent risk factors for predicting malignant thyroid nodule.The sensitivity,specificity,and Youden index of the logistic regression model were 73.33%,80.00%,and 0.53,respectively,and the area under receiver operating characteristic curve was 0.799(95%CI=0.662-0.899). Conclusion MFI facilitates the differential diagnosis of benign and malignant thyroid nodules and has the potential to be applied in the future.
Diagnosis, Differential ; Humans ; ROC Curve ; Retrospective Studies ; Thyroid Nodule/diagnostic imaging* ; Ultrasonography/methods*

China ; Data Systems ; Humans ; Radiology ; Retrospective Studies ; Thyroid Neoplasms/diagnostic imaging* ; Thyroid Nodule ; Ultrasonography/methods* ; United States

Objective To explore the roles of conventional ultrasound and contrast-enhanced ultrasound in distinguishing between benign and malignant thyroid nodules with calcification. Methods A total of 102 solid thyroid nodules with calcification in 76 patients were evaluated by conventional ultrasound alone and conventional ultrasound combined with contrast-enhanced ultrasound.The features obtained through conventional ultrasound alone and that combined with contrast-enhanced ultrasound were scored,and the diagnostic performance of the two methods was analyzed based on the final pathological results. Results The distribution of microcalcification(
Calcinosis/diagnostic imaging* ; Contrast Media ; Diagnosis, Differential ; Humans ; Sensitivity and Specificity ; Thyroid Neoplasms ; Thyroid Nodule/diagnostic imaging* ; Ultrasonography

Objective: To analyze the features of degenerating cystic thyroid nodules (DCTN) on conventional ultrasound and contrast-enhanced ultrasound (CEUS), and to explore the differentiation between DCTN and papillary thyroid carcinomas (PTC). Methods: A total of 46 DCTN (39 cases, including 12 males and 27 females, with an age range of 25 to 76 years) and 36 PTC (32 cases, including 8 males and 24 females, with an age range of 23 to 68 years) diagnosed via fine- needle aspiration (FNA) or surgery from February 2019 to January 2020 in the First Affiliated Hospital of Nanchang University were enrolled. The size, shape, margin, echogenicity, presence of shadowing, calcification and vascularity of DCTN and PTC were retrospectively evaluated, and 28 DCTN and 30 PTC underwent CEUS were separately analyzed and compared.The t test, χ² test or Fisher's exact test were implemented to compare the features of ultrasound among the two groups. The binary Logistic regression test was performed to determine whether the feature whose difference was statistically significant was an independent predictive risk factor. Results: A univariate analysis indicated that DCTN more frequently showed wider-than-tall shapes, marked hypoechogenicity, well-defined margin and no or dot-lined enhancement (wider-than-tall shapes: 36 vs. 17, χ²=8.511; well-defined margin: 30 vs. 15, χ²=4.523; marked hypoechogenicity: 27 vs. 9, χ²=9.310; no or dot-lined enhancement: 24 vs. 3, χ²=33.369; all P＜0.05). A multivariate analysis demonstrated that wider-than-tall shapes, well-defined margin and marked hypoechogenicity were independent predictors for DCTN (OR values were 5.204, 3.134 and 5.042, P values were 0.003, 0.031, and 0.003, respectively). Among 28 DCTN, 15 showed a decrease in mean maximum diameter (24.3±11.4 mm) with a mean time span of (18.6±10.5) months between the presence and absence of suspicious ultrasound features. Conclusions: Compared with PTC, DCTN shows the ultrasound characteristics of wider-than-tall shapes, well-defined margin, marked hypoechogenicity and no or dot-lined enhancement pattern. Ultrasound follow-up can help to identify spontaneous DCTN.
Adult ; Aged ; Carcinoma, Papillary/diagnostic imaging* ; Female ; Humans ; Male ; Middle Aged ; Retrospective Studies ; Thyroid Cancer, Papillary/diagnostic imaging* ; Thyroid Neoplasms/diagnostic imaging* ; Thyroid Nodule/diagnostic imaging* ; Ultrasonography