Ultrasound Viscosity Imaging in Breast Lesions: A Multicenter Prospective Study.

RATIONALE AND OBJECTIVES: To explore and validate the clinical value of ultrasound (US) viscosity imaging in differentiating breast lesions by combining with BI-RADS, and then comparing the diagnostic performances with BI-RADS alone. MATERIALS AND METHODS: This multicenter, prospective study enrolled participants with breast lesions from June 2021 to November 2022. A development cohort (DC) and validation cohort (VC) were established. Using histological results as reference standard, the viscosity-related parameter with the highest area under the receiver operating curve (AUC) was selected as the optimal one. Then the original BI-RADS would upgrade or not based on the value of this parameter. Finally, the results were validated in the VC and total cohorts. In the DC, VC and total cohorts, all breast lesions were divided into the large lesion, small lesion and overall groups respectively. RESULTS: A total of 639 participants (mean age, 46 years ± 14) with 639 breast lesions (372 benign and 267 malignant lesions) were finally enrolled in this study including 392 participants in the DC and 247 in the VC. In the DC, the optimal viscosity-related parameter in differentiating breast lesions was calculated to be A'-S2-Vmax, with the AUC of 0.88 (95% CI: 0.84, 0.91). Using > 9.97 Pa.s as the cutoff value, the BI-RADS was then modified. The AUC of modified BI-RADS significantly increased from 0.85 (95% CI: 0.81, 0.88) to 0.91 (95% CI: 0.87, 0.93), 0.85 (95% CI: 0.80, 0.89) to 0.90 (95% CI: 0.85, 0.93) and 0.85 (95% CI: 0.82, 0.87) to 0.90 (95% CI: 0.88, 0.92) in the DC, VC and total cohorts respectively (P < .05 for all). CONCLUSION: The quantitative viscous parameters evaluated by US viscosity imaging contribute to breast cancer diagnosis when combined with BI-RADS.

An efficient framework for lesion segmentation in ultrasound images using global adversarial learning and region-invariant loss.

Lesion segmentation in ultrasound images is an essential yet challenging step for early evaluation and diagnosis of cancers. In recent years, many automatic CNN-based methods have been proposed to assist this task. However, most modern approaches often lack capturing long-range dependencies and prior information making it difficult to identify the lesions with unfixed shapes, sizes, locations, and textures. To address this, we present a novel lesion segmentation framework that guides the model to learn the global information about lesion characteristics and invariant features (e.g., morphological features) of lesions to improve the segmentation in ultrasound images. Specifically, the segmentation model is guided to learn the characteristics of lesions from the global maps using an adversarial learning scheme with a self-attention-based discriminator. We argue that under such a lesion characteristics-based guidance mechanism, the segmentation model gets more clues about the boundaries, shapes, sizes, and positions of lesions and can produce reliable predictions. In addition, as ultrasound lesions have different textures, we embed this prior knowledge into a novel region-invariant loss to constrain the model to focus on invariant features for robust segmentation. We demonstrate our method on one in-house breast ultrasound (BUS) dataset and two public datasets (i.e., breast lesion (BUS B) and thyroid nodule from TNSCUI2020). Experimental results show that our method is specifically suitable for lesion segmentation in ultrasound images and can outperform the state-of-the-art approaches with Dice of 0.931, 0.906, and 0.876, respectively. The proposed method demonstrates that it can provide more important information about the characteristics of lesions for lesion segmentation in ultrasound images, especially for lesions with irregular shapes and small sizes. It can assist the current lesion segmentation models to better suit clinical needs.

Thyroid ultrasound diagnosis improvement via multi-view self-supervised learning and two-stage pre-training.

Thyroid nodule classification and segmentation in ultrasound images are crucial for computer-aided diagnosis; however, they face limitations owing to insufficient labeled data. In this study, we proposed a multi-view contrastive self-supervised method to improve thyroid nodule classification and segmentation performance with limited manual labels. Our method aligns the transverse and longitudinal views of the same nodule, thereby enabling the model to focus more on the nodule area. We designed an adaptive loss function that eliminates the limitations of the paired data. Additionally, we adopted a two-stage pre-training to exploit the pre-training on ImageNet and thyroid ultrasound images. Extensive experiments were conducted on a large-scale dataset collected from multiple centers. The results showed that the proposed method significantly improves nodule classification and segmentation performance with limited manual labels and outperforms state-of-the-art self-supervised methods. The two-stage pre-training also significantly exceeded ImageNet pre-training.

Machine Learning-Based Breast Tumor Ultrasound Radiomics for Pre-operative Prediction of Axillary Sentinel Lymph Node Metastasis Burden in Early-Stage Invasive Breast Cancer.

OBJECTIVE: The aim of the work described here was to assess the application of ultrasound (US) radiomics with machine learning (ML) classifiers to the prediction of axillary sentinel lymph node metastasis (SLNM) burden in early-stage invasive breast cancer (IBC). METHODS: In this study, 278 early-stage IBC patients with at least one SLNM (195 in the training set and 83 in the test set) were studied at our institution. Pathologic SLNM burden was used as the reference standard. The US radiomics features of breast tumors were extracted by using 3D-Slicer and PyRadiomics software. Four ML classifiers-linear discriminant analysis (LDA), support vector machine (SVM), random forest (RF) and decision tree (DT)-were used to construct radiomics models for the prediction of SLNM burden. The combined clinicopathologic-radiomics models were also assessed with respect to sensitivity, specificity, accuracy and areas under the curve (AUCs). RESULTS: Among the US radiomics models, the SVM classifier achieved better predictive performance with an AUC of 0.920 compared with RF (AUC = 0.874), LDA (AUC = 0.835) and DT (AUC = 0.800) in the test set. The clinicopathologic model had low efficacy, with AUCs of 0.678 and 0.710 in the training and test sets, respectively. The combined clinicopathologic (C) factors and SVM classifier (C + SVM) model improved the predictive ability with an AUC of 0.934, sensitivity of 86.7%, specificity of 89.9% and accuracy of 91.0% in the test set. CONCLUSION: ML-based US radiomics analysis, as a novel and promising predictive tool, is conducive to a precise clinical treatment strategy.

Generalizability and Diagnostic Performance of AI Models for Thyroid US.

Background Artificial intelligence (AI) models have improved US assessment of thyroid nodules; however, the lack of generalizability limits the application of these models. Purpose To develop AI models for segmentation and classification of thyroid nodules in US using diverse data sets from nationwide hospitals and multiple vendors, and to measure the impact of the AI models on diagnostic performance. Materials and Methods This retrospective study included consecutive patients with pathologically confirmed thyroid nodules who underwent US using equipment from 12 vendors at 208 hospitals across China from November 2017 to January 2019. The detection, segmentation, and classification models were developed based on the subset or complete set of images. Model performance was evaluated by precision and recall, Dice coefficient, and area under the receiver operating characteristic curve (AUC) analyses. Three scenarios (diagnosis without AI assistance, with freestyle AI assistance, and with rule-based AI assistance) were compared with three senior and three junior radiologists to optimize incorporation of AI into clinical practice. Results A total of 10 023 patients (median age, 46 years [IQR 37-55 years]; 7669 female) were included. The detection, segmentation, and classification models had an average precision, Dice coefficient, and AUC of 0.98 (95% CI: 0.96, 0.99), 0.86 (95% CI: 0.86, 0.87), and 0.90 (95% CI: 0.88, 0.92), respectively. The segmentation model trained on the nationwide data and classification model trained on the mixed vendor data exhibited the best performance, with a Dice coefficient of 0.91 (95% CI: 0.90, 0.91) and AUC of 0.98 (95% CI: 0.97, 1.00), respectively. The AI model outperformed all senior and junior radiologists (P < .05 for all comparisons), and the diagnostic accuracies of all radiologists were improved (P < .05 for all comparisons) with rule-based AI assistance. Conclusion Thyroid US AI models developed from diverse data sets had high diagnostic performance among the Chinese population. Rule-based AI assistance improved the performance of radiologists in thyroid cancer diagnosis. © RSNA, 2023 Supplemental material is available for this article.

Markedly hypoechoic: a new definition improves the diagnostic performance of thyroid ultrasound.

OBJECTIVES: To determine the contribution of a modified definition of markedly hypoechoic in the differential diagnosis of thyroid nodules. METHODS: A total of 1031 thyroid nodules were included in this retrospective multicenter study. All of the nodules were examined with US before surgery. The US features of the nodules were evaluated, in particular, the classical markedly hypoechoic and modified markedly hypoechoic (decreased or similar echogenicity relative to the adjacent strap muscles). The sensitivity, specificity, and AUC of classical/modified markedly hypoechoic and the corresponding ACR-TIRADS, EU-TIRADS, and C-TIRADS categories were calculated and compared. The inter- and intraobserver variability in the evaluation of the main US features of the nodules was assessed. RESULTS: There were 264 malignant nodules and 767 benign nodules. Compared with classical markedly hypoechoic as a diagnostic criterion for malignancy, using modified markedly hypoechoic as the criterion resulted in a significant increase in sensitivity (28.03% vs. 63.26%) and AUC (0.598 vs. 0.741), despite a significant decrease in specificity (91.53% vs. 84.88%) (p < 0.001 for all). Compared to the AUC of the C-TIRADS with the classical markedly hypoechoic, the AUC of the C-TIRADS with the modified markedly hypoechoic increased from 0.878 to 0.888 (p = 0.01); however, the AUCs of the ACR-TIRADS and EU-TIRADS did not change significantly (p > 0.05 for both). There was substantial interobserver agreement (κ = 0.624) and perfect intraobserver agreement (κ = 0.828) for the modified markedly hypoechoic. CONCLUSION: The modified definition of markedly hypoechoic resulted in a significantly improved diagnostic efficacy in determining malignant thyroid nodules and may improve the diagnostic performance of the C-TIRADS. CLINICAL RELEVANCE STATEMENT: Our study found that, compared with the original definition, modified markedly hypoechoic significantly improved the diagnostic performance in differentiating malignant from benign thyroid nodules and the predictive efficacy of the risk stratification systems. KEY POINTS: • Compared with the classical markedly hypoechoic as a diagnostic criterion for malignancy, the modified markedly hypoechoic resulted in a significant increase in sensitivity and AUC. • The C-TIRADS with the modified markedly hypoechoic achieved higher AUC and specificity than that with the classical markedly hypoechoic (p = 0.01 and < 0.001, respectively).

Virtual elastography ultrasound via generative adversarial network for breast cancer diagnosis.

Elastography ultrasound (EUS) imaging is a vital ultrasound imaging modality. The current use of EUS faces many challenges, such as vulnerability to subjective manipulation, echo signal attenuation, and unknown risks of elastic pressure in certain delicate tissues. The hardware requirement of EUS also hinders the trend of miniaturization of ultrasound equipment. Here we show a cost-efficient solution by designing a deep neural network to synthesize virtual EUS (V-EUS) from conventional B-mode images. A total of 4580 breast tumor cases were collected from 15 medical centers, including a main cohort with 2501 cases for model establishment, an external dataset with 1730 cases and a portable dataset with 349 cases for testing. In the task of differentiating benign and malignant breast tumors, there is no significant difference between V-EUS and real EUS on high-end ultrasound, while the diagnostic performance of pocket-sized ultrasound can be improved by about 5% after V-EUS is equipped.

Deep learning algorithm using bispectrum analysis energy feature maps based on ultrasound radiofrequency signals to detect breast cancer.

Background: Ultrasonography is an important imaging method for clinical breast cancer screening. As the original echo signals of ultrasonography, ultrasound radiofrequency (RF) signals provide abundant tissue macroscopic and microscopic information and have important development and utilization value in breast cancer detection. Methods: In this study, we proposed a deep learning method based on bispectrum analysis feature maps to process RF signals and realize breast cancer detection. The bispectrum analysis energy feature maps with frequency subdivision were first proposed and applied to breast cancer detection in this study. Our deep learning network was based on a weight sharing network framework for the input of multiple feature maps. A feature map attention module was designed for multiple feature maps input of the network to adaptively learn both feature maps and features that were conducive to classification. We also designed a similarity constraint factor, learning the similarity and difference between feature maps by cosine distance. Results: The experiment results showed that the areas under the receiver operating characteristic curves of our proposed method in the validation set and two independent test sets for benign and malignant breast tumor classification were 0.913, 0.900, and 0.885, respectively. The performance of the model combining four ultrasound bispectrum analysis energy feature maps in breast cancer detection was superior to that of the model using an ultrasound grayscale image and the model using a single bispectrum analysis energy feature map in this study. Conclusion: The combination of deep learning technology and our proposed ultrasound bispectrum analysis energy feature maps effectively realized breast cancer detection and was an efficient method of feature extraction and utilization of ultrasound RF signals.

Novel Human Artificial Intelligence Hybrid Framework Pinpoints Thyroid Nodule Malignancy and Identifies Overlooked Second-Order Ultrasonographic Features.

We present a Human Artificial Intelligence Hybrid (HAIbrid) integrating framework that reweights Thyroid Imaging Reporting and Data System (TIRADS) features and the malignancy score predicted by a convolutional neural network (CNN) for nodule malignancy stratification and diagnosis. We defined extra ultrasonographical features from color Doppler images to explore malignancy-relevant features. We proposed Gated Attentional Factorization Machine (GAFM) to identify second-order interacting features trained via a 10 fold distribution-balanced stratified cross-validation scheme on ultrasound images of 3002 nodules all finally characterized by postoperative pathology (1270 malignant ones), retrospectively collected from 131 hospitals. Our GAFM-HAIbrid model demonstrated significant improvements in Area Under the Curve (AUC) value (p-value < 10−5), reaching about 0.92 over the standalone CNN (~0.87) and senior radiologists (~0.86), and identified a second-order vascularity localization and morphological pattern which was overlooked if only first-order features were considered. We validated the advantages of the integration framework on an already-trained commercial CNN system and our findings using an extra set of ultrasound images of 500 nodules. Our HAIbrid framework allows natural integration to clinical workflow for thyroid nodule malignancy risk stratification and diagnosis, and the proposed GAFM-HAIbrid model may help identify novel diagnosis-relevant second-order features beyond ultrasonography.

Multi-Attribute Attention Network for Interpretable Diagnosis of Thyroid Nodules in Ultrasound Images.

Ultrasound (US) is the primary imaging technique for the diagnosis of thyroid cancer. However, accurate identification of nodule malignancy is a challenging task that can elude less-experienced clinicians. Recently, many computer-aided diagnosis (CAD) systems have been proposed to assist this process. However, most of them do not provide the reasoning of their classification process, which may jeopardize their credibility in practical use. To overcome this, we propose a novel deep learning (DL) framework called multi-attribute attention network (MAA-Net) that is designed to mimic the clinical diagnosis process. The proposed model learns to predict nodular attributes and infer their malignancy based on these clinically-relevant features. A multi-attention scheme is adopted to generate customized attention to improve each task and malignancy diagnosis. Furthermore, MAA-Net utilizes nodule delineations as nodules spatial prior guidance for the training rather than cropping the nodules with additional models or human interventions to prevent losing the context information. Validation experiments were performed on a large and challenging dataset containing 4554 patients. Results show that the proposed method outperformed other state-of-the-art methods and provides interpretable predictions that may better suit clinical needs.

Boundary-rendering network for breast lesion segmentation in ultrasound images.

Breast Ultrasound (BUS) has proven to be an effective tool for the early detection of cancer in the breast. A lesion segmentation provides identification of the boundary, shape, and location of the target, and serves as a crucial step toward accurate diagnosis. Despite recent efforts in developing machine learning algorithms to automate this process, problems remain due to the blurry or occluded edges and highly irregular nodule shapes. Existing methods often produce over-smooth or inaccurate results, failing the need of identifying detailed boundary structures which are of clinical interest. To overcome these challenges, we propose a novel boundary-rendering framework that explicitly highlights the importance of boundary for automated nodule segmentation in BUS images. It utilizes a boundary selection module to automatically focuses on the ambiguous boundary region and a graph convolutional-based boundary rendering module to exploit global contour information. Furthermore, the proposed framework embeds nodule classification via semantic segmentation and encourages co-learning across tasks. Validation experiments were performed on different BUS datasets to verify the robustness of the proposed method. Results show that the proposed method outperforms states-of-art segmentation approaches (Dice=0.854, IOU=0.919, HD=17.8) in nodule delineation, as well as obtains a higher classification accuracy than classical classification models.

Mass fractions, solubility, speciation and isotopic compositions of iron in coal and municipal waste fly ash.

Deposition of anthropogenic aerosols may contribute significantly to dissolved Fe in the open ocean, affecting marine primary production and biogeochemical cycles; however, fractional solubility of Fe is not well understood for anthropogenic aerosols. This work investigated mass fractions, solubility, speciation and isotopic compositions of Fe in coal and municipal waste fly ash. Compared to desert dust (3.1 ± 1.1%), the average mass fraction of Fe was higher in coal fly ash (6.2 ± 2.7%) and lower in municipal waste fly ash (2.6 ± 0.4%), and the average Fe/Al ratios were rather similar for the three types of particles. Municipal waste fly ash showed highest Fe solubility (1.98 ± 0.43%) in acetate buffer (pH: 4.3), followed by desert dust (0.43 ± 0.30%) and coal fly ash (0.24 ± 0.28%), suggesting that not all the anthropogenic aerosols showed higher Fe solubility than desert dust. For the samples examined in our work, amorphous Fe appeared to be an important controlling factor for Fe solubility, which was not correlated with particle size or BET surface area. Compared to desert dust (-0.05‰ to 0.21‰), coal and municipal waste fly ash showed similar or even higher δ56Fe values for total Fe (range: 0.05‰ to 0.75‰), implying that the presence of coal or municipal waste fly ash may not be able to explain significantly smaller δ56Fe values reported for total Fe in ambient aerosols affected by anthropogenic sources.

Comprehensive Risk System Based on Shear Wave Elastography and BI-RADS Categories in Assessing Axillary Lymph Node Metastasis of Invasive Breast Cancer-A Multicenter Study.

Purpose: To develop a risk stratification system that can predict axillary lymph node (LN) metastasis in invasive breast cancer based on the combination of shear wave elastography (SWE) and conventional ultrasound. Materials and Methods: A total of 619 participants pathologically diagnosed with invasive breast cancer underwent breast ultrasound examinations were recruited from a multicenter of 17 hospitals in China from August 2016 to August 2017. Conventional ultrasound and SWE features were compared between positive and negative LN metastasis groups. The regression equation, the weighting, and the counting methods were used to predict axillary LN metastasis. The sensitivity, specificity, and the areas under the receiver operating characteristic curve (AUC) were calculated. Results: A significant difference was found in the Breast Imaging Reporting and Data System (BI-RADS) category, the "stiff rim" sign, minimum elastic modulus of the internal tumor and peritumor region of 3 mm between positive and negative LN groups (p < 0.05 for all). There was no significant difference in the diagnostic performance of the regression equation, the weighting, and the counting methods (p > 0.05 for all). Using the counting method, a 0-4 grade risk stratification system based on the four characteristics was established, which yielded an AUC of 0.656 (95% CI, 0.617-0.693, p < 0.001), a sensitivity of 54.60% (95% CI, 46.9%-62.1%), and a specificity of 68.99% (95% CI, 64.5%-73.3%) in predicting axillary LN metastasis. Conclusion: A 0-4 grade risk stratification system was developed based on SWE characteristics and BI-RADS categories, and this system has the potential to predict axillary LN metastases in invasive breast cancer.

Clinical Application of Computer-Aided Diagnosis for Breast Ultrasonography: Factors That Lead to Discordant Results in Radial and Antiradial Planes.

OBJECTIVE: To investigate factors that may lead to discordant results in radial and antiradial planes when applying a computer-aided diagnostic system (S-DetectTM) for breast ultrasound (US). METHODS: From May 2019 to September 2019, a total of 288 breast lesions from 286 consecutive women were analyzed. Diagnostic performance and diagnostic agreement of the CAD system between the radial and antiradial planes were calculated. Based on the CAD results in the radial and antiradial planes, the lesions were classified into two groups: the agreement group and the discordant group. Ultrasound imaging and clinicopathologic factors in the two groups were compared. RESULTS: Of the 288 breast lesions, 187 (64.7%) were benign, and 101 (35.3%) were malignant. There was no difference in diagnostic performance of the CAD system between the radial and antiradial planes. The diagnostic agreement of the CAD system between these two orthogonal planes was good (ĸ = 0.645). Compared to category 3, 4C and 5 lesions, category 4A and 4B lesions were more likely to have discordant CAD results. In the subgroup of malignant tumors, lesions diagnosed as carcinoma in situ (P = 0.014), staged as T1 (P = 0.013) and with low Ki-67 status (P = 0.024) were significantly associated with discordant CAD results. CONCLUSION: The CAD system for breast ultrasound has a favorable diagnostic performance, and discordant CAD results between the radial and antiradial planes were more frequent for BI-RADS category 4A-4B lesions and less invasive malignant lesions.

Shear wave elastography and pulsed doppler for breast lesions: Similar diagnostic performance and positively correlated stiffness and blood flow resistance.

PURPOSE: To compare the diagnostic performance of shear wave elastography (SWE) and pulsed Doppler ultrasound in breast lesions, and to explore whether the quantitative SWE parameters correlated with pulsed Doppler ultrasound parameters. MATERIALS AND METHODS: Seventy-nine patients with 79 breast lesions who had undergone conventional ultrasound, pulsed Doppler ultrasound and SWE examination were included. All of them underwent core needle biopsy or surgery within one week. Parameters including Emax (the maximum elastic modulus), Emean (mean elastic modulus), Emin (minimum elastic modulus), Esd (elastic modulus standard deviation), and RI (resistive index), PI (pulsatility index), PSV (peak systolic velocity) and EDV (end diastolic velocity) were obtained for statistical analysis. RESULTS: Almost all SWE parameters were significantly different between benign and malignant breast lesions (P＜0.05). There was no significant difference between Esd and PI (P＞0.05), which had the best AUC among SWE and vascular parameters respectively (0.877 vs. 0.871). Emax showed a moderate correlation with PI (P = 0.000, r = 0.552) and RI (P = 0.000, r = 0.544), and Esd moderately correlated with PI (P = 0.000, r = 0.567) and RI (P = 0.000, r = 0.546). For the benign group, no parameters showed any significant correlation (P＞0.05), while for the malignant group, Emax and Esd also significantly correlated with PI or RI. CONCLUSIONS: SWE and pulsed Doppler ultrasound had similar diagnostic efficacy for breast lesions. SWE and pulsed Doppler parameters were significantly correlated in breast lesions, especially in malignant ones, indicating the potential association between elastographic and vascular characteristics of breast tumors.

AW3M: An auto-weighting and recovery framework for breast cancer diagnosis using multi-modal ultrasound.

Recently, more clinicians have realized the diagnostic value of multi-modal ultrasound in breast cancer identification and began to incorporate Doppler imaging and Elastography in the routine examination. However, accurately recognizing patterns of malignancy in different types of sonography requires expertise. Furthermore, an accurate and robust diagnosis requires proper weights of multi-modal information as well as the ability to process missing data in practice. These two aspects are often overlooked by existing computer-aided diagnosis (CAD) approaches. To overcome these challenges, we propose a novel framework (called AW3M) that utilizes four types of sonography (i.e. B-mode, Doppler, Shear-wave Elastography, and Strain Elastography) jointly to assist breast cancer diagnosis. It can extract both modality-specific and modality-invariant features using a multi-stream CNN model equipped with self-supervised consistency loss. Instead of assigning the weights of different streams empirically, AW3M automatically learns the optimal weights using reinforcement learning techniques. Furthermore, we design a light-weight recovery block that can be inserted to a trained model to handle different modality-missing scenarios. Experimental results on a large multi-modal dataset demonstrate that our method can achieve promising performance compared with state-of-the-art methods. The AW3M framework is also tested on another independent B-mode dataset to prove its efficacy in general settings. Results show that the proposed recovery block can learn from the joint distribution of multi-modal features to further boost the classification accuracy given single modality input during the test.

Using Deep Neural Network to Diagnose Thyroid Nodules on Ultrasound in Patients With Hashimoto's Thyroiditis.

OBJECTIVE: The aim of this study is to develop a model using Deep Neural Network (DNN) to diagnose thyroid nodules in patients with Hashimoto's Thyroiditis. METHODS: In this retrospective study, we included 2,932 patients with thyroid nodules who underwent thyroid ultrasonogram in our hospital from January 2017 to August 2019. 80% of them were included as training set and 20% as test set. Nodules suspected for malignancy underwent FNA or surgery for pathological results. Two DNN models were trained to diagnose thyroid nodules, and we chose the one with better performance. The features of nodules as well as parenchyma around nodules will be learned by the model to achieve better performance under diffused parenchyma. 10-fold cross-validation and an independent test set were used to evaluate the performance of the algorithm. The performance of the model was compared with that of the three groups of radiologists with clinical experience of <5 years, 5-10 years, >10 years respectively. RESULTS: In total, 9,127 images were collected from 2,932 patients with 7,301 images for the training set and 1,806 for the test set. 56% of the patients enrolled had Hashimoto's Thyroiditis. The model achieved an AUC of 0.924 for distinguishing malignant and benign nodules in the test set. It showed similar performance under diffused thyroid parenchyma and normal parenchyma with sensitivity of 0.881 versus 0.871 (p = 0.938) and specificity of 0.846 versus 0.822 (p = 0.178). In patients with HT, the model achieved an AUC of 0.924 to differentiate malignant and benign nodules which was significantly higher than that of the three groups of radiologists (AUC = 0.824, 0.857, 0.863 respectively, p < 0.05). CONCLUSION: The model showed high performance in diagnosing thyroid nodules under both normal and diffused parenchyma. In patients with Hashimoto's Thyroiditis, the model showed a better performance compared to radiologists with various years of experience.

Thyroid Parenchyma Microcalcifications on Ultrasound for Predicting Lymph Node Metastasis in Papillary Thyroid Carcinoma: A Prospective Multicenter Study in China.

Objective: Cervical lymph node metastasis (LNM) was found to be clinically significant prognostic factors of patients with papillary thyroid carcinomas (PTC). Ultrasound (US) characteristics of thyroid nodules and thyroid parenchyma may be used to predict LNM. To investigate the value of nodular US features as well as thyroid parenchymal microcalcifications on US in predicting LNM in patients with PTC. Methods: This prospective study was approved by the Institutional Review Board. From January 2018 to June 2019, 971 consecutive patients with solitary PTC who underwent preoperative neck US evaluation were included from six hospitals in China. The US features of thyroid nodules as well as thyroid parenchyma microcalcifications were carefully evaluated based on the static images and dynamic clips. Univariate and multivariate analyses were performed to determine independent predictors of LNM. Results: Of the 971 patients, 760 were female, 211 were male. According to the pathological examination, 241(24.82%) patients were found with cervical LNM (LNM positive group), while 730 (75.18%) patients were not (LNM negative group). Multiple logistic regression analysis showed that young age (<55 years old) (OR = 1.522, P = 0.047), large size (>10 mm) (OR = 1.814, P < 0.001), intratumoral microcalcifications (OR = 1.782, P = 0.002) and thyroid parenchyma microcalcifications (OR = 1.635, P = 0.046) were independent risk factors for LNM of PTC. Conclusions: Young age, large nodule size, intratumoral microcalcifications, as well as thyroid parenchyma microcalcifications on US are independent predictors of cervical LNM for patients with PTC.

Thyroid imaging reporting and data system (TIRADS) for ultrasound features of nodules: multicentric retrospective study in China.

PURPOSE: To establish a practical and simplified Chinese thyroid imaging reporting and data system (C-TIRADS) based on the Chinese patient database. METHODS: A total of 2141 thyroid nodules that were neither cystic nor spongy were used in the current study. These specimens were derived from 2141 patients in 131 alliance hospitals of the Chinese Artificial Intelligence Alliance for Thyroid and Breast Ultrasound. The ultrasound features, including location, orientation, margin, halo, composition, echogenicity, echotexture, echogenic foci and posterior features were assessed. Univariate and multivariate analyses were performed to investigate the association between ultrasound features and malignancy. The regression equation, the weighting, and the counting methods were used to determine the malignant risk of the thyroid nodules. The areas under the receiver operating characteristic curve (Az values) were calculated. RESULTS: Of the 2141 thyroid nodules, 1572 were benign, 565 were malignant, and 4 were borderline. Vertical orientation, ill-defined, or irregular margin (including extrathyroidal extension), microcalcifications, solid, and markedly hypoechoic were positively associated with malignancy, while comet-tail artifacts were negatively associated with malignancy. The logistic regression equation yielded the highest Az value of 0.913, which was significantly higher than that obtained using the weighting method (0.893) and the counting method (0.890); however, no significant difference was found between the latter two. The C-TIRADS, based on the counting method, was designed following the principle of balancing the diagnostic performance and sensitivity of the risk stratification with the ease of use. CONCLUSIONS: A relatively simple C-TIRADS was established using the counting value of positive and negative ultrasound features.

Lymph node metastasis prediction of papillary thyroid carcinoma based on transfer learning radiomics.

Non-invasive assessment of the risk of lymph node metastasis (LNM) in patients with papillary thyroid carcinoma (PTC) is of great value for the treatment option selection. The purpose of this paper is to develop a transfer learning radiomics (TLR) model for preoperative prediction of LNM in PTC patients in a multicenter, cross-machine, multi-operator scenario. Here we report the TLR model produces a stable LNM prediction. In the experiments of cross-validation and independent testing of the main cohort according to diagnostic time, machine, and operator, the TLR achieves an average area under the curve (AUC) of 0.90. In the other two independent cohorts, TLR also achieves 0.93 AUC, and this performance is statistically better than the other three methods according to Delong test. Decision curve analysis also proves that the TLR model brings more benefit to PTC patients than other methods.