Thyroid Nodules In Ultrasound Images Research Articles

A Risk Stratification Study of Ultrasound Images of Thyroid Nodules Based on Improved DETR

ABSTRACTThe Chinese Thyroid Imaging Reporting and Data System (C‐TIRADS) standard is based on the Chinese current medical context. However, at present, there is a lack of C‐TIRADS‐based automatic computer‐aided diagnosis system for thyroid nodule ultrasound images, and the existing algorithms for detecting and recognizing thyroid nodules are basically for the dichotomous classification of benign and malignant. We used the DETR (detection transformer) model as a baseline model and carried out model enhancements to address the shortcomings of unsatisfactory classification accuracy and difficulty in detecting small thyroid nodules in the DETR model. First, to investigate the method of acquiring multi‐scale features of thyroid nodule ultrasound images, we choose TResNet‐L as the feature extraction network and introduce multi‐scale feature information and group convolution, thereby enhancing the model's multi‐label classification accuracy. Second, a parallel decoder architecture for multi‐label thyroid nodule ultrasound image classification is designed to enhance the learning of correlation between pathological feature class labels, aiming to improve the multi‐label classification accuracy of the detection model. Third, the loss function of the detection model is improved. We propose a linear combination of Smooth L1‐Loss and CIoU Loss as the model's bounding box loss function and asymmetric loss as the model's multi‐label classification loss function, aiming to further improve the detection model's detection accuracy for small thyroid nodules. The experiment results show that the improved DETR model achieves an AP of 92.4% and 81.6% with IoU thresholds of 0.5 and 0.75, respectively.

Deep learning model for diagnosis of thyroid nodules with size less than 1 cm: A multicenter, retrospective study

ObjectiveTo develop a ultrasound images based dual-channel deep learning model to achieve accurate early diagnosis of thyroid nodules less than 1 cm. MethodsA dual-channel deep learning model called thyroid nodule transformer network (TNT-Net) was proposed. The model has two input channels for transverse and longitudinal ultrasound images of thyroid nodules, respectively. A total of 9649 nodules from 8455 patients across five hospitals were retrospectively collected. The data were divided into a training set (8453 nodules, 7369 patients), an internal test set (565 nodules, 512 patients), and an external test set (631 nodules, 574 patients). ResultsTNT-Net achieved an area under the curve (AUC) of 0.953 (95 % confidence interval (CI): 0.934, 0.969) on the internal test set and 0.941 (95 % CI: 0.921, 0.957) on the external test set, significantly outperforming traditional deep convolutional neural network models and single-channel swin transformer model, whose AUCs ranged from 0.800 (95 % CI: 0.759, 0.837) to 0.856 (95 % CI: 0.819, 0.881). Furthermore, feature heatmap visualization showed that TNT-Net could extract richer and more energetic malignant nodule patterns. ConclusionThe proposed TNT-Net model significantly improved the recognition capability for thyroid nodules with size less than 1 cm. This model has the potential to reduce overdiagnosis and overtreatment of such nodules, providing essential support for precise management of thyroid nodules while complementing fine-needle aspiration biopsy.

Mamba- and ResNet-Based Dual-Branch Network for Ultrasound Thyroid Nodule Segmentation.

Accurate segmentation of thyroid nodules in ultrasound images is crucial for the diagnosis of thyroid cancer and preoperative planning. However, the segmentation of thyroid nodules is challenging due to their irregular shape, blurred boundary, and uneven echo texture. To address these challenges, a novel Mamba- and ResNet-based dual-branch network (MRDB) is proposed. Specifically, the visual state space block (VSSB) from Mamba and ResNet-34 are utilized to construct a dual encoder for extracting global semantics and local details, and establishing multi-dimensional feature connections. Meanwhile, an upsampling-convolution strategy is employed in the left decoder focusing on image size and detail reconstruction. A convolution-upsampling strategy is used in the right decoder to emphasize gradual feature refinement and recovery. To facilitate the interaction between local details and global context within the encoder and decoder, cross-skip connection is introduced. Additionally, a novel hybrid loss function is proposed to improve the boundary segmentation performance of thyroid nodules. Experimental results show that MRDB outperforms the state-of-the-art approaches with DSC of 90.02% and 80.6% on two public thyroid nodule datasets, TN3K and TNUI-2021, respectively. Furthermore, experiments on a third external dataset, DDTI, demonstrate that our method improves the DSC by 10.8% compared to baseline and exhibits good generalization to clinical small-scale thyroid nodule datasets. The proposed MRDB can effectively improve thyroid nodule segmentation accuracy and has great potential for clinical applications.

STU3Net: An Improved U‐Net With Swin Transformer Fusion for Thyroid Nodule Segmentation

ABSTRACTThyroid nodules are a common endocrine system disorder for which accurate ultrasound image segmentation is important for evaluation and diagnosis, as well as a critical step in computer‐aided diagnostic systems. However, the accuracy and consistency of segmentation remains a challenging task due to the presence of scattering noise, low contrast and resolution in ultrasound images. Therefore, we propose a deep learning‐based CAD (computer‐aided diagnosis) method, STU3Net in this paper, aiming at automatic segmentation of thyroid nodules. The method employs a modified Swin Transformer combined with a CNN encoder, which is capable of extracting morphological features and edge details of thyroid nodules in ultrasound images. In decoding through the features for image reconstruction, we introduce a modified three‐layer U‐Net network with cross‐layer connectivity to further enhance image reduction. This cross‐layer connectivity enhances the network's capture and representation of the contained image feature information by creating skip connections between different layers and merging the detailed information of the shallow network with the abstract information of the deeper network. Through comparison experiments with current mainstream deep learning methods on the TN3K and BUSI datasets, we validate the superiority of the STU3Net method in thyroid nodule segmentation performance. The experimental results show that STU3Net outperforms most of the mainstream models on the TN3K dataset, with Dice and IoU reaching 0.8368 and 0.7416, respectively, which are significantly better than other methods. The method demonstrates excellent performance on these datasets and provides radiologists with an effective auxiliary tool to accurately detect thyroid nodules in ultrasound images.

Thyroid-DETR: Thyroid nodule detection model with transformer in ultrasound images

Deep learning-based object detection methods using ultrasound images to assist clinicians in rapid and accurate diagnosis of thyroid nodules has been an attractive area of research. Despite the great progress made by object detectors, these may still produce unsatisfactory performance due to certain factors, such as the presence of abundant background information and significant nodule size variations in ultrasound images of thyroid nodules. In this study, we draw on the idea of a Transformer and design a powerful thyroid nodule ultrasound imaging detection model, namely Thyroid-DETR. The proposed Thyroid-DETR has the following features: (i) a deformable convolution v2 is introduced flexibly capture small nodules of irregular shapes and small sizes, (ii) a multi-head self-attention structure is designed to acquire and learn broader correlated information in the shallow layer, (iii) a dual-stream training structure with BatchFormerV2 is developed to fully capture the effective information of medical images; (iv) fine-tuning methods to optimize the model are used to improve the diagnostic accuracy for detecting thyroid nodules. To evaluate the efficacy and effectiveness of the proposed approach. Extensive experiments and comprehensive comparisons are conducted on real thyroid nodule ultrasound images from the hospital. The results show that our method significantly outperforms the state-of-the-art methods.

A Multi-View Deep Learning Model for Thyroid Nodules Detection and Characterization in Ultrasound Imaging.

Thyroid Ultrasound (US) is the primary method to evaluate thyroid nodules. Deep learning (DL) has been playing a significant role in evaluating thyroid cancer. We propose a DL-based pipeline to detect and classify thyroid nodules into benign or malignant groups relying on two views of US imaging. Transverse and longitudinal US images of thyroid nodules from 983 patients were collected retrospectively. Eighty-one cases were held out as a testing set, and the rest of the data were used in five-fold cross-validation (CV). Two You Look Only Once (YOLO) v5 models were trained to detect nodules and classify them. For each view, five models were developed during the CV, which was ensembled by using non-max suppression (NMS) to boost their collective generalizability. An extreme gradient boosting (XGBoost) model was trained on the outputs of the ensembled models for both views to yield a final prediction of malignancy for each nodule. The test set was evaluated by an expert radiologist using the American College of Radiology Thyroid Imaging Reporting and Data System (ACR-TIRADS). The ensemble models for each view achieved a mAP0.5 of 0.797 (transverse) and 0.716 (longitudinal). The whole pipeline reached an AUROC of 0.84 (CI 95%: 0.75-0.91) with sensitivity and specificity of 84% and 63%, respectively, while the ACR-TIRADS evaluation of the same set had a sensitivity of 76% and specificity of 34% (p-value = 0.003). Our proposed work demonstrated the potential possibility of a deep learning model to achieve diagnostic performance for thyroid nodule evaluation.

Semi-Supervised Thyroid Nodule Detection in Ultrasound Videos.

Deep learning techniques have been investigated for the computer-aided diagnosis of thyroid nodules in ultrasound images. However, most existing thyroid nodule detection methods were simply based on static ultrasound images, which cannot well explore spatial and temporal information following the clinical examination process. In this paper, we propose a novel video-based semi-supervised framework for ultrasound thyroid nodule detection. Especially, considering clinical examinations that need to detect thyroid nodules at the ultrasonic probe positions, we first construct an adjacent frame guided detection backbone network by using adjacent supporting reference frames. To further reduce the labour-intensive thyroid nodule annotation in ultrasound videos, we extend the video-based detection in a semi-supervised manner by using both labeled and unlabeled videos. Based on the detection consistency in sequential neighbouring frames, a pseudo label adaptation strategy is proposed for the refinement of unpredicted frames. The proposed framework is validated on 996 transverse viewed and 1088 longitudinal viewed ultrasound videos. Experimental results demonstrated the superior performance of our proposed method in the ultrasound video-based detection of thyroid nodules.

Localization and Risk Stratification of Thyroid Nodules in Ultrasound Images Through Deep Learning

ObjectiveDeep learning algorithms have commonly been used for the differential diagnosis between benign and malignant thyroid nodules. In this study, we aimed to develop an integrated system that combines deep learning model and clinical standard thyroid imaging reporting and data system (TI-RADS) for the simultaneous segmentation and risk stratification of thyroid nodules. Methods304 ultrasound images from two independent sites with TI-RADS 4 thyroid nodules were collected. Edge-connection and Criminisi algorithm was used to remove manually-induced markers in ultrasound images. An integrated system based on TI-RADS and mask region-based convolution neural network (Mask R-CNN) was proposed to stratify subclasses of TI-RADS 4 thyroid nodules and to segment thyroid nodules in the ultrasound images. Accuracy and precision-recall curve were used to evaluate stratification performance, dice similarity coefficient (DSC) between the segmentation of Mask R-CNN and the radiologist's contour was used to evaluate segmentation performance of the model. ResultsThe combined approach could significantly enhance the performance of the proposed integrated system. Overall stratification accuracy of TI-RADS 4 thyroid nodules, mean average precision, and mean DSC of the proposed model in the independent test set were 90.79%, 0.8579, and 0.83, respectively. Specifically, stratification accuracy for TI-RADS 4a, 4b and 4c thyroid nodules were 95.83%, 84.21%, and 77.78%. ConclusionAn integrated system combining TI-RADS and deep learning model was developed. The system can provide clinicians with not only diagnostic assistance from TI-RADS, but also accurate segmentation of thyroid nodules, which improves the applicability of the system in clinical practice.

An Improved k-Nearest Neighbor Algorithm for Recognition and Classification of Thyroid Nodules.

To complete the task of automatic recognition and classification of thyroid nodules and solve the problem of high classification error rates when the samples are imbalanced. An improved k-nearest neighbor (KNN) algorithm is proposed and a method for automatic thyroid nodule classification based on the improved KNN algorithm is established. In the improved KNN algorithm, we consider not only the number of class labels for various classes of data in KNNs, but also the corresponding weights. And we use the Minkowski distance measure instead of the Euclidean distance measure. A total of 508 ultrasound images of thyroid nodules, including 415 benign nodules and 93 malignant nodules, were used in the paper. Experimental results show the improved KNN has 0.872549 accuracy, 0.867347 precision, 1 recall, and 0.928962 F1-score. At the same time, we also considered the influence of different distance weights, the value of k, different distance measures on the classification results. A comparison result shows that our method has a better performance than the traditional KNN and other classical machine learning methods.

Interobserver variability in thyroid ultrasound.

Ultrasound evaluation of thyroid nodules is the preferred technique, but it is dependent on operator interpretation, leading to inter-observer variability. The current study aimed to determine the inter-physician consensus on nodular characteristics, risk categorization in the classification systems, and the need for fine needle aspiration puncture. Four endocrinologists from the same center blindly evaluated 100 ultrasound images of thyroid nodules from 100 different patients. The following ultrasound features were evaluated: composition, echogenicity, margins, calcifications, and microcalcifications. Nodules were also classified according to ATA, EU-TIRADS, K-TIRADS, and ACR-TIRADS classifications. Krippendorff's alpha test was used to assess interobserver agreement. The interobserver agreement for ultrasound features was: Krippendorff's coefficient 0.80 (0.71-0.89) for composition, 0.59 (0.47-0.72) for echogenicity, 0.73 (0.57-0.88) for margins, 0.55 (0.40-0.69) for calcifications, and 0.50 (0.34-0.67) for microcalcifications. The concordance for the classification systems was 0.7 (0.61-0.80) for ATA, 0.63 (0.54-0.73) for EU-TIRADS, 0.64 (0.55-0.73) for K-TIRADS, and 0.68 (0.60-0.77) for K-TIRADS. The concordance in the indication of fine needle aspiration puncture (FNA) was 0.86 (0.71-1), 0.80 (0.71-0.88), 0.77 0.67-0.87), and 0.73 (0.64-0.83) for systems previously described respectively. Interobserver agreement was acceptable for the identification of nodules requiring cytologic study using various classification systems. However, limited concordance was observed in risk stratification and many ultrasonographic characteristics of the nodules.

Real-time reliable semantic segmentation of thyroid nodules in ultrasound images

Objective. Low efficiency in medical image segmentation is a common issue that limits computer-aided diagnosis development. Due to the varying positions and sizes of nodules, it is not easy to accurately segment ultrasound images. This study aims to propose a segmentation model that maintains high efficiency while improving accuracy. Approach. We propose a novel layer that integrates the advantages of dense connectivity, dilated convolution, and factorized filters to maintain excellent efficiency while improving accuracy. Dense connectivity optimizes feature reuse, dilated convolution redesigns layers, and factorized convolution improves efficiency. Moreover, we propose a loss function optimization method from a pixel perspective to increase the network's accuracy further. Main results. Experiments on the Thyroid dataset show that our method achieves 81.70% intersection-over-union (IoU), 90.50% true positive rate (TPR), and 0.25% false positive rate (FPR). In terms of accuracy, our method outperforms the state-of-the-art methods, with twice faster inference and nearly 400 times fewer parameters. Meanwhile, in a test on an External Thyroid dataset, our method achieves 77.03% IoU, 82.10% TPR, and 0.16% FPR, demonstrating our proposed model's robustness. Significance. We propose a real-time semantic segmentation architecture for thyroid nodule segmentation in ultrasound images called fully convolution dense dilated network (FCDDN). Our method runs fast with a few parameters and is suitable for medical devices requiring real-time segmentation.

A multitask approach for automated detection and segmentation of thyroid nodules in ultrasound images

An increase in the incidence and diagnosis of thyroid nodules and thyroid cancer underscores the need for a better approach to nodule detection and risk stratification in ultrasound (US) images that can reduce healthcare costs, patient discomfort, and unnecessary invasive procedures. However, variability in ultrasound technique and interpretation makes the diagnostic process partially subjective. Therefore, an automated approach that detects and segments nodules could improve performance on downstream tasks, such as risk stratification. Ultrasound studies were acquired from 280 patients at UCLA Health, totaling 9888 images, and annotated by collaborating radiologists. Current deep learning architectures for segmentation are typically semi-automated because they are evaluated solely on images known to have nodules and do not assess ability to identify suspicious images. However, the proposed multitask approach both detects suspicious images and segments potential nodules; this allows for a clinically translatable model that aptly parallels the workflow for thyroid nodule assessment. The multitask approach is centered on an anomaly detection (AD) module that can be integrated with any UNet architecture variant to improve image-level nodule detection. Of the evaluated multitask models, a UNet with a ImageNet pretrained encoder and AD achieved the highest F1 score of 0.839 and image-wide Dice similarity coefficient of 0.808 on the hold-out test set. Furthermore, models were evaluated on two external validations datasets to demonstrate generalizability and robustness to data variability. Ultimately, the proposed architecture is an automated multitask method that expands on previous methods by successfully both detecting and segmenting nodules in ultrasound.

CRSANet: Class Representations Self-Attention network for the segmentation of thyroid nodules

Segmenting accurate thyroid nodules in medical ultrasound images is always non-trivial due to the large variation in size, shape and texture of nodules, and also the presence of surrounding tissues and organs with similar intensity values. Existing algorithms solve this by designing multi-scale spatial context modeling modules and introducing the emerging self-attention based paradigm, while failing to learn the class-wise relational context around nodules. To this end, we propose a novel coarse-to-fine framework, namely, CRSANet, which leverages both class-wise relational context and spatial multi-scale context for comprehensive context perception. Specifically, we first introduce the concept of Class Representations, which describe the overall representation of thyroid nodules in ultrasound images from a categorical perspective. Secondly, we propose a backbone-free dual-branch Class Representation Self-Attention Module (CRSA) to refine the coarse segmentation results from segmentation backbones. Concretely, Fine Class Region (FCR) branch is proposed to impose consistency between pixels and their corresponding Class Representations to correct the wrongly classified pixels. Meanwhile, Class Contextual Representation (CCR) branch is introduced to enhance the ability of pixel representations with the corresponding Class Representations. We evaluate the effectiveness and generality of CRSANet using three types of backbone networks. In the experiments, our proposed U-CRSANet achieves a mIoU of 78.86% and 81.35%, an accuracy of 97.11% and 98.63%, a Dice value of 74.26% and 87.57%, respectively, which outperforms most state-of-the-art methods on the public DDTI and TN-SCUI datasets.

Research on Thyroid Nodule Segmentation Using an Improved U-Net Network

To develop a precise neural network model designed for segmenting ultrasound images of thyroid nodules. The deep learning U-Net network model was utilized as the main backbone, with improvements made to the convolutional operations and the implementation of multilayer perceptron modeling at the lower levels, using the more effective BCEDice loss function. The modified network achieved enhanced segmentation precision and robust generalization capabilities, with a Dice coefficient of 0.9062, precision of 0.9153, recall of 0.9023, and an F1 score of 0.9062, indicating improvements over the U-Net and Swin-Unet to various extents. The U-Net network enhancement presented in this study outperforms the original U-Net across all performance indicators. This advancement could help physicians make more precise and efficient diagnoses, thereby minimizing medical errors.

Using multimodal ultrasound including full-time-series contrast-enhanced ultrasound cines for identifying the nature of thyroid nodules.

Based on the conventional ultrasound images of thyroid nodules, contrast-enhanced ultrasound (CEUS) videos were analyzed to investigate whether CEUS improves the classification accuracy of benign and malignant thyroid nodules using machine learning (ML) radiomics and compared with radiologists. The B-mode ultrasound (B-US), real-time elastography (RTE), color doppler flow images (CDFI) and CEUS cines of patients from two centers were retrospectively gathered. Then, the region of interest (ROI) was delineated to extract the radiomics features. Seven ML algorithms combined with four kinds of radiomics data (B-US, B-US + CDFI + RTE, CEUS, and B-US + CDFI + RTE + CEUS) were applied to establish 28 models. The diagnostic performance of ML models was compared with interpretations from expert and nonexpert readers. A total of 181 thyroid nodules from 181 patients of 64 men (mean age, 42 years +/- 12) and 117 women (mean age, 46 years +/- 12) were included. Adaptive boosting (AdaBoost) achieved the highest area under the receiver operating characteristic curve (AUC) of 0.89 in the test set among 28 models when combined with B-US + CDFI + RTE + CEUS data and an AUC of 0.72 and 0.66 when combined with B-US and B-US + CDFI + RTE data. The AUC achieved by senior and junior radiologists was 0.78 versus (vs.) 0.69 (p > 0.05), 0.79 vs. 0.64 (p < 0.05), and 0.88 vs. 0.69 (p < 0.05) combined with B-US, B-US+CDFI+RTE and B-US+CDFI+RTE+CEUS, respectively. With the addition of CEUS, the diagnostic performance was enhanced for all seven classifiers and senior radiologists based on conventional ultrasound images, while no enhancement was observed for junior radiologists. The diagnostic performance of ML models was similar to senior radiologists, but superior to those junior radiologists.

Ultrasound radiomics based XGBoost model to differential diagnosis thyroid nodules and unnecessary biopsy rate: Individual application of SHapley additive exPlanations.

Radiomics-based eXtreme gradient boosting (XGBoost) model was developed to differentiate benign thyroid nodules from malignant thyroid nodules and to prevent unnecessary thyroid biopsies, including positive and negative effects. The study evaluated a data set of ultrasound images of thyroid nodules in patients retrospectively, who initially received ultrasound-guided fine-needle aspiration biopsy (FNAB) for diagnostic purposes. According to ACR TI-RADS, a total of five ultrasound feature categories and the maximum size of the nodule were determined by four radiologists. A radiomics score was developed by the LASSO algorithm from the ultrasound-based radiomics features. An interpretative method based on Shapley additive explanation (SHAP) was developed. XGBoost was compared with ACR TI-RADS for its diagnostic performance and FNAB rate and was compared with six other machine learning models to evaluate the model performance. Finally, 191 thyroid nodules were examined from 177 patients. The radiomics score were calculated using 8 features, which were selected among 789 candidate features generated from the ultrasound images. The model yielded an AUC of 93% in the training cohort and 92% in the test cohort. It outperformed traditional machine learning models in assessing the nature of thyroid nodules. Compared with ACR TI-RADS, the FNAB rate decreased from 34% to 30% in training and from 35% to 41% in test. The radiomics-based XGBoost model proposed could distinguish benign and malignant thyroid nodules, thereby reduced significantly the number of unnecessary FNAB. It was effective in making preoperative decisions and managing selected patients using the SHAP visual interpretation tools.

Automatic Detection of Thyroid Nodule Characteristics From 2D Ultrasound Images.

Thyroid cancer is one of the common types of cancer worldwide, and Ultrasound (US) imaging is a modality normally used for thyroid cancer diagnostics. The American College of Radiology Thyroid Imaging Reporting and Data System (ACR TIRADS) has been widely adopted to identify and classify US image characteristics for thyroid nodules. This paper presents novel methods for detecting the characteristic descriptors derived from TIRADS. Our methods return descriptions of the nodule margin irregularity, margin smoothness, calcification as well as shape and echogenicity using conventional computer vision and deep learning techniques. We evaluate our methods using datasets of 471 US images of thyroid nodules acquired from US machines of different makes and labeled by multiple radiologists. The proposed methods achieved overall accuracies of 88.00%, 93.18%, and 89.13% in classifying nodule calcification, margin irregularity, and margin smoothness respectively. Further tests with limited data also show a promising overall accuracy of 90.60% for echogenicity and 100.00% for nodule shape. This study provides an automated annotation of thyroid nodule characteristics from 2D ultrasound images. The experimental results showed promising performance of our methods for thyroid nodule analysis. The automatic detection of correct characteristics not only offers supporting evidence for diagnosis, but also generates patient reports rapidly, thereby decreasing the workload of radiologists and enhancing productivity.

A multi-task model for reliable classification of thyroid nodules in ultrasound images.

Thyroid nodules are common, and patients with potential malignant lesions are usually diagnosed using ultrasound imaging to determine further treatment options. This study aims to propose a computer-aided diagnosis method for benign and malignant classification of thyroid nodules in ultrasound images. We propose a novel multi-task framework that combines the advantages of dense connectivity, Squeeze-and-Excitation (SE) connectivity, and Atrous Spatial Pyramid Pooling (ASPP) layer to enhance feature extraction. The Dense connectivity is used to optimize feature reuse, the SE connectivity to optimize feature weights, the ASPP layer to fuse feature information, and a multi-task learning framework to adjust the attention of the network. We evaluate our model using a 10-fold cross-validation approach based on our established Thyroid dataset. We assess the performance of our method using six average metrics: accuracy, sensitivity, specificity, positive predictive value, negative predictive value, and AUC, which are 93.49, 95.54, 91.52, 91.63, 95.47, and 96.84%, respectively. Our proposed method outperforms other classification networks in all metrics, achieving optimal performance. We propose a multi-task model, DSMA-Net, for distinguishing thyroid nodules in ultrasound images. This method can further enhance the diagnostic ability of doctors for suspected cancer patients and holds promise for clinical applications.

Automatic ultrasound diagnosis of thyroid nodules: a combination of deep learning and KWAK TI-RADS

Objective. There has been a considerable amount of computer-aided diagnosis (CAD) methods highlighted in the field of ultrasonic examination (USE) of thyroid nodules. However, few researches focused on the automatic risk classification, which was the basis for determining whether fine needle aspiration (FNA) was needed. The aim of this work was to implement automatic risk level assessment of thyroid nodules. Approach. Firstly, 1862 cases of thyroid nodules with the results of USE and FNA were collected as the dataset. Then, an improved U-Net++ model was utilized for segmenting thyroid nodules in ultrasound images automatically. Finally, the segmentation result was imported into a multi-task convolutional neural network (MT-CNN), the design of which was based on the clinical guideline called KWAK TI-RADS. Apart from the category of benign and malignant, the MT-CNN also exported the classification result of four malignant features, solid component (SC), hypoechogenicity or marked hypoechogenicity (HMH), microlobulated or irregular margin (MIM), microcalcification (MC), which were used for counting the risk level in KWAK TI-RADS. Main results. The performance of the improved U-Net++ was evaluated on our test set, including 302 cases. The Dice coefficient and intersection over union reached 0.899, 0.816, respectively. The classification accuracy rates of SC, HMH, MIM, MC, were 94.5%, 92.8%, 86.1%, 88.9%, while the false positive (FP) rate was 6.0%, 5.6%, 10.6%, 12.9% respectively. As for the category of benign and malignant, the precision and recall rates were 93.7% and 94.4%. Significance. The proposed CAD method showed favourable performance in the diagnosis of thyroid nodules. Compared with other methods, it could provide reports closer to clinical practice for doctors.

TS-DSANN: Texture and shape focused dual-stream attention neural network for benign-malignant diagnosis of thyroid nodules in ultrasound images.

Recently, accurate diagnosis of thyroid nodules has played a critical role in precision medicine and healthcare system management. Due to complicated changes in ultrasound features of texture, and similar visual appearance of benign-malignant nodules, the identification of cancerous thyroid lesions from a given ultrasound image still faces challenges for even experienced radiologists. Learning-based computer-aided diagnosis (CAD) systems have accordingly attracted more and more attention recently. However, little research is committed to developing a deep learning-based CAD system that has greater conformity with radiologists' diagnostic decision-making. In this study, we devise a texture and shape focused dual-stream attention neural network, dubbed TS-DSANN. Specifically, in the texture focused stream, we utilize the ImageNet pre-trained ResNet34 to guide the network to recognize texture-related nodule attributes. Meanwhile, in the shape focused stream, in addition to using ResNet34 backbone, jointly learning from scratch with the contour obtained by contour detection module to enhance the extraction of shape features. Afterward, we employ a concatenation operation to aggregate the abovementioned two stream networks for capturing richer and more representative features. Finally, we further utilize an online class activation mapping mechanism to assist the dual-stream network in generating a localization heatmap to obtain more visualization attention to the nodule from the whole image, and supervise classifier's attention in decision-making. Experimental results conducted on the two-center thyroid nodule ultrasound datasets verify that our proposed method has improved the classification performance, superior to the state-of-the-art methods.