ARB U-Net: An Improved Neural Network for Suprapatellar Bursa Effusion Ultrasound Image Segmentation

To develop and validate a deep learning radiomics fusion model (DLR) based on ultrasound (US) images to identify bone erosion in rheumatoid arthritis (RA) patients. A total of 432 patients with RA at two institutions were collected. Three hundred twelve patients from center 1 were randomly divided into a training set (N = 218) and an internal test set (N = 94) in a 7:3 ratio; meanwhile, 124 patients from center 2 were as an external test set. Radiomics (Rad) and deep learning (DL) features were extracted based on hand-crafted radiomics and deep transfer learning networks. The least absolute shrinkage and selection operator regression was employed to establish DLR fusion feature from the Rad and DL features. Subsequently, 10 machine learning algorithms were used to construct models and the final optimal model was selected. The performance of models was evaluated using receiver operating characteristic (ROC) and decision curve analysis (DCA). The diagnostic efficacy of sonographers was compared with and without the assistance of the optimal model. LR was chosen as the optimal algorithm for model construction account for superior performance (Rad/DL/DLR: area under the curve [AUC] = 0.906/0.974/0.979) in the training set. In the internal test set, DLR_LR as the final model had the highest AUC (AUC = 0.966), which was also validated in the external test set (AUC = 0.932). With the aid of DLR_LR model, the overall performance of both junior and senior sonographers improved significantly (P < 0.05), and there was no significant difference between the junior sonographer with DLR_LR model assistance and the senior sonographer without assistance (P > 0.05). DLR model based on US images is the best performer and is expected to become an important tool for identifying bone erosion in RA patients. Key Points • DLR model based on US images is the best performer in identifying BE in RA patients. • DLR model may assist the sonographers to improve the accuracy of BE evaluations.

The objective of this retrospective study is to develop and validate an artificial intelligence model constrained by the anatomical structure of the brain with the aim of improving the accuracy of prenatal diagnosis of fetal cerebellar hypoplasia using ultrasound imaging. Fetal central nervous system dysplasia is one of the most prevalent congenital malformations, and cerebellar hypoplasia represents a significant manifestation of this anomaly. Accurate clinical diagnosis is of great importance for the purpose of prenatal screening of fetal health. Although ultrasound has been extensively utilized to assess fetal development, the accurate assessment of cerebellar development remains challenging due to the inherent limitations of ultrasound imaging, including low resolution, artifacts, and acoustic shadowing of the skull. This retrospective study included 302 cases diagnosed with cerebellar hypoplasia and 549 normal pregnancies collected from Maternal and Child Health Hospital of Hubei Province between September 2019 and September 2023. For each case, experienced ultrasound physicians selected appropriate brain ultrasound images to delineate the boundaries of the skull, cerebellum, and cerebellomedullary cistern. These cases were divided into one training set and two test sets, based on the examination dates. This study then proposed a dual-branch deep learning classification network, anatomical structure-constrained network (ASC-Net), which took ultrasound images and anatomical structure masks as separate inputs. The performance of the ASC-Net was extensively evaluated and compared with several state-of-the-art deep learning networks. The impact of anatomical structures on the performance of ASC-Net was carefully examined. ASC-Net demonstrated superior performance in the diagnosis of cerebellar hypoplasia, achieving classification accuracies of 0.9778 and 0.9222, as well as areas under the receiver operating characteristic curve of 0.9986 and 0.9265 on the two test sets. These results significantly outperformed several state-of-the-art networks on the same dataset. In comparison to other studies on cerebellar hypoplasia auxiliary diagnosis, ASC-Net also demonstrated comparable or even better performance. A subgroup analysis revealed that ASC-Net was more capable of distinguishing cerebellar hypoplasia in cases with gestational weeks greater than 30weeks. Furthermore, when constrained by anatomical structures of both the cerebellum and cistern, ASC-Net exhibited the best performance compared to other kinds of structural constraint. The development and validation of ASC-Net have significantly enhanced the accuracy of prenatal diagnosis of cerebellar hypoplasia using ultrasound images. This study highlights the importance of anatomical structures of the fetal cerebellum and cistern on the performance of the diagnostic artificial intelligence model in ultrasound. This might provide new insights for clinical diagnosis of cerebellar hypoplasia, assist clinicians in providing more targeted advice and treatment during pregnancy, and contribute to improved perinatal healthcare. ASC-Net is open-sourced and publicly available in a GitHub repository at https://github.com/Wwwwww111112/ASC-Net .

Simple SummaryIn this study, deep learning combined with udder ultrasonography of buffalo was used for the detection of mastitis for the first time, with the aim of establishing an accurate, rapid, and inexpensive method to detect buffalo mastitis instead of routine laboratory examination. This method provides a basis for mastitis detection in buffaloes mostly raised by small farmers and has the opportunity to be used in a variety of dairy animals in the future.Mastitis is one of the most predominant diseases with a negative impact on ranch products worldwide. It reduces milk production, damages milk quality, increases treatment costs, and even leads to the premature elimination of animals. In addition, failure to take effective measures in time will lead to widespread disease. The key to reducing the losses caused by mastitis lies in the early detection of the disease. The application of deep learning with powerful feature extraction capability in the medical field is receiving increasing attention. The main purpose of this study was to establish a deep learning network for buffalo quarter-level mastitis detection based on 3054 ultrasound images of udders from 271 buffaloes. Two data sets were generated with thresholds of somatic cell count (SCC) set as 2 × 105 cells/mL and 4 × 105 cells/mL, respectively. The udders with SCCs less than the threshold value were defined as healthy udders, and otherwise as mastitis-stricken udders. A total of 3054 udder ultrasound images were randomly divided into a training set (70%), a validation set (15%), and a test set (15%). We used the EfficientNet_b3 model with powerful learning capabilities in combination with the convolutional block attention module (CBAM) to train the mastitis detection model. To solve the problem of sample category imbalance, the PolyLoss module was used as the loss function. The training set and validation set were used to develop the mastitis detection model, and the test set was used to evaluate the network’s performance. The results showed that, when the SCC threshold was 2 × 105 cells/mL, our established network exhibited an accuracy of 70.02%, a specificity of 77.93%, a sensitivity of 63.11%, and an area under the receiver operating characteristics curve (AUC) of 0.77 on the test set. The classification effect of the model was better when the SCC threshold was 4 × 105 cells/mL than when the SCC threshold was 2 × 105 cells/mL. Therefore, when SCC ≥ 4 × 105 cells/mL was defined as mastitis, our established deep neural network was determined as the most suitable model for farm on-site mastitis detection, and this network model exhibited an accuracy of 75.93%, a specificity of 80.23%, a sensitivity of 70.35%, and AUC 0.83 on the test set. This study established a 1/4 level mastitis detection model which provides a theoretical basis for mastitis detection in buffaloes mostly raised by small farmers lacking mastitis diagnostic conditions in developing countries.

The risk of malignancy in cystic-solid thyroid nodules (CSTN) varies greatly and may be underestimated. This study aimed to explore the value of a comprehensive model that integrates deep transfer learning (DTL), ultrasound radiomics, and clinical, and ultrasound features in predicting the risk of malignancy of CSTN. A retrospective analysis was conducted on 278 patients with CSTN confirmed by pathology from the First Affiliated Hospital of Guangxi Medical University from January 2023 to December 2023. Radiomics features were manually extracted from ultrasound images, and DTL features were extracted using deep learning networks. The least absolute shrinkage and selection operator (LASSO) regression was utilized to select non-zero coefficient features from radiomics and DTL features. The comprehensive model nomogram was constructed using a logistic regression algorithm that integrates clinical, ultrasound features, deep learning, and radiomics features. The predictive performance was assessed using the receiver operating characteristic (ROC) curve and the area under the curve (AUC), calibration curves, and decision curves. Subsequently, DeLong testing was performed for comparative analysis of the AUC, with parameter estimates including a 95% confidence interval (CI), and a P value of less than 0.05 was considered statistically significant. The AUC of each model was compared, revealing that the comprehensive model outperformed the individual models in predicting the malignancy risk of CSTN, demonstrating good predictive performance with sensitivity and specificity of 87.50% and 82.90%, respectively. Additionally, the AUC of the comprehensive model in the testing set was 0.913 (95% CI: 0.844-0.982), which was higher than the radiomics model (0.913 vs. 0.898, P=0.67), and the DTL model (0.913 vs. 0.848, P=0.38). In the training set, the AUC was 0.973 (95% CI: 0.949-0.997), outperforming the radiomics model (0.973 vs. 0.926, P=0.09) and the DTL model (0.973 vs. 0.943, P=0.01). The novel comprehensive model based on ultrasound demonstrates excellent performance in predicting the malignancy risk of CSTN, providing clinicians with a preoperative non-invasive screening method to predict the malignancy risk of CSTN.

High-resolution magnetic resonance vessel wall imaging (HR-VWI) offers enhanced visualization of vascular structures, thereby facilitating the deep learning (DL) network's acquisition of more extensive and detailed image information. This study aimed to develop a high-precision integrated model leveraging DL with an attention mechanism based on HR-VWI for predicting recurrent stroke in patients with symptomatic intracranial atherosclerotic stenosis (sICAS). A retrospective study was conducted involving 363 sICAS patients who underwent HR-VWI, with data divided into a training set (n=254) from Center 1 (The First Affiliated Hospital of Xinxiang Medical University) and a test set (n=109) from Center 2 (The Sixth People's Hospital of Shanghai Jiao Tong University). Two convolutional neural network (CNN) models, ResNet50 and DenseNet169, were employed as feature extractors to capture image information from culprit plaques in HR-VWI. Integrating the Transformer attention mechanism, an advanced ensemble model, Trans-CNN, was constructed to predict stroke recurrence in sICAS patients. Model performance was evaluated using receiver operating characteristic (ROC) curves, with DeLong's test for comparing models. Additionally, decision curve analysis (DCA) and calibration curves were utilized to assess the model's practical and clinical value. Trans-CNN demonstrated superior predictive performance, outperforming other models in both the training and test sets. Specifically, in the training set, Trans-CNN achieved an area under the curve (AUC) of 0.951 [95% confidence interval (CI): 0.923-0.974], accuracy of 0.880 (95% CI: 0.797-0.937), sensitivity of 0.900 (95% CI: 0.836-1.000), and specificity of 0.882 (95% CI: 0.757-0.948). Similarly, in the test set, it achieved an AUC of 0.912 (95% CI: 0.839-0.969), accuracy of 0.858 (95% CI: 0.743-0.936), sensitivity of 0.880 (95% CI: 0.693-1.000), and specificity of 0.810 (95% CI: 0.690-0.976). The AUC improvement of Trans-CNN over all other models was statistically significant (DeLong's test, P<0.05). Calibration curve analysis revealed good agreement between predicted probabilities and observed outcomes in both sets. DCA further underscored the potential value of Trans-CNN in guiding clinical decision-making. The integrated model combining DL with an attention mechanism based on HR-VWI exhibits excellent performance in assessing the risk of stroke recurrence in sICAS patients. This advancement holds significant potential in assisting clinicians in diagnosis and developing individualized treatment strategies.

ObjectivesThis study aimed to differentially diagnose thyroid nodules (TNs) of Thyroid Imaging Reporting and Data System (TI-RADS) 3–5 categories using a deep learning (DL) model based on multimodal ultrasound (US) images and explore its auxiliary role for radiologists with varying degrees of experience.MethodsPreoperative multimodal US images of 1,138 TNs of TI-RADS 3–5 categories were randomly divided into a training set (n = 728), a validation set (n = 182), and a test set (n = 228) in a 4:1:1.25 ratio. Grayscale US (GSU), color Doppler flow imaging (CDFI), strain elastography (SE), and region of interest mask (Mask) images were acquired in both transverse and longitudinal sections, all of which were confirmed by pathology. In this study, fivefold cross-validation was used to evaluate the performance of the proposed DL model. The diagnostic performance of the mature DL model and radiologists in the test set was compared, and whether DL could assist radiologists in improving diagnostic performance was verified. Specificity, sensitivity, accuracy, positive predictive value, negative predictive value, and area under the receiver operating characteristics curves (AUC) were obtained.ResultsThe AUCs of DL in the differentiation of TNs were 0.858 based on (GSU + SE), 0.909 based on (GSU + CDFI), 0.906 based on (GSU + CDFI + SE), and 0.881 based (GSU + Mask), which were superior to that of 0.825-based single GSU (p = 0.014, p< 0.001, p< 0.001, and p = 0.002, respectively). The highest AUC of 0.928 was achieved by DL based on (G + C + E + M)US, the highest specificity of 89.5% was achieved by (G + C + E)US, and the highest accuracy of 86.2% and sensitivity of 86.9% were achieved by DL based on (G + C + M)US. With DL assistance, the AUC of junior radiologists increased from 0.720 to 0.796 (p< 0.001), which was slightly higher than that of senior radiologists without DL assistance (0.796 vs. 0.794, p > 0.05). Senior radiologists with DL assistance exhibited higher accuracy and comparable AUC than that of DL based on GSU (83.4% vs. 78.9%, p = 0.041; 0.822 vs. 0.825, p = 0.512). However, the AUC of DL based on multimodal US images was significantly higher than that based on visual diagnosis by radiologists (p< 0.05).ConclusionThe DL models based on multimodal US images showed exceptional performance in the differential diagnosis of suspicious TNs, effectively increased the diagnostic efficacy of TN evaluations by junior radiologists, and provided an objective assessment for the clinical and surgical management phases that follow.

Kidney Disease (KD) is characterized by a gradual decline in kidney function, which can eventually lead to kidney damage or failure. As the disease progresses, diagnosis becomes more challenging. Incorporating routine clinical data to assess different stages of KD can aid in early detection and timely intervention. Advanced stages of KD are associated with a higher risk of cardiovascular complications and mortality. Ultrasound (US) imaging is widely used in clinical practice for predicting KD due to its safety, convenience, and affordability. However, manual analysis of US images is time-consuming, prone to errors, and requires highly skilled professionals. In recent years, Deep Learning (DL) has shown promising results in medical image analysis. This research introduces a hybrid DL network, Convolutional Neural Network (CNN)-Transformer, designed to predict KD from US images. To conduct the study, US images of both healthy and diseased kidneys were collected from Aadhar Diagnostic Centre, Maharashtra. The collected raw images underwent several pre-processing steps, including resizing and augmentation. The processed dataset was then split into training, validation, and test sets in a 7:2:1 ratio. The proposed hybrid network was compared with well-known DL networks, ResNet and DenseNet. All three models were trained, validated, and tested under identical conditions, including the same number of images, epochs, and hyperparameters, to ensure a fair comparison. The models were tested on 25 healthy and 25 diseased images. The results showed that DenseNet and ResNet correctly predicted 44 and 43 cases, respectively, while the proposed CNN-Transformer network achieved 49 correct predictions out of 50 samples. The proposed network attained the highest accuracy of 98%, whereas DenseNet and ResNet achieved 88% and 86%, respectively. In addition to accuracy, other evaluation metrics, including Precision, Recall, and F1-Score, were also significantly higher for the proposed network compared to the other two networks. These findings demonstrate that the proposed CNN-Transformer network delivers promising results for KD prediction using US images.

The pathological classification and imaging manifestation of parotid gland tumors are complex, while accurate preoperative identification plays a crucial role in clinical management and prognosis assessment. This study aims to construct and compare the performance of clinical models, traditional radiomics models, deep learning (DL) models, and deep learning radiomics (DLR) models based on ultrasound (US) images in differentiating between benign parotid gland tumors (BPGTs) and malignant parotid gland tumors (MPGTs). Retrospective analysis was conducted on 526 patients with confirmed PGTs after surgery, who were randomly divided into a training set and a testing set in the ratio of 7:3. Traditional radiomics and three DL models (DenseNet121, VGG19, ResNet50) were employed to extract handcrafted radiomics (HCR) features and DL features followed by feature fusion. Seven machine learning classifiers including logistic regression (LR), support vector machine (SVM), RandomForest, ExtraTrees, XGBoost, LightGBM and multi-layer perceptron (MLP) were combined to construct predictive models. The most optimal model was integrated with clinical and US features to develop a nomogram. Receiver operating characteristic (ROC) curve was employed for assessing performance of various models while the clinical utility was assessed by decision curve analysis (DCA). The DLR model based on ExtraTrees demonstrated superior performance with AUC values of 0.943 (95% CI: 0.918-0.969) and 0.916 (95% CI: 0.861-0.971) for the training and testing set, respectively. The combined model DLR nomogram (DLRN) further enhanced the performance, resulting in AUC values of 0.960 (95% CI: 0.940- 0.979) and 0.934 (95% CI: 0.876-0.991) for the training and testing sets, respectively. DCA analysis indicated that DLRN provided greater clinical benefits compared to other models. DLRN based on US images shows exceptional performance in distinguishing BPGTs and MPGTs, providing more reliable information for personalized diagnosis and treatment plans in clinical practice.

Deep Learning for Distinguishing Mucinous Breast Carcinoma From Fibroadenoma on Ultrasound

Biometrics based personal verification for mobile phone devices are currently well-known. In this study, a verification approach is suggested depending on fingerphoto pictures. Couple of Deep Fingerphotos Learning (CDFL) approach is proposed, where two Deep Learning (DL) networks are involved. A fingerphoto picture of the index finger is verified using the first DL network. To recognize a fingerphoto picture of a middle finger, another DL network is used. Then, the outputs of the two networks are integrated. Fingerphoto pictures from the IIITD smartphone fingerphoto dataset are used in this work. The results yield that the accuracy of the first DL network is reported as 76.95% and the accuracy of the second DL network is reported as 86.33%. Whereas, the overall accuracy of the proposed CDFL method after integrating both DL networks is benchmarked as 96.48%.

Deep learning has shown considerable promise in the differential diagnosis of lung lesions. However, the majority of previous studies have focused primarily on X-ray, computed tomography (CT), and magnetic resonance imaging (MRI), with relatively few investigations exploring the predictive value of ultrasound imaging. This study aims to develop a deep learning model based on ultrasound imaging to differentiate between benign and malignant peripheral lung tumors. A retrospective analysis was conducted on a cohort of 371 patients who underwent ultrasound-guided percutaneous lung tumor procedures across two centers. The dataset was divided into a training set (n = 296) and a test set (n = 75) in an 8:2 ratio for further analysis and model evaluation. Five distinct deep learning models were developed using ResNet152, ResNet101, ResNet50, ResNet34, and ResNet18 algorithms. Receiver Operating Characteristic (ROC) curves were generated, and the Area Under the Curve (AUC) was calculated to assess the diagnostic performance of each model. DeLong's test was employed to compare the differences between the groups. Among the five models, the one based on the ResNet18 algorithm demonstrated the highest performance. It exhibited statistically significant advantages in predictive accuracy (p < 0.05) compared to the models based on ResNet152, ResNet101, ResNet50, and ResNet34 algorithms. Specifically, the ResNet18 model showed superior discriminatory power. Quantitative evaluation through Net Reclassification Improvement (NRI) analysis revealed that the NRI values for the ResNet18 model, when compared with ResNet152, ResNet101, ResNet50, and ResNet34, were 0.180, 0.240, 0.186, and 0.221, respectively. All corresponding p-values were less than 0.05 (p < 0.05 for each comparison), further confirming that the ResNet18 model significantly outperformed the other four models in reclassification ability. Moreover, its predictive outcomes led to marked improvements in risk stratification and classification accuracy. The ResNet18-based deep learning model demonstrated superior accuracy in distinguishing between benign and malignant peripheral lung tumors, providing an effective and non-invasive tool for the early detection of lung cancer.

Segmentation of knee cartilage from Ultrasound (US) images is essential for various clinical tasks in diagnosis and treatment planning of Osteoarthritis. Moreover, the potential use of US imaging for guidance in robotic knee arthroscopy is presently being investigated. The femoral cartilage being the main organ at risk during the operation, it is paramount to be able to segment this structure, to make US guidance feasible. In this paper, we set forth a deep learning network, Mask R-CNN, based femoral cartilage segmentation in 2D US images for these types of applications. While the traditional imaging approaches showed promising results, they are mostly not real-time and involve human interaction. This being the case, in recent years, deep learning has paved its way into medical imaging showing commendable results. However, deep learning-based segmentation in US images remains unexplored. In the present study we employ Mask R-CNN on US images of the knee cartilage. The performance of the method is analyzed in various scenarios, with and without Gaussian filter preprocessing and pretraining the network with different datasets. The best results are observed when the images are preprocessed and the network is pretrained with COCO 2016 image dataset. A maximum Dice Similarity Coefficient (DSC) of 0.88 and an average DSC of 0.80 is achieved when tested on 55 images indicating that the proposed method has a potential for clinical applications.

Objectives: The research aims to enhance breast cancer detection accuracy and effectiveness using deep transfer learning and pre-trained neural networks. It analyses breast ultrasound images and identifies important characteristics using pre-trained networks. The goal is to create a more efficient and accurate automated system for breast cancer detection. Methods: The study uses breast ultrasound cancer image data from the Kaggle Data Repository to extract informative features, identify cancer-related characteristics, and classify them into benign, malignant, and normal tissue. Pre-trained Deep Neural Networks (DNNs) extract these features and feed them into a 10-fold cross-validation SVM classifier. The SVM is evaluated using various kernel functions to identify the best kernel for separating data points. This methodology aims to achieve accurate classification of breast cancer in ultrasound images. Findings: The study confirms the effectiveness of deep transfer learning for breast cancer detection in ultrasound images, with Inception V3 outperforming VGG-16 and VGG-19 in extracting relevant features. The combination of Inception V3 and the SVM classifier with a polynomial kernel achieved the highest classification accuracy, indicating its ability to model complex relationships. The study demonstrated an AUC of 0.944 and a classification accuracy of 87.44% using the Inception V3 + SVM polynomial. Novelty: This research demonstrates the potential of deep transfer learning and SVM classifiers for accurate breast cancer detection in ultrasound images. It integrates Inception V3, VGG-16, and VGG-19 for breast cancer detection, demonstrating improved classification accuracy. The combination of Inception V3 and SVM (polynomial) achieved a significant AUC (0.944) and classification accuracy (87.44%), outperforming other models tested. This research underscores the potential of these technologies for accurate breast cancer detection in ultrasound images. Keywords: Breast Cancer, Deep Learning, Feature Extraction, Inception-v3, SVM, Transfer Learning

Predicting Atrial Fibrillation from Automated Measurements of Left Atrial Volume Using Routine Chest CT Examination: Overlooked and Underrecognized Risk Factors.

Functional adrenal tumors (FATs) are mainly diagnosed by biochemical analysis. Traditional imaging tests have limitations and cannot be used alone to diagnose FATs. In this study, we aimed to establish an artificially intelligent diagnostic model based on computed tomography (CT) images to distinguish different types of FATs. A cohort study of 375 patients diagnosed with hyperaldosteronism (HA), Cushing's syndrome (CS), and pheochromocytoma in our center between March 2015 and June 2020 was conducted. Retrospectively, patients were randomly divided into three data sets: the training set (270 cases), the testing set (60 cases), and the retrospective trial set (45 cases). An artificially intelligent diagnostic model based on CT images was established by transferring data from the training set into the deep learning network. The testing set was then used to evaluate the accuracy of the model compared to that of physicians' judgments. The retrospective trial set was used to evaluate the quantification and distinction performance. The deep learning model achieved an average area under the receiver operating characteristic (ROC) curve (AUC) of 0.915, and the AUCs in all three FAT types were greater than 0.882. The AUC of the model tested on the retrospective dataset reached above 0.849. In the quantitative evaluation of tumor lesion area recognition, the diagnostic model also obtained a segmentation Dice coefficient of 0.69. With the help of the proposed model, clinicians reached 92.5% accuracy in distinguishing FATs, compared to 80.6% accuracy when using only their judgment (P<0.05). The result of our study shows that the diagnostic model based on a deep learning network can distinguish and quantify three common FAT types based on texture features of contrast-enhanced CT images. The model can quantify and distinguish functional tumors without any endocrine tests and can assist clinicians in the diagnostic procedure.