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).

Effects of heterogeneous reaction with NO2 on ice nucleation activities of feldspar and Arizona Test Dust.

Mineral dust is an important type of ice nucleating particles in the troposphere; however, the effects of heterogeneous reactions on ice nucleation (IN) activities of mineral dust remain to be elucidated. A droplet-freezing apparatus (Guangzhou Institute of Geochemistry Ice Nucleation Apparatus, GIGINA) was developed in this work to measure IN activities of atmospheric particles in the immersion freezing mode, and its performance was validated by a series of experimental characterizations. This apparatus was then employed to measure IN activities of feldspar and Arizona Test Dust (ATD) particles before and after heterogeneous reaction with NO2 (10±0.5 ppmv) at 40% relative humidity. The surface coverage of nitrate, θ(NO3-), increased to 3.1±0.2 for feldspar after reaction with NO2 for 6 hr, and meanwhile the active site density per unit surface area (ns) at -20°C was reduced from 92±5 to <1.0 cm-2 by about two orders of magnitude; however, no changes in nitrate content or IN activities were observed for further increase in reaction time (up to 24 hr). Both nitrate content and IN activities changed continuously with reaction time (up to 24 hr) for ATD particles; after reaction with NO2 for 24 hr, θ(NO3-) increased to 1.4±0.1 and ns at -20°C was reduced from 20±4 to 9.7±1.9 cm-2 by a factor of ∼2. Our work suggests that heterogeneous reaction with NO2, an abundant reactive nitrogen species in the troposphere, may significantly reduce IN activities of mineral dust in the immersion freezing mode.

Heterogeneous reaction of NO2 with feldspar, three clay minerals and Arizona Test Dust.

Heterogeneous reaction of NO2 with mineral dust aerosol may play important roles in troposphere chemistry, and has been investigated by a number of laboratory studies. However, the influence of mineralogy on this reaction has not been well understood, and its impact on aerosol hygroscopicity is not yet clear. This work investigated heterogeneous reactions of NO2 (∼10 ppmv) with K-feldspar, illite, kaolinite, montmorillonite and Arizona Test Dust (ATD) at room temperature as a function of relative humidity (<1% to 80%) and reaction time (up to 24 hr). Heterogeneous reactivity towards NO2 was low for illite, kaolinite, montmorillonite and ATD, and uptake coefficients of NO2, γ(NO2), were determined to be around or smaller than 1×10-8; K-feldspar exhibited higher reactivity towards NO2, and CaCO3 is most reactive among the nine mineral dust samples considered in this and previous work. After heterogeneous reaction with NO2 for 24 hr, increase in hygroscopicity was nearly insignificant for illite, kaolinite and montmorillonite, and small but significant for K-feldspar; in addition, large increase in hygroscopicity was observed for ATD, although the increase in hygroscopicity was still smaller than CaCO3.

An Improved Pedestrian Ttracking Method Based on Wi-Fi Fingerprinting and Pedestrian Dead Reckoning.

Wi-Fi based positioning has great potential for use in indoor environments because Wi-Fi signals are near-ubiquitous in many indoor environments. With a Reference Fingerprint Map (RFM), fingerprint matching can be adopted for positioning. Much assisting information can be adopted for increasing the accuracy of Wi-Fi based positioning. One of the most adopted pieces of assisting information is the Pedestrian Dead Reckoning (PDR) information derived from inertial measurements. This is widely adopted because the inertial measurements can be acquired through a Commercial Off The Shelf (COTS) smartphone. To integrate the information of Wi-Fi fingerprinting and PDR information, many methods have adopted filters, such as Kalman filters and particle filters. A new methodology for integration of Wi-Fi fingerprinting and PDR is proposed using graph optimization in this paper. For the Wi-Fi based fingerprinting part, our method adopts the state-of-art hierarchical structure and the Penalized Logarithmic Gaussian Distance (PLGD) metric. In the integration part, a simple extended Kalman filter (EKF) is first used for integration of Wi-Fi fingerprinting and PDR results. Then, the tracking results are adopted as initial values for the optimization block, where Wi-Fi fingerprinting and PDR results are adopted to form an concentrated cost function (CCF). The CCF can be minimized with the aim of finding the optimal poses of the user with better tracking results. With both real-scenario experiments and simulations, we show that the proposed method performs better than classical Kalman filter based and particle filter based methods with both less average and maximum positioning error. Additionally, the proposed method is more robust to outliers in both Wi-Fi based and PDR based results, which is commonly seen in practical situations.

LncRNA PRAL is closely related to clinical prognosis of multiple myeloma and the bortezomib sensitivity.

Chromosome 17p deletions (del(17p)) are present in about 11% of newly diagnosed multiple myeloma (MM) patients and related to inferior prognosis. Bortezomib (BTZ), the first proteasome inhibitor anticancer drug, has a good therapeutic effect for newly diagnosed, relapsed or refractory MM, but is unable to improve the outcome of MM patients with del(17p). Long noncoding RNA (lncRNA) PRAL is located on chromosome 17p, and is associated with the progression and prognosis of different types of cancers. However, little is known about its role in MM. Here, we found that PRAL was downregulated in primary MM cells and cell lines, especially in MM cells with del(17p), and was associated with ISS (international staging system) stage and Durie-Salmon stage in MM patients. Survival curves showed that MM patients with low PRAL expression had a significantly shorter disease-free survival (DFS) and overall survival (OS) than the patients with high PRAL expression. Multivariate Cox regression analysis showed that PRAL expression was an independent predictor for DFS and OS. Then cell proliferation, viability, Ki67 expression, and caspase-3 activity detection showed that PRAL promoted MM cell growth inhibition and apoptosis, and potentiated the anti-MM effect of BTZ in vitro. We further identified and confirmed that miR-210 was the target of PRAL, and miR-210 overexpression overturned the potentiation effect of PRAL on BTZ efficacy. Subsequently, bone morphogenetic protein 2 (BMP2) was confirmed to be the target of miR-210, and PRAL positively regulated the derepression of BMP2 by sponging miR-210. Overexpression of BMP2 potentiated the anti-MM effect of BTZ in vitro. In addition, animal experiments further confirmed that PRAL potentiated BTZ efficacy in vivo. Collectively, our study first revealed a critical role of PRAL-miR-210-BMP2 axis in MM progression, prognosis and treatment with BTZ, and PRAL could become a novel diagnostic, prognostic and therapeutic candidate for MM patients especially for the MM patients harboring del(17p) in the future.

Preoperative Axillary Ultrasound in the Selection of Patients With a Heavy Axillary Tumor Burden in Early-Stage Breast Cancer: What Leads to False-Positive Results?

OBJECTIVES: To determine whether imaging and clinicopathologic features could predict false-positive axillary ultrasound (US) results in the selection of patients with breast cancer who had a heavy axillary tumor burden (≥3 tumor-involved nodes). METHODS: Among 788 patients with histologically confirmed invasive breast cancer at Ruijin Hospital from October 2014 to September 2015, 162 patients (cT1-T2, cN0) with 167 axillae had suspicious axillary US findings. Ultrasound findings were considered suspicious for metastasis if cortical thickening of greater than 3 mm or effacement of the fatty hilum was present. The false-positive rate of suspicious axillary US results for identifying 3 or more positive lymph nodes in the final pathologic examination was calculated. Univariate and multivariate analyses were used to evaluate imaging and clinicopathologic factors related to the false-positive results. RESULTS: Axillary US showed a false-positive rate of 60.5% (101 of 167) in the patients with breast cancer and a heavy nodal burden. By logistic regression analyses, we found false-positive axillary US results more frequently in patients who had a T1 stage tumor (P = .005), an estrogen receptor/progesterone receptor-negative tumor (P < .001), solitary suspicious nodes identified on axillary US (P < .001), and a cortical thickness of the most suspicious lymph node of 3.5 mm or less (P = .015). CONCLUSIONS: Imaging and clinicopathologic features can be used to identify axillae with less than 3 metastatic nodes in patients with early-stage breast cancer who have positive axillary US results. In the post-American College of Surgeons Oncology Group Z0011 trial era, conducting a secondary evaluation either clinically or by axillary imaging before the use of a US-guided biopsy of suspicious nodes can potentially avoid the additional morbidity of axillary lymph node dissection and reduce the preoperative workload.

Spatial variability of available soil microelements in an ecological functional zone of Beijing.

Understanding the spatial variability of soil microelements and its influencing factors is of importance for a number of applications such as scientifically formulated fertilizer and environmental protection. This study used descriptive statistics and geostatistics to investigate the spatial variability of available soil Fe, Mn, Cu, and Zn contents in agricultural topsoil (0-20 cm) in an ecological functional zone located at Yanqing County, Beijing, China. Kriging method was applied to map the spatial patterns of available soil Fe, Mn, Cu, and Zn contents. Results showed that the available soil Cu had a widest spatial correlation distance (e.g., 9.6 km), which for available soil Fe, Mn, and Zn were only 1.29, 2.58, and 0.99 km, respectively. The values of C 0/sill for available soil Fe and Zn were 0.12 and 0.11, respectively, demonstrating that the spatial heterogeneity was mainly due to structural factors. The available soil Mn and Cu had the larger values of C 0/sill (i.e., 0.50 and 0.44 for Mn and Cu, respectively), which showed a medium spatial correlation. Mapping of the spatial patterns of the four microelements showed that the decrease trend of available soil Fe and Mn were from northeast to southwest across the study area. The highest amount of available soil Cu was distributed in the middle of the study area surrounding urban region which presented as a "single island". The highest amount of available soil Zn was mainly distributed in the north and south of the study area. One-way analysis of variance for the influencing factors showed that the lithology of parental materials, soil organic matter, and pH were important factors affecting spatial variability of the available microelements. The topography only had a significant influence on the spatial variability of available soil Fe and Mn contents, parental materials, and the land use types had little influence on the spatial variability.

Elastography ultrasound for breast lesions: fat-to-lesion strain ratio vs gland-to-lesion strain ratio.

OBJECTIVE: To compare the diagnostic performance of the fat-to-lesion strain ratio (FLR) and gland-to-lesion strain ratio (GLR) in the diagnosis of breast lesions. METHODS: This was an institutional ethics committee approved prospective study. One hundred and ninety-three breast lesions in 193 women (mean age, 46.03 ± 13.60 years, range 18-82 years) were examined with conventional and elastographic US. Both the FLR and GLR of the lesions were calculated. The elasticity scores of the lesions were also evaluated using the five-point elasticity scoring system. For diagnostic performance, the sensitivity, specificity and receiver operating characteristic (ROC) analysis were obtained. RESULTS: Seventy lesions were malignant and 123 were benign. Both the FLR and GLR of the lesions were significantly higher in malignant cases than in benign ones (P < 0.001 for both). The Az values for the FLR (0.847) and the elasticity score (0.829) were significantly higher than that for the GLR (0.752) (P = 0.009, P = 0.029, respectively). However, there was no significant difference in the Az value between the FLR and the elasticity score (P = 0.443). CONCLUSION: The FLR yielded better diagnostic performance than GLR in breast lesions; the fatty tissue was better than glandular tissue as the reference normal tissue for calculating strain ratio in breast elastography. KEY POINTS: FLR was higher than GLR in both malignant and benign breast lesions. Both FLR and GLR were higher in malignant than in benign breast lesions. FLR yielded better diagnostic performance than GLR in breast lesions.

The Effect of Wear on the Performance of a Rotary Lip Seal.

Theoretical analysis, combined with experimental verification, is used to study the effect of wear on the performance of a rotary lip seal as characterized by the pumping rate and friction torque. The performance of a rotary lip seal is determined by the sealing lip surface microscopic characteristics and contact characteristics at the sealing zone. The variation of the contact characteristics with wear has been established based on the variation of the sealing lip profiles measured by using a trilinear coordinate measuring instrument. The impression method is used to copy the lip surface microtopography at different wear times and then an optical interferometer is used to measure the impression surface microtopography to obtain the variation of roughness with wear. The variations of the roughness, contact characteristics, and approximate contact temperature with wear are inserted into a mixed lubrication model to obtain the variations of the pumping rate and friction torque. A comparison of the simulated results with those from experimental measurement verifies the validity of the theoretical analysis.

Shrub Growth Improves Morphological Features of Nebkhas: A Case Study of Nitraria tangutorum in the Tengger Desert.

To understand the role of shrubs in nebkha development, a comparative analysis of nebkha morphology and shrub features was conducted in two different habitats at the southeast margin of the Tengger Desert, Northern China. Morphometric variables of 184 Nitraria tangutorum nebkhas were measured in a semi-fixed lake-basin lowland site (site 1, n = 102) and a salinized fixed sand site (site 2, n = 82). Mean length, width, projected area, and accumulated sand volume were all greater in nebkhas in site 1 than in site 2 (p < 0.05); however, mean height (i.e., sand burial depth) did not differ significantly in nebkhas between the two sites (p > 0.05). The larger nebkha volume in site 1 relative to site 2 (mean, 88.19 m3 vs. 33.16 m3) implied that the projected area influenced the accumulated sand volume. Nebkhas in site 1 tended to have large areas, low densities, and high spatial autocorrelation, while nebkhas in site 2 exhibited opposite trends with stochastic distribution. Mean vegetation density was significantly higher in site 1 than in site 2 (p < 0.05), while mean vegetation height exhibited an opposite trend (p < 0.05). In addition, there was higher vegetation coverage in site 1 than in site 2 (p > 0.05). According to the results, plant species (i.e., N. tangutorum) limited nebkha height under similar wind regimes regardless of the transport distance of aeolian material, while aeolian deposition and its effect on shrub growth jointly increased nebkha size.

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.

Chlorate and perchlorate in tea leaves from major producing regions in China and related human exposure risk.

Chlorate and perchlorate are emerging pollutants that may interfere with thyroid function. Since they are highly water soluble, chlorate and perchlorate in tea leaves cause health concerns but have scarcely been studied. In this study, chlorate and perchlorate concentrations in 216 tea samples from different regions of China were determined. Perchlorate was detected in all the samples with a median concentration of 44.1 µg kg-1, while the chlorate detection frequency was 15.7%. We observed regional differences in perchlorate contents in tea leaves, with the highest quantity found in the central region of China. Except for dark tea, the concentration of perchlorate in tea infusions decreased with the increased number of times the tea leaves were brewed. The hazard quotients (HQs) of chlorate and perchlorate in all the samples were less than 1, suggesting negligible health risks caused by these pollutants from tea consumption. To the best of our knowledge, this is the first study to investigate chlorate and perchlorate contamination in tea infusions by simulating brewing behavior.

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.

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.

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 Analytical Model for the Normal Contact Stiffness of Mechanical Joint Surfaces Based on Parabolic Cylindrical Asperities.

The analytical results of normal contact stiffness for mechanical joint surfaces are quite different from the experimental data. So, this paper proposes an analytical model based on parabolic cylindrical asperity that considers the micro-topography of machined surfaces and how they were made. First, the topography of a machined surface was considered. Then, the parabolic cylindrical asperity and Gaussian distribution were used to create a hypothetical surface that better matches the real topography. Second, based on the hypothetical surface, the relationship between indentation depth and contact force in the elastic, elastoplastic, and plastic deformation intervals of the asperity was recalculated, and the theoretical analytical model of normal contact stiffness was obtained. Finally, an experimental test platform was then constructed, and the numerical simulation results were compared with the experimental results. At the same time, the numerical simulation results of the proposed model, the J. A. Greenwood and J. B. P. Williamson (GW) model, the W. R. Chang, I. Etsion, and D. B. Bogy (CEB) model, and the L. Kogut and I. Etsion (KE) model were compared with the experimental results. The results show that when roughness is Sa 1.6 µm, the maximum relative errors are 2.56%, 157.9%, 134%, and 90.3%, respectively. When roughness is Sa 3.2 µm, the maximum relative errors are 2.92%, 152.4%, 108.4%, and 75.1%, respectively. When roughness is Sa 4.5 µm, the maximum relative errors are 2.89%, 158.07%, 68.4%, and 46.13%, respectively. When roughness is Sa 5.8 µm, the maximum relative errors are 2.89%, 201.57%, 110.26%, and 73.18%, respectively. The comparison results demonstrate that the suggested model is accurate. This new method for examining the contact characteristics of mechanical joint surfaces uses the proposed model in conjunction with a micro-topography examination of an actual machined surface.

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.

LARNeXt: End-to-End Lie Algebra Residual Network for Face Recognition.

Face recognition has always been courted in computer vision and is especially amenable to situations with significant variations between frontal and profile faces. Traditional techniques make great strides either by synthesizing frontal faces from sizable datasets or by empirical pose invariant learning. In this paper, we propose a completely integrated embedded end-to-end Lie algebra residual architecture (LARNeXt) to achieve pose robust face recognition. First, we explore how the face rotation in the 3D space affects the deep feature generation process of convolutional neural networks (CNNs), and prove that face rotation in the image space is equivalent to an additive residual component in the feature space of CNNs, which is determined solely by the rotation. Second, on the basis of this theoretical finding, we further design three critical subnets to leverage a soft regression subnet with novel multi-fusion attention feature aggregation for efficient pose estimation, a residual subnet for decoding rotation information from input face images, and a gating subnet to learn rotation magnitude for controlling the strength of the residual component that contributes to the feature learning process. Finally, we conduct a large number of ablation experiments, and our quantitative and visualization results both corroborate the credibility of our theory and corresponding network designs. Our comprehensive experimental evaluations on frontal-profile face datasets, general unconstrained face recognition datasets, and industrial-grade tasks demonstrate that our method consistently outperforms the state-of-the-art ones.

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.