Background: Studies on deep learning (DL)-based models in breast ultrasound (US) remain at the early stage due to a lack of large datasets for training and independent test sets for verification. We aimed to develop a DL model for differentiating benign from malignant breast lesions on US using a large multicenter dataset and explore the model’s ability to assist the radiologists.Methods: A total of 14,043 US images from 5012 women were prospectively collected from 32 hospitals. To develop the DL model, the patients from 30 hospitals were randomly divided into a training cohort (n = 4149) and an internal test cohort (n = 466). The remaining 2 hospitals (n = 397) were used as the external test cohorts (ETC). We compared the model with the prospective Breast Imaging Reporting and Data System assessment and five radiologists. We also explored the model’s ability to assist the radiologists using two different methods.Results: The model demonstrated excellent diagnostic performance with the ETC, with a high area under the receiver operating characteristic curve (AUC, 0.913), sensitivity (88.84%), specificity (83.77%), and accuracy (86.40%). In the comparison set, the AUC was similar to that of the expert (p = 0.5629) and one experienced radiologist (p = 0.2112) and significantly higher than that of three inexperienced radiologists (p < 0.01). After model assistance, the accuracies and specificities of the radiologists were substantially improved without loss in sensitivities.Conclusions: The DL model yielded satisfactory predictions in distinguishing benign from malignant breast lesions. The model showed the potential value in improving the diagnosis of breast lesions by radiologists.

Application of superb microvascular imaging in focal liver lesions

Objective.: To investigate skin elasticity using acoustic radiation force impulse (ARFI) quantification in systemic sclerosis (SSc), and compare the modified Rodnan skin score (mRSS) with measured shear wave velocity (SWV) and thickness of the skin.Methods.: Fifteen patients with diffuse cutaneous SSc (dcSSc) and 15 age-matched and sex-matched healthy controls were evaluated. The SWV and thickness of skin were measured at 17 sites corresponding to those assessed in the mRSS in each participant. The SWV measurements of skin were compared between patients with dcSSc and healthy controls. The correlations between the mRSS and the skin SWV and thickness were explored using Spearman’s correlation.Results.: The SWV values were higher in patients with dcSSc compared with healthy controls at right hand dorsum, right forearm, left hand dorsum, left forearm, right foot dorsum, and left foot dorsum (p < 0.05). In patients with dcSSc, the SWV values of uninvolved skin were higher than those of controls (p < 0.001), and the SWV values increased with increasing skin scores except for skin score 3 (p < 0.05). The sum of the SWV values correlated with total clinical skin score (r = 0.841, p < 0.001), and the sum of the skin thickness correlated with total clinical skin score (r = 0.740, p = 0.002).Conclusion.: ARFI quantification is feasible and reliable for assessing the skin involvement in dcSSc. ARFI quantification could identify early skin change that may precede palpable skin involvement, and may be a valuable adjunct to skin evaluation in patients with SSc.

Purpose: The purpose of our study was to prospectively evaluate the diagnostic performance of the vascular index (VI, defined as the ratio of Doppler signal pixels to pixels in the total lesion) measured via Smart 3-D superb microvascular imaging (SMI) for breast lesions. Patients and methods: Two hundred and thirty-two consecutive patients with 236 breast lesions referred for biopsy at Peking Union Medical College Hospital were enrolled in the study from December 2016 to November 2017. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy of VI were calculated with histopathologic results as the reference standard. Results: Of the 236 breast lesions, 121 were malignant and 115 were benign. The mean VI was significantly higher in malignant lesions (9.7±8.2) than that in benign ones (3.4±3.3) (P<0.0001). Sensitivity, specificity, PPV, NPV and accuracy of VI (4.0 as the threshold) were respectively: 76.0%, 66.1%, 70.2%, 72.4% and 71.2% (P<0.05). Conclusion: Smart three-dimensional (3-D) SMI is a noninvasive tool using two-dimensional (2-D) scanning to generate 3-D vascular architecture with a high-resolution image of micro-vessels. This can be used as a qualitative guide to identify the optimal 2-D SMI plane with the most abundant vasculature to guide VI quantitative measurements of breast lesions. Smart 3-D SMI may potentially serve as a noninvasive tool to accurately characterize benign versus malignant breast lesions.