Search

The aim of this study was to discuss the diagnostic value of high-resolution ultrasound and virtual touch tissue imaging quantification (VTIQ) for distinguishing metastatic and benign central lymph nodes (CLNs) in patients with papillary thyroid carcinoma. This retrospective study involved 86 pathologically proven benign lymph nodes (LNs) and 118 metastatic LNs in patients with papillary thyroid carcinoma. We analyzed the sonographic features of CLNs (size, shape, distribution, hilum, echogenicity, cystic change, calcification, vascularity, shear-wave velocity [SWV]). The prevalence of sonographic features and the SWV was compared between metastatic and benign CLNs. The size, shape, margin, distribution, presence of hilum, echogenicity, calcification, and vascularity were significantly different between benign and metastatic CLNs (P < 0.05 for all). The mean maximum SWV for malignant CLNs was 3.139 ± 0.408 m/s, whereas that of benign CLNs was 2.418 ± 0.369 m/s (P < 0.05). The cutoff point of the SWV for differentiating benign and malignant LNs was 2.675 m/s. Logistic regression analysis showed that round or irregular shape, aggregation or fusion, calcification, and VTIQ value greater than 2.675 m/s of CLNs were independent risk factors for malignancy, with an odds ratio of 5.77, 3.05, 3.23, and 62.85, respectively. High-resolution ultrasound and VTIQ can provide valuable information for distinguishing metastatic from benign CLNs.

Sea Surface Temperature (SST) plays an important role in marine ecology. SST prediction raises considerable attention in ocean-related fields. Recently, deep learning models are widely used in SST prediction, but it is not easy to obtain optimal prediction results using historical observation data directly. More than temporal information, SST data also contains other features, such as trend, periodicity and disturbance. In this paper, we proposed a Multi-Channel LSTM (MC-LSTM) model to improve SST prediction. Firstly, a wavelet transform is used to decompose time-sequence data into multiple sequences representing trend, period and disturbance respectively. Secondly, we use multiple LSTM channels to train these data in parallel, and then obtain the combined prediction results. MC-LSTM can predict Sea surface temperature only by historical SST data, without the help of other spatial and climatic information, so it is easy to get the data. Compared with the direct use of LSTM for prediction, MC-LSTM can improve the prediction accuracy by 26%.

Breast cancer is still considered as the most common form of cancer as well as the leading causes of cancer deaths among women all over the world. Fortunately, the mortality of breast cancer can be significantly reduced via the early detection and diagnosis of breast cancer. As one of the most continually used diagnosis tools, ultrasonography (US) scan plays an important role in the detection and classification of the breast tumor. In this paper, we introduce a large breast ultrasound image database which stored breast ultrasound images and pathology results from breast tumor patients as well as their clinic diagnostic information. Furthermore, we design a web-based training system based on the database using a feature scoring scheme which based on the fifth edition of Breast Imaging Reporting and Data System (BI-RADS) lexicon for US. This online training system (new web-based teaching framework) automatically creates case-based exercises to train and guide the newly employed or resident sonographers for diagnosis of breast cancer using breast ultrasound images based on the BI-RADS.

Background and Objective Breast cancer is still considered as the most common form of cancer as well as the leading causes of cancer deaths among women all over the world. We aim to provide a web-based breast ultrasound database for online training inexperienced radiologists and giving computer-assisted diagnostic information for detection and classification of the breast tumor. Methods We introduce a web database which stores breast ultrasound images from breast cancer patients as well as their diagnostic information. A web-based training system using a feature scoring scheme based on Breast Imaging Reporting and Data System (BI-RADS) US lexicon was designed. A computer-aided diagnosis (CAD) subsystem was developed to assist the radiologists to make scores on the BI-RADS features for an input case. The training system possesses 1669 scored cases, where 412 cases are benign and 1257 cases are malignant. It was tested by 31 users including 12 interns, 11 junior radiologists, and 8 experienced senior radiologists. Results This online training system automatically creates case-based exercises to train and guide the newly employed or resident radiologists for the diagnosis of breast cancer using breast ultrasound images based on the BI-RADS. After the trainings, the interns and junior radiologists show significant improvement in the diagnosis of the breast tumor with ultrasound imaging (p-value .05). Conclusions The online training system can improve the capabilities of early-career radiologists in distinguishing between the benign and malignant lesions and reduce the misdiagnosis of breast cancer in a quick, convenient and effective manner.