Abstract: Disclosed are an extended U-Net for multi-information extraction and an application method thereof in low-dose X-ray imaging. The follow technical solution is adopted: expanding the number of output channels in upsampling of a U-Net model, adding a fully connected layer at the end to realize extraction and superposition calculation of multiple information, processing low-dose X-ray images by using the extended U-Net, and taking normalized images, noise and image salient features as outputs for learning and training.

Abstract: Disclosed is an ultrasonic Doppler flow imaging method. The method includes: using eigen decomposition on an ultrasonic image to obtain eigenvectors matrices with different power values, so that the SNR and power distribution of each point of the ultrasonic image are calculated, thereby accurately extracting flow data with high reliability and obtaining accurate and stable flow images. The method has the beneficial effects of filtering out tissue signals and noise signals on the characteristic dimension, so as to better extract flow data by using eigen decomposition on ultrasonic images, retaining complete flow data and providing stable and clear imaging of fine flow without the need of contrast agent injection that may cause injuries to human bodies.

Abstract: Disclosed is a compound method of shear-wave elastography (SWE) and quasi-static elastography (QSE). The method includes: firstly, pre-scanning is carried out on a target region to obtain a rough comparison relation curve, then real-time scanning is carried out. The rough comparison relation curve is optimized and corrected to obtain an accurate comparison relation curve. Finally, the real-time shear modulus of a large area is calculated according to the real-time elastic relative value of the large area, so that accurate real-time elastography is realized.

Abstract: An automatic myocardial segmentation method in two-dimensional echocardiography includes: calculating and extracting ridge information in two-dimensional echocardiography; extracting image skeleton data, and connecting fragmented skeleton data to obtain continuous skeleton data by using interpolation; screening the continuous skeleton data, and extracting two optimal paths from a top to a bottom based on a myocardial feature to obtain left and right myocardial skeletons; extracting a set of feature point candidates from the left and right myocardial skeletons, and selecting a pair of key matching feature points from the set of feature point candidates as anchor points based on the myocardial feature; entering anchor point data and left and right myocardial skeleton data in a mathematical model of heart chambers to fit the morphology of heart chambers and obtain an initial myocardial contour; and constructing a myocardial feature energy equation.