Segmentation of Echocardiography Based on Deep Learning Model

In order to achieve the classification of mitral regurgitation, a deep learning network VDS-UNET was designed automatically segment critical regions echocardiography with three sections apical two-chamber, three-chamber, and four-chamber. First, an expert-labeled dataset 153 echocardiographic videos 2183 images from 49 subjects constructed. Then, convolution layer in VGG16 used replace contraction path original UNet extract image features, depth supervision added expansion segmentation LA, LV, MV. The results showed that Dice coefficients MV were 0.935, 0.915, 0.757, respectively. proposed can simultaneous accurate multi-section echocardiography, laying foundation for quantitative measurement clinical parameters related regurgitation.