Automatic ventricular region segmentation method and device of cardiac ultrasound image and storage medium

Abstract

The invention provides an automatic ventricular region segmentation method and device for a cardiac ultrasound image and a storage medium, and relates to the technical field of image processing. According to the method, a ventricular region automatic segmentation model is utilized to obtain a segmentation result of each pixel in a to-be-segmented ultrasonic image. According to the ventricular region automatic segmentation model, compared with a traditional UNet method, the network structure of the model is improved, an up-sampling deconvolution layer of a decoder of the model is a multi-dimensional splicing type deconvolution up-sampling layer, the decoder can better obtain image information from multiple dimensions, and then the segmentation accuracy is improved. According to the automatic ventricular region segmentation model, the model training speed and accuracy can be effectively improved, and then the speed and accuracy of automatically segmenting the left ventricular region in the echocardiogram are improved.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a method, device and storage medium for automatic segmentation of ventricle regions of cardiac ultrasound images. Background technique [0002] Cardiac ejection fraction is an important indicator for judging the type of heart failure and one of the important methods for measuring heart function. Cardiac ejection fraction refers specifically to the percentage of blood that the heart ejects during systole after filling with blood during diastole. The first step to obtain this index relies on the segmentation of the ventricles of the heart image. [0003] In recent years, many different methods have been proposed and used for the automatic segmentation of cardiac images, which are mainly divided into two categories from the general direction: one is the expansion and improvement of traditional segmentation algorithms; the other is the combination of deep learning a...

AI Technical Summary

Problems solved by technology

[0004] 1) Echocardiography has obvious noise, and the position of the region to be segmented is relatively fixed.

Existing segmentation techniques using neural networks do not express enough information from multiple scales, and there is still much room for improvement

[0005] 2) At present, the optimization objective function of the image segmentation model training still uses the traditional MSE (mean square error function), which will make the model training results unable to converge

Images