Ultrasonic image left ventricular myocardium segmentation method and system and application

Abstract

The invention belongs to the technical field of image processing, and discloses an ultrasonic image left ventricular myocardium segmentation method and system and an application. The method comprisesthe steps of obtaining ultrasonic data, dividing the data into a training set, a verification set and a test set, and performing marking; enhancing the diversity of training set samples, interceptingan approximate left ventricular myocardium region, and performing histogram equalization and normalization operation on the data; using Pytorch to realize network segmentation, and storing a model having the best performance on the verification set; and measuring the thickness based on a segmentation result. According to the ultrasonic image left ventricular myocardium segmentation method based ona convolutional neural network, the left ventricular myocardium at the end of diastole can be automatically segmented, shape information of the left ventricular myocardium is added into the network to assist network learning, proposed mixed loss functions are optimized from three angles, and boundary information is further enhanced during learning; and the thickness can be automatically measuredbased on the segmentation result, and no post-processing is needed in the whole process.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a method, system and application for segmenting left ventricular myocardium in an ultrasonic image. Background technique [0002] At present, hypertrophic cardiomyopathy (HCM) is a common autosomal dominant cardiovascular disease, with an incidence of about 1:500-1:200 in the population and a mortality rate of about 1.4%-2.2%. , can lead to serious consequences such as chest tightness, chest pain, dyspnea, repeated syncope, atrial fibrillation, ventricular tachycardia, heart failure and even sudden death, and is the most common cause of sudden death in young people and athletes. At present, the main diagnostic criteria for HCM is left ventricular wall hypertrophy, which usually refers to interventricular septum or left ventricular wall thickness ≥ 15 mm measured by two-dimensional echocardiography, or thickness ≥ 13 mm in patients with a clear family history,...

AI Technical Summary

Problems solved by technology

[0005] (1) Existing technology to judge whether a patient has myocardial hypertrophy mainly relies on the rich clinical experience of doctors to find the boundary between the endocardium and epicardium to measure the thickness and see if the measured value meets the diagnostic criteria for hypertrophic cardiomyopathy. However, the repeatability of the process is poor, and due to different clinical experience, the measured thickness is different

[0006] (2) The traditional machine learning method first needs to manually extract features, but due to the large differences in the anatomical structure of the left ventricle of different HCM patients; and the traditional segmentation method such as atlas segmentation takes a long time

[0007] The difficulty in solving the above problems and defects is: the amount of data is crucial to the generalization of the model, but compared with natural images, medical images are often scarce, and the process of labeling data is time-consuming and labor-intensive and requires extensive clinical experience , so it is necessary to solve the problem of small amount of data; in the ultrasound image, the left ventricular myocardium only occupies a small part of the whole image, so it is necessary to solve the problem of the imbalance between positive and negative samples; and the boundary between endocardium and epicardium is not clear The problem

Images