Heart MRI segmentation method and system based on adversarial network

Abstract

The invention discloses a heart MRI segmentation method and a heart MRI segmentation system based on an adversarial network. The adversarial network comprises a divider and a discriminator. The methodcomprises the following specific steps: inputting heart MRI to be segmented; preprocessing the input heart MRI to be segmented; training an adversarial network; calling a divider model subjected to adversarial training to divide the preprocessed heart MRI; completing ventricular structure segmentation, and outputting left ventricle, right ventricle and myocardial segmentation results. And evaluating whether the left ventricle, the right ventricle and the myocardium obtained by segmentation are accurate or not according to the Dice coefficient and the Hough distance. The evaluation result shows that the method can automatically output the high-precision heart MRI segmentation result, thereby relieving the tension of public health resources and improving the diagnosis efficiency of doctors.

Description

technical field [0001] The present disclosure relates to the technical field of medical image processing, in particular to a cardiac MRI segmentation method and system based on an adversarial network. Background technique [0002] The statements in this section merely mention background art related to the present disclosure and do not necessarily constitute prior art. [0003] With the development of medical technology and the improvement of public health conditions, the fatality rate of epidemics has continued to decline. Correspondingly, chronic diseases have become the main cause of death for modern people. The heart is the power pump for blood circulation, and lesions in the heart area are extremely fatal. With the improvement of modern people's living standards, cardiovascular disease has become the number one cause of death for humans. Before medical image technology was applied to clinical diagnosis, doctors could only diagnose heart disease through electrocardiogra...

AI Technical Summary

Problems solved by technology

However, manual segmentation is time-consuming and labor-intensive. Since the shape of the heart is different and the pathological features vary widely, the segmentation results of different doctors are inconsistent, and the same doctor may also have different results in two segmentations, which leads to poor repeatability of manual segmentation.

[0008] To sum up, most of the current ventricular segmentation algorithms have shortcomings such as excessive calculation and complicated steps.

Images