The invention relates to an Attention-based segmentation and classification method for multi-scale dilated myocardiumMulti-scale dilated cardiac muscle segmentation and classification method based on Attention

Abstract

The invention relates to a multi-scale dilated cardiac muscle segmentation and classification method based on Attention. The method comprises the following steps: collecting a plurality of cases suffering from the dilated cardiomyopathy, and collecting MRI image data of the cardiomyopathy; P; performing artificial edge labeling on the lesion area of the MRI image data collected in the last step layer by layer to serve as label data; carrying out standardized preprocessing on the label data obtained in the previous step and converting the label data into a two-dimensional data set; constructinga multilayer two-dimensional convolutional neural network based on Attention, and training the multilayer two-dimensional convolutional neural network by using the two-dimensional data set in the previous step; For to-be-segmented and classified MRI image data of the myocardial part, collecting medical images of the same mode of the same part, and carrying out standardization processing on the collected images; and performing automatic segmentation and classification on the to-be-segmented and classified MRI image data of the myocardial part through the trained network model. According to theinvention, automatic segmentation and classification of the expansive myocardial region can be realized, and higher precision can be obtained in comparison with a mainstream network.

Description

technical field [0001] The present invention relates to the technical field of image segmentation and classification of dilated myocardium in the field of image segmentation, in particular to an Attention-based multi-scale dilated myocardium segmentation and classification method. Background technique [0002] Dilated cardiomyopathy (dilated cardiomyopathy, DCM) is a primary disease of unknown cause, and it is one of the high incidence myocardial diseases in my country. The incidence rate is very high. The annual incidence rate of DCM is 5 / 100000~8 / 100000 people, and there is a rising trend. There are more men than women (2.5:1), and the average age of onset is about 40 years old. The clinical manifestations of patients vary in severity, and many symptomatic patients are progressively deteriorating, and 10% to 15% of patients develop symptoms of heart failure within one year. It is estimated that the annual mortality rate of the typical patient population with heart failure...

AI Technical Summary

Problems solved by technology

However, this approach will cause two problems: a large number of redundant calculations lead to low time efficiency, and the network cannot learn global features

However, this method does not highlight effective feature information, nor suppress useless feature information, so it cannot make full use of feature information that is beneficial to classification

[0010] In summary, there are two problems in the above method. On the one hand, there is a lack of reuse of original feature information and insufficient learning of global feature information.

On the other hand, it lacks the prominence of useful feature information and the suppression of useless feature information

Images