Method for identifying benign and malignant breast nodules based on shear wave elastic diagram of deep learning

Abstract

The invention relates to a medical image processing technology, and aims to provide a method for identifying benign and malignant breast nodules based on a shear wave elastic diagram of deep learning.The method comprises the following steps: collecting case data of an ultrasonic mammary gland common B-ultrasonic image and a shear wave elastic image with nodules, and constructing a benign and malignant classification data set according to a pathological result; selecting a basic network structure, replacing part of the convolution layer with a rotary pooling convolution layer, and constructinga benign and malignant identification network structure; adding nodule mask information to network input, performing data enhancement on a training set, and using separable Dropout calculation in thetraining process to improve the generalization ability of the model; and inputting a test image, carrying out multi-image-block and multi-model test to evaluate model performance, and carrying out breast nodule benign and malignant judgment on the image. Compared with a method for identifying benign and malignant parts by using a common B-ultrasonic image, the method can improve the benign and malignant diagnosis accuracy. According to the method, the rotation invariance can be learned, and the model recognition accuracy can be improved under the condition that the network calculation complexity is not increased.

Description

technical field [0001] The invention relates to the technical field of medical image processing, in particular to a method for identifying benign and malignant breast nodules based on shear wave elastic images of deep convolutional neural networks. Background technique [0002] As the incidence of malignant breast tumors increases year by year, breast examination has become an indispensable item in female physical examination. The current breast examination is mainly based on ultrasound and mammography, but mammography has X-ray radiation and examination. The method is more painful and other factors, so ultrasound is the first choice for breast examination due to its unique examination characteristics. Routine breast examinations mainly use two-dimensional, color Doppler and other examination methods to observe the shape, boundary, internal echo and blood supply of lesions. Later, the appearance of ultrasound elastography provided more clear texture information for examinati...

AI Technical Summary

Problems solved by technology

[0005] The deep convolutional neural network can learn the commonality of the same category in the image and the main differences between different categories. The highly abstract features output by the last few layers of the network have certain translation, scale, brightness, etc. invariance, but do not have rotation. Denaturation, resulting in insufficient stability of learned features

When augmenting the simulated data of the input image, although random rotation transformation can be added, the parameters need to be set in advance, and the parameter ranges of different tasks are different, it is

Images