A Semantic Segmentation Method for Masses in Mammography Images Based on Deep Residual Networks

Abstract

The present invention proposes a method for semantic segmentation of lumps in breast mammography images based on deep residual networks. and the corresponding label images are divided into training samples and test samples; after preprocessing the training samples, a training data set is formed; a deep residual network is constructed, and the training data set is used to train the network to obtain a deep residual network training model; the preprocessing is to be After segmenting the mammography image mass, the deep residual network training model is used to perform binary classification and post-processing on the mammography image pixels to be segmented, and output the mass segmentation image to realize the semantic segmentation of the mammography image mass. The invention can effectively improve the automation and intelligence level of mammography image mass segmentation, and can be applied in technical fields such as assisting radiologists in medical diagnosis.

Description

technical field [0001] The invention relates to the fields of machine learning and digital medical image processing and analysis, in particular to a method for semantic segmentation of tumors in mammography images based on a deep residual network. Background technique [0002] Breast cancer has become a common malignancy among women worldwide and is the leading cause of cancer death in women. The incidence of female breast cancer in my country is getting younger and rising year by year, and as many as 200,000 people die from breast cancer every year, which has brought catastrophic panic to women's health. Early detection to improve breast cancer outcomes and survival remains the cornerstone of breast cancer control. Mammography has high spatial resolution and can display the early symptoms of breast cancer. It is recognized as the most reliable and convenient method for early diagnosis of breast cancer. With the rapid development of computer and image processing technology...

AI Technical Summary

Problems solved by technology

However, the CV model has its own inevitable defects. When the distribution in the foreground and background regions is uneven, the internal and external characteristic parameters of the level set in the CV model cannot accurately express the local characteristics of the image.

On the other hand, the normal tissue near the tumor in the mammography image is very similar to the tumor, and the intensity distribution of these areas is also very uneven

Therefore, when CV is dealing with ROI (region of interest, region of interest) images with low contrast and large fluctuations in gray levels inside and outside the mass, the segmentation results are often not ideal.

Images