Deep learning classification method for mammography images based on lightweight neural network

Abstract

The invention relates to a deep learning classification method for mammography images based on a lightweight neural network. The method uses an image classification algorithm based on deep learning to realize the breast density classification of mammography images, and uses a deep learning framework based on a lightweight neural network. The method of the invention significantly improves the adaptability on small-scale image data sets, further improves the accuracy and processing speed of mammary gland density classification, and can realize automatic mammary gland density classification of mammary gland molybdenum target images.

Description

Technical field [0001] The present invention is in the field of biomedicine, in particular, to mammography image depth learning classification method based on neural network lightweight. Background technique [0002] Mammography full name of mammography X-ray photographic examination, also known as molybdenum examination, the diagnosis of breast disease is the preferred and most convenient, most reliable non-invasive means of detection, the pain is relatively small, simple, and high-resolution good reproducibility, to retain images available before and after comparison, no age limit in shape, has as a routine means of checking. Mammography as a relatively non-invasive examination method, more comprehensive and correctly reflect substantially the entire breast anatomy, physiological factors influence was observed on a variety of structures such as the breast of the menstrual cycle, pregnancy, lactation, and can be dynamically observation; help identify benign lesions and malignant...

AI Technical Summary

Problems solved by technology

However, due to the small number of samples, large differences, and uneven density distribution of medical mammography images, for the application of mammography image processing and analysis, manual recognition can only simply divide the boundaries of breast regions and determine the density of breasts in the region. Qualitative estimation has been difficult to meet the accuracy and speed requirements of breast density classification, and the traditional automatic density classification method of mammography images also has shortcomings that seriously affect the analysis results: the breast itself and its shape are different, it is difficult to use the traditional The method based on the morphological model segments various tissues, resulting in inaccurate boundary division between the mammary gland and the image background; the density distribution of various tissues inside the mammary gland is extremely uneven, which makes the statistical results of the density distribution histogram easy to be partial and complete, resulting in mammary gland An error occurred in the statistical analysis of the overall density, which seriously affected the discrimination accuracy and processing speed of mammography density classification

Images