Breast ultrasonic image tumor classification method based on attention neural network

Abstract

The invention provides a breast ultrasonic image tumor classification method based on an attention neural network, and the method comprises the following steps: S1, a data expansion step: carrying out the image enhancement through employing breast ultrasonic images of a patient on a training set, and expanding a data set of the training set; S2, a feature re-calibration step: designing an attention mechanism module SE of a neuron dimension, and carrying out feature re-calibration on neuron features; S3, an over-fitting relieving step: designing a classification module containing a global average pooling layer, a full connection layer, a batch normalization layer, ReLu activation and a Dropout layer, reducing parameters, and relieving over-fitting; and S4, a test evaluation step: performing training on the training set, inputting a tested data set into the final model, evaluating the model performance through multiple classification indexes, and comparing the model performance with a traditional VGG16 network before improvement. According to the invention, the improved convolutional neural network based on the attention mechanism is applied to the mammary gland ultrasonic images, so that benign and malignant mammary gland ultrasonic images can be automatically classified.

Description

technical field [0001] The present invention relates to the technical field of biomedicine, and more specifically relates to a method for classifying breast ultrasound image tumors based on attentional neural network. Background technique [0002] Today, breast cancer has become the most common cancer among women in the world. Early screening of breast cancer and timely treatment can effectively reduce mortality. At present, the mainstream breast cancer screening method in my country is breast imaging technology, which mainly includes mammography, magnetic resonance imaging and ultrasound imaging. X-ray imaging has certain ionizing radiation, which will cause harm to the human body; in addition, because this method can only produce two-dimensional images, it is difficult to distinguish tumors in the images for dense breasts; Pain. MRIs are long, difficult and expensive to make appointments. The ionizing radiation of ultrasound imaging technology is very small, and the ima...

AI Technical Summary

Problems solved by technology

[0004] Traditional breast ultrasound image classification algorithms generally require manual feature engineering, which is complex and less robust

Images