End-to-end mammary gland ultrasound image segmentation method based on Distance-AttU-Net (Distance-AttU-Net)

Abstract

The invention discloses an end-to-end mammary gland ultrasonic image segmentation method based on Distance-AttU-Net. The method comprises the following steps: firstly, an original mammary gland ultrasonic image is directly used as input of a network, a trained loss value is calculated by using a self-defined loss function, and a network model is trained by optimizing the loss value and updating model parameters through back propagation until the model converges. Through the training of the network, the network itself has the capability of distinguishing the lesion from the similar regions around the lesion, and finally the lesion region in the mammary gland ultrasonic image is segmented. Through testing on the public data set B, the network model has high segmentation capability, it is proved that the mammary gland lesion segmentation model based on the ultrasonic image with excellent performance can be trained under the condition that training samples are few, and the mammary gland lesion segmentation model based on the ultrasonic image has the advantages of end-to-end segmentation, multi-scale lesion processing, high discrimination and the like.

Description

technical field [0001] The invention relates to the fields of deep learning, computer vision and medical images, in particular to an end-to-end breast ultrasound image segmentation method based on Distance-AttU-Net. Background technique [0002] There are many ways to diagnose breast cancer. Ultrasound imaging has become the most effective way to diagnose breast lesions due to its advantages of no radiation, low cost, and fast imaging. However, the judgment and understanding of breast ultrasound images requires rich clinical experience of doctors, and the workload of interpretation is relatively large. Relying entirely on manual processing by doctors is prone to fatigue, leading to a further increase in misdiagnosis rates. Therefore, in order to reduce the burden on doctors and improve diagnostic results, it is extremely important to study reliable methods for intelligent segmentation of ultrasound breast lesions. [0003] At present, the methods for ultrasonic breast image...

AI Technical Summary

Problems solved by technology

[0007] In order to solve the problem of low accuracy and complicated data processing steps in the segmentation technology of existing breast ultrasound images when dealing with multi-size lesions

Images