Abstract:Breast ultrasound image analysis is commonly used to help breast cancer diagnosis.Breast tumor segmentation is the first step of image analysis and plays an important role for the automatic breast ultrasound image analysis.However,complex characteristics such as intensity inhomogeneity,variance of texture and shape in breast ultrasound images makes accurate tumor segmentation difficult.In order to deal with the complex characteristics appeared in ultrasound images,this paper proposes a hierarchical breast tumor segmentation method.The method is developed with active contour model framework.In the proposed framework,active contour model is first to be used as low-level segmentation model for initial segmentation.Based on the initial segmentation result,a high-level seg-mentation model based on superpixel and Support Vector Machine(SVM)is proposed.Simple Linear Interactive Cluster(SLIC)is firstly used to divide the image into superpixels.And then,intensity,texture and local features are extracted for each superpixel.At last,SVM is trained based on these superpixels.Because the high-model is trained by introducing the segmentation result information,the high-model can be used for recognizing tumor regions misclassified by low-level segmentation model.In the process of image segmentation by using low-level segmentation model,complex characteristics of the ultrasound images may result in segmentation error of some local tumor regions which appear large intensity difference or texture variation.However,high-level segmentation model has ability to learn certain characteristics of the low-level segmentation model,which can be complementary to low-level segmentation model.Therefore,the high-level segmentation model can be used to correct the segmentation error caused by using low-level segmentation model,which can be used to improve the segmentation performance.We conduct the experiments on our self-constructed database and the experimental results demonstrate the effectiveness and robustness of the proposed method.

Breast Anatomy Enriched Tumor Saliency Estimation

Tumor saliency estimation for breast ultrasound images via breast anatomy modeling

A Hybrid Framework for Tumor Saliency Estimation

Attention Enriched Deep Learning Model for Breast Tumor Segmentation in Ultrasound Images

Fuzzy Semantic Segmentation of Breast Ultrasound Image with Breast Anatomy Constraints

Breast Tumor Segmentation in Ultrasound Images Based on U-NET Model

Saliency map-guided hierarchical dense feature aggregation framework for breast lesion classification using ultrasound image

BreastSAM: A Study of Segment Anything Model for Breast Tumor Detection in Ultrasound Images

ESTAN: Enhanced Small Tumor-Aware Network for Breast Ultrasound Image Segmentation

ACL-DUNet: A tumor segmentation method based on multiple attention and densely connected breast ultrasound images

BUSnet: A Deep Learning Model of Breast Tumor Lesion Detection for Ultrasound Images

Breast Tumor Detection In Ultrasound Images Using Deep Learning

Hierarchical segmentation of breast tumor in ultrasound image

Breast Tumor Classification in Ultrasound Images by Fusion of Deep Convolutional Neural Network and Shallow LBP Feature

Spatial Attention Lesion Detection On Automated Breast Ultrasound

New one-step model of breast tumor locating based on deep learning.

Computer-aided Tumor Diagnosis in Automated Breast Ultrasound using 3D Detection Network

A2DMN: Anatomy-Aware Dilated Multiscale Network for Breast Ultrasound Semantic Segmentation

An RDAU-NET model for lesion segmentation in breast ultrasound images

BreastSAM: Adapting the Segmentation Anything Model for Breast Tumor Segmentation in Ultrasound Imaging

3D Multi-View Tumor Detection in Automated Whole Breast Ultrasound Using Deep Convolutional Neural Network