Deep Convolutional Neural Networks-Based Automatic Breast Segmentation and Mass Detection in DCE-MRI.

Yihua Huang,Han Jiao,Li Li,Xinhua Jiang,Zhiyong Pang,Xiaofeng Lin

doi:10.1155/2020/2413706

Abstract

Breast segmentation and mass detection in medical images are important for diagnosis and treatment follow-up. Automation of these challenging tasks can assist radiologists by reducing the high manual workload of breast cancer analysis. In this paper, deep convolutional neural networks (DCNN) were employed for breast segmentation and mass detection in dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). First, the region of the breasts was segmented from the remaining body parts by building a fully convolutional neural network based on U-Net++. Using the method of deep learning to extract the target area can help to reduce the interference external to the breast. Second, a faster region with convolutional neural network (Faster RCNN) was used for mass detection on segmented breast images. The dataset of DCE-MRI used in this study was obtained from 75 patients, and a 5-fold cross validation method was adopted. The statistical analysis of breast region segmentation was carried out by computing the Dice similarity coefficient (DSC), Jaccard coefficient, and segmentation sensitivity. For validation of breast mass detection, the sensitivity with the number of false positives per case was computed and analyzed. The Dice and Jaccard coefficients and the segmentation sensitivity value for breast region segmentation were 0.951, 0.908, and 0.948, respectively, which were better than those of the original U-Net algorithm, and the average sensitivity for mass detection achieved 0.874 with 3.4 false positives per case.

Highlights

Breast cancer is one of the most common cancers amongst women worldwide [1]
We proposed breast region segmentation by training a U-Net++ model, which removes interference from different Dynamic contrastenhanced magnetic resonance imaging (DCE-MRI) series external to the breast region
We trained the U-Net++ model and the U-Net model as comparison for 50 epochs using stochastic gradient descent (SGD) as the optimizer and used the last epoch to predict validation datasets. e batch size was set to five given the limitation of GPU memory. e training processes of U-Net and U-Net++ are shown in Figures 6 and 7, respectively, which show the final convergence of the networks

Summary

Introduction

Breast cancer is one of the most common cancers amongst women worldwide [1]. Early diagnosis and treatment are proven to reduce the mortality rate [2]. E correct resolution of DCE-MRI image of the breast depends largely on the quality of visualization, operation experience, and the time needed for data analysis. Automatic segmentation of the breast region reduces result bias and can accelerate data processing. Jiang et al [8] used dynamic programing combined with preprocessing to segment the breast region in fat-suppressed transverse DCE-MRI. While these traditional methods have shown good performance, the robustness of these algorithms is insufficient, in that they tend to fail in some specific data because they can only process the underlying information. Xu et al used a 2D U-Net for automatic breast region segmentation in DCE-MRI [12]. Our results indicated the effectiveness and accuracy of this method in biomedical segmentation tasks

Objectives

Methods

Results

Conclusion