Automated Segmentation of Mass Regions in DBT Images Using a Dilated DCNN Approach.

Jianming Ye,Jianqing Wang,Liuyi Li,Gang Chen,Weiji Yang,Xiangcai Wang,Xiaobo Lai,Xiaomei Xu,Chun Xie,Guangming Zhang

doi:10.1155/2022/9082694

Abstract

To overcome the limitations of conventional breast screening methods based on digital mammography, a quasi-3D imaging technique, digital breast tomosynthesis (DBT) has been developed in the field of breast cancer screening in recent years. In this work, a computer-aided architecture for mass regions segmentation in DBT images using a dilated deep convolutional neural network (DCNN) is developed. First, to improve the low contrast of breast tumour candidate regions and depress the background tissue noise in the DBT image effectively, the constraint matrix is established after top-hat transformation and multiplied with the DBT image. Second, input image patches are generated, and the data augmentation technique is performed to create the training data set for training a dilated DCNN architecture. Then the mass regions in DBT images are preliminarily segmented; each pixel is divided into two different kinds of labels. Finally, the postprocessing procedure removes all false-positives regions with less than 50 voxels. The final segmentation results are obtained by smoothing the boundaries of the mass regions with a median filter. The testing accuracy (ACC), sensitivity (SEN), and the area under the receiver operating curve (AUC) are adopted as the evaluation metrics, and the ACC, SEN, as well as AUC are 86.3%, 85.6%, and 0.852 for segmenting the mass regions in DBT images on the entire data set, respectively. The experimental results indicate that our proposed approach achieves promising results compared with other classical CAD-based frameworks.

Highlights

Breast cancer is one of the leading causes of diseases in women worldwide, and it is the most common cause of cancer deaths in women
A dilated deep convolutional neural network (DCNN) architecture is designed for the dense prediction of mass regions in digital breast tomosynthesis (DBT) images, which systematically aggregates multiscale contextual information without losing resolution
We will discuss the research work of Kim et al [28], Fotin et al [29], and Samara et al [30] in detail. ey applied deep learning to the detection and segmentation of breast mass regions in DBT images. eir research works evaluated the automated segmentation computer-aided diagnosis (CAD) frameworks for breast masses in DBT images using the hand-crafted feature- and DCNN-based models. e DCNN model proposed by Kim et al [28] extracted low-level features from the regions of interest (ROIs) and corresponding ROIs, respectively, through the convolutional layers separately, which can recognize the latent bilateral feature representations of breast masses in the reconstructed DBT volumes

Summary

Introduction

Breast cancer is one of the leading causes of diseases in women worldwide, and it is the most common cause of cancer deaths in women. Since the late 70s of the last century, the incidence of breast cancer worldwide has been increasing. According to the report “ e Status and Trends of Cancer in China 2017” released by the National Cancer Center, the incidence of breast cancer ranks first among female malignant tumours [1]. Diagnosis and treatment can effectively reduce the mortality of breast cancer patients and improve their quality of life [2]. In developed countries, organized and opportunistic screening programs have significantly reduced breast cancer mortality. Two-dimensional mammography uses a new detector, it is well known that it still has its limitations because the normal structures and pathological structures may overlap each other when obtaining the transmission X-ray image [3]

Methods

Results

Conclusion