Abstract:<section class="article-section__content"><h3 class="article-section__sub-title section1"> Purpose</h3><p>Automatic breast ultrasound (ABUS) imaging has become an essential tool in breast cancer diagnosis since it provides complementary information to other imaging modalities. Lesion segmentation on ABUS is a prerequisite step of breast cancer computer‐aided diagnosis (CAD). This work aims to develop a deep learning‐based method for breast tumor segmentation using three‐dimensional ABUS automatically.</p></section><section class="article-section__content"><h3 class="article-section__sub-title section1"> Methods</h3><p>For breast tumor segmentation in ABUS, we developed a Mask scoring region‐based convolutional neural network (R‐CNN) that consists of five subnetworks, i.e., a backbone, a regional proposal network, a region CNN head, a mask head, and a mask score head. A network block building direct correlation between mask quality and region class was integrated into a Mask scoring R‐CNN based framework for the segmentation of new ABUS images with ambiguous regions of interest (ROIs). For segmentation accuracy evaluation, we retrospectively investigated 70 patients with breast tumor confirmed with needle biopsy and manually delineated on ABUS, of which 40 were used for five‐fold cross‐validation and 30 were used for hold‐out test. The comparison between the automatic breast tumor segmentations and the manual contours was quantified by I) six metrics including Dice similarity coefficient (DSC), Jaccard index, 95% Hausdorff distance (HD95), mean surface distance (MSD), residual mean square distance (RMSD), and center of mass distance (CMD); II) Pearson correlation analysis and Bland‐Altman analysis.</p></section><section class="article-section__content"><h3 class="article-section__sub-title section1"> Results</h3><p>The mean (median) DSC was 85% ± 10.4% (89.4%) and 82.1% ± 14.5% (85.6%) for cross validation and hold‐out test, respectively. The corresponding HD95, MSD, RMSD and CMD of the two tests was 1.646 ± 1.191 and 1.665 ± 1.129 mm, 0.489 ± 0.406 and 0.475 ± 0.371 mm, 0.755 ± 0.755 and 0.751 ± 0.508 mm, and 0.672 ± 0.612 and 0.665 ± 0.729 mm. The mean volumetric difference (mean and ± 1.96 standard deviation) was 0.47 cc ([‐0.77, 1.71)) for the cross‐validation and 0.23 cc ([‐0.23 0.69]) for hold‐out test respectively.</p></section><section class="article-section__content"><h3 class="article-section__sub-title section1"> Conclusion</h3><p>We developed a novel Mask scoring R‐CNN approach for the automated segmentation of the breast tumor in ABUS images and demonstrated its accuracy for breast tumor segmentation. Our learning‐based method can potentially assist the clinical CAD of breast cancer using 3D ABUS imaging.</p></section>

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve This paper aims to develop a deep learning-based method for the segmentation of breast tumors in automated breast ultrasound (ABUS) images. Specifically, the authors propose an improved Mask Scoring R-CNN method to enhance the accuracy and robustness of automatic breast tumor segmentation. ### Background and Motivation 1. **Early Detection of Breast Cancer**: Breast cancer is one of the most common cancers among women, and early detection is crucial for improving treatment outcomes and reducing mortality rates. 2. **Advantages of ABUS**: 3D automated breast ultrasound (ABUS) is a cost-effective and safe imaging technology that has been proven to have significant advantages in detecting clinically important, small, invasive, and lymph node-negative breast cancers. 3. **Limitations of Manual Segmentation**: Currently, breast tumor segmentation in most centers still relies on manually outlining the lesion contours, which is not only time-consuming but also prone to inter-observer variability, affecting the accuracy of subsequent analyses. 4. **Shortcomings of Existing Methods**: Existing fully automatic breast tumor segmentation methods perform poorly when dealing with low-quality images (e.g., speckle noise, shadow effects, and low contrast), leading to suboptimal segmentation results. ### Research Objectives 1. **Develop a New Segmentation Method**: Propose a new Mask Scoring R-CNN method that integrates a mask scoring mechanism to improve segmentation accuracy. 2. **Validate the Method's Effectiveness**: Evaluate the performance of the proposed method in breast tumor segmentation through a retrospective study and compare it with existing methods. ### Method Overview 1. **Network Structure**: - **Backbone Network**: Use 3D ResNet to extract features. - **Region Proposal Network (RPN)**: Used for coarse localization of candidate regions for breast tumors. - **Region CNN Head**: Further refines the positions of candidate regions. - **Mask Head**: Generates binary masks of the tumors. - **Mask Scoring Head**: Establishes a direct correlation between segmentation quality and detection performance, improving the rationality of the final segmentation results. 2. **Dataset**: - Use ABUS images from 70 patients confirmed to have breast cancer via needle biopsy for evaluation. - 40 patients are used for 5-fold cross-validation, and 30 patients are used for independent testing. 3. **Performance Evaluation**: - Use multiple metrics (e.g., Dice similarity coefficient, Jaccard index, 95% Hausdorff distance, mean surface distance, residual mean square distance, and centroid distance) to evaluate segmentation accuracy. - Conduct volume correlation analysis and Bland-Altman analysis. ### Main Contributions 1. **Mask Scoring Head**: By introducing the mask scoring head, a direct correlation between segmentation quality and detection performance is established, improving the rationality of the segmentation results. 2. **Fully Automatic Segmentation**: The proposed method is fully automated, requiring no manual intervention, making it suitable for large-scale tasks. 3. **Comprehensive Performance Evaluation**: The segmentation performance is comprehensively evaluated using multiple metrics, ensuring the reliability and effectiveness of the method. ### Conclusion The authors successfully developed a Mask Scoring R-CNN-based method for automatic breast tumor segmentation and validated its segmentation accuracy and robustness on 3D ABUS images through experiments. The method outperforms existing segmentation methods across multiple performance metrics and holds promise for aiding the development of clinical breast cancer computer-aided diagnosis (CAD) systems.

Breast tumor segmentation in 3D automatic breast ultrasound using Mask scoring R‐CNN

Automatic Tumor Segmentation in 3D Automated Breast Ultrasound Using Convolutional Neural Network

Deep Learning-Based Automatic Segmentation for Size and Volumetric Measurement of Breast Cancer on Magnetic Resonance Imaging.

Deep Learning-based Breast Tumor Detection and Segmentation in 3D Ultrasound Image

Tumor segmentation in automated whole breast ultrasound using bidirectional LSTM neural network and attention mechanism

Cross-Model Attention-Guided Tumor Segmentation for 3D Automated Breast Ultrasound (ABUS) Images

Medical Breast Ultrasound Image Segmentation by Machine Learning

Tumor Detection in Automated Breast Ultrasound Using 3-D CNN and Prioritized Candidate Aggregation

Breast MRI Tumor Automatic Segmentation and Triple-Negative Breast Cancer Discrimination Algorithm Based on Deep Learning

CNN-Based Quality Assurance for Automatic Segmentation of Breast Cancer in Radiotherapy.

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

Automated 3D ultrasound image segmentation for assistant diagnosis of breast cancer

Mass Detection and Segmentation in Digital Breast Tomosynthesis Using 3D-Mask Region-Based Convolutional Neural Network: A Comparative Analysis

Prostate cancer of magnetic resonance imaging automatic segmentation and detection of based on 3D-Mask RCNN

Automatic tumor segmentation in breast ultrasound images using a dilated fully convolutional network combined with an active contour model.

Multi-task Learning for Segmentation and Classification of Tumors in 3D Automated Breast Ultrasound Images

Image Segmentation of Triple-Negative Breast Cancer by Incorporating Multiscale and Parallel Attention Mechanisms

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

Fully automatic tumor segmentation of breast ultrasound images with deep learning

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

Automatic Breast Tumor Detection in ABVS Images Based on Convolutional Neural Network and Superpixel Patterns