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

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Accurate breast cancer diagnosis and monitoring rely on precise segmentation and mass detection in medical images.
• Manual analysis of breast cancer imaging is time-consuming and prone to high workloads for radiologists.
• Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) is a key modality for breast cancer assessment.

Purpose of the Study:

• To develop and evaluate deep learning models for automated breast segmentation and mass detection in DCE-MRI.
• To improve the efficiency and accuracy of breast cancer analysis in radiological workflows.
• To compare the performance of advanced deep learning architectures against baseline methods.

Main Methods:

• Utilized a U-Net++ architecture for fully convolutional neural network-based breast segmentation.
• Employed a Faster Region-based Convolutional Neural Network (Faster R-CNN) for mass detection on segmented breast images.
• Validated the models on a dataset of DCE-MRI from 75 patients using 5-fold cross-validation.

Main Results:

• Achieved high performance in breast region segmentation with Dice Similarity Coefficient (DSC) of 0.951, Jaccard coefficient of 0.908, and segmentation sensitivity of 0.948.
• Demonstrated superior segmentation performance compared to the original U-Net algorithm.
• Attained an average sensitivity of 0.874 for mass detection with 3.4 false positives per case.

Conclusions:

• Deep convolutional neural networks (DCNNs) effectively automate breast segmentation and mass detection in DCE-MRI.
• The proposed deep learning approach significantly enhances the accuracy and efficiency of breast cancer analysis.
• These automated methods show promise in assisting radiologists and reducing manual workload in clinical practice.