Related Experiment Video
Updated: Aug 29, 2025
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
1.9K
Breast Masses Detection and Segmentation in Full-Field Digital Mammograms using Unified Convolution Neural Network
Summary
This study introduces Unified-CNN, a novel deep learning model for automated breast mass segmentation and detection. It enhances early cancer diagnosis by improving accuracy and efficiency in mammographic analysis.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Breast cancer is a leading cause of mortality in women globally.
- Early detection and treatment significantly improve patient outcomes.
- Automated segmentation and classification of breast tumors are challenging due to tumor variability and image quality.
Purpose of the Study:
- To develop a unified Convolutional Neural Network (CNN) architecture for simultaneous segmentation and detection of breast masses.
- To improve the accuracy and efficiency of automated breast cancer detection in mammography.
Main Methods:
- Designed a novel Unified-CNN architecture with a new convolution module incorporating an additional offset.
- Employed Random Region Selection (RRS) for data augmentation to enhance boundary region selection.
- Utilized ROI pooling for precise boundary detection and optimized model training.
Main Results:
- The Unified-CNN architecture demonstrated high prediction accuracy.
- Evaluated performance using True Positive Rate at False Positive per Image (FPI) and Dice Index on the INBreast dataset.
- Achieved competitive results compared to existing methodologies.
Conclusions:
- The Unified-CNN offers a promising approach for integrated breast mass segmentation and detection.
- The novel convolution module and data augmentation strategy contribute to improved high-level feature extraction and prediction.
- This automated method has the potential to aid radiologists in earlier and more accurate breast cancer diagnosis.
