Related Experiment Video
Updated: Jan 13, 2026
15:48
Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging
Published on: December 15, 2014
23.0K
Advanced deep learning framework for breast cancer detection using digital breast tomosynthesis images.
G Bharatha Sreeja1, S Sudha2, T M Inbamalar3
1231516 ECE Department, Easwari Engineering College , Chennai, Tamil Nadu, India.
Biomedizinische Technik. Biomedical Engineering
|January 11, 2026
Summary
This study introduces a deep learning framework for detecting breast cancer using digital breast tomosynthesis (DBT) images. The advanced model significantly improved diagnostic performance, offering a promising tool for early breast cancer detection.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Early and accurate breast cancer detection is crucial for improving patient survival rates.
- Digital Breast Tomosynthesis (DBT) offers advanced imaging capabilities for breast cancer diagnosis.
Purpose of the Study:
- To develop a robust deep learning framework for breast cancer detection using DBT images.
- To leverage both single-slice and multi-slice DBT inputs for enhanced detection performance.
Main Methods:
- Image preprocessing including normalization, resizing, and Laplacian Pyramid Enhancement (LPE).
- Feature extraction, fusion, and selection using Exhaustive Feature Selection (EFS).
- A hybrid classification model combining ResNet V2, MobileNet V3, and Inception V3+, with XGBoost ensemble learning and hyperparameter optimization.
Main Results:
- The hybrid model utilizing multi-slice DBT inputs demonstrated superior accuracy, sensitivity, specificity, and AUC.
- Laplacian Pyramid Enhancement (LPE), feature fusion, and Exhaustive Feature Selection (EFS) substantially improved diagnostic performance.
Conclusions:
- Advanced deep learning methods show significant potential for enhancing breast cancer detection performance.
- Future work includes integration with clinical decision support systems, multi-center datasets, and other imaging modalities.
