Related Experiment Video
Updated: Dec 17, 2025
15:48
Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging
Published on: December 15, 2014
22.9K
A deep learning methodology for improved breast cancer diagnosis using multiparametric MRI
Qiyuan Hu1, Heather M Whitney2,3, Maryellen L Giger2
1Committee on Medical Physics, Department of Radiology, The University of Chicago, 5841 S Maryland Ave., Chicago, IL, MC202660637, USA. qhu@uchicago.edu.
Scientific Reports
|July 1, 2020
Summary
This study developed a deep learning method using multiparametric MRI to diagnose breast cancer. Feature fusion significantly improved diagnostic accuracy, potentially reducing false positives in breast imaging.
Area of Science:
- Radiology
- Artificial Intelligence
- Oncology
Background:
- Multiparametric magnetic resonance imaging (mpMRI) enhances breast cancer diagnosis.
- Machine learning, specifically deep transfer learning, offers potential for automated diagnostic tools.
Purpose of the Study:
- To develop and evaluate a deep transfer learning computer-aided diagnosis (CADx) system for breast cancer detection using mpMRI.
- To compare different fusion strategies for integrating DCE-MRI and T2w-MRI sequences.
Main Methods:
- A retrospective study analyzed 927 lesions from 616 women using mpMRI (DCE and T2w sequences).
- A convolutional neural network (CNN) extracted features, and support vector machines (SVMs) classified lesions.
- Three fusion methods (image, feature, classifier) were investigated.
Main Results:
- Single-sequence classifiers achieved AUCs of 0.85 (DCE) and 0.78 (T2w).
- Multiparametric fusion methods showed improved performance, with feature fusion yielding an AUC of 0.87.
- Feature fusion statistically significantly outperformed DCE alone (P < 0.001).
Conclusions:
- The proposed deep transfer learning CADx method for mpMRI shows promise in improving breast cancer diagnostic performance.
- This approach may enhance diagnostic accuracy by reducing false positives and increasing the positive predictive value in breast imaging interpretation.
