Related Experiment Video
Updated: Jan 1, 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
Differentiating benign and malignant mass and non-mass lesions in breast DCE-MRI using normalized frequency-based
Fazael Ayatollahi1,2, Shahriar B Shokouhi3, Jonas Teuwen4,5
1Electrical Engineering Department, Iran University of Science and Technology (IUST), Tehran, Iran. fazael_ayatollahi@elec.iust.ac.ir.
International Journal of Computer Assisted Radiology and Surgery
|December 16, 2019
Summary
This study introduces novel frequency textural features for computer-aided diagnosis (CADx) of breast lesions in DCE-MRI. The new method effectively distinguishes between malignant and benign mass and non-mass lesions with high accuracy.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Biomedical Engineering
Background:
- Breast cancer diagnosis relies heavily on medical imaging like DCE-MRI.
- Distinguishing between benign and malignant lesions remains a challenge.
- Computer-aided diagnosis (CADx) systems can aid radiologists in lesion characterization.
Purpose of the Study:
- To develop a novel computer-aided diagnosis (CADx) system for breast DCE-MRI.
- To differentiate between malignant and benign mass and non-mass lesions.
- To introduce and evaluate new frequency textural features for lesion characterization.
Main Methods:
- Utilized the dual-tree complex wavelet transform to extract normalized frequency-based features from MRI slices.
- Employed a support vector machine (SVM) classifier for lesion classification.
- Investigated strategies to handle class imbalance, including modified cost functions and sampling techniques.
Main Results:
- Achieved an area under the curve (AUC) of 0.98 for mass lesions and 0.94 for non-mass lesions.
- Reported high diagnostic performance, including accuracies of 96.9% (mass) and 89.8% (non-mass).
- Demonstrated excellent sensitivity and specificity for both lesion types.
Conclusions:
- Normalized frequency-based features are efficient for characterizing benign and malignant breast lesions in DCE-MRI.
- The proposed CADx method shows high potential for clinical application.
- Combining frequency features with 3D shape descriptors further enhances diagnostic performance.
