Related Experiment Video
Updated: Sep 3, 2025
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
43.0K
BI-RADS-Based Classification of Mammographic Soft Tissue Opacities Using a Deep Convolutional Neural Network
Albin Sabani1, Anna Landsmann1, Patryk Hejduk1
1Institute of Diagnostic and Interventional Radiology, University Hospital of Zurich, University of Zurich, 8091 Zurich, Switzerland.
Diagnostics (Basel, Switzerland)
|July 27, 2022
Summary
This study demonstrates that a deep convolutional neural network (dCNN) can accurately classify breast cancer soft tissue opacities in mammograms. The AI model achieved high specificity, mimicking human radiologist performance for BI-RADS classification.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Mammography is crucial for breast cancer screening.
- Classifying soft tissue opacities is key for accurate diagnosis.
- Current methods can be subjective and observer-dependent.
Purpose of the Study:
- To evaluate a deep convolutional neural network (dCNN) for classifying breast cancer soft tissue opacities.
- To assess the dCNN's performance independently of other mammographic features.
- To compare the dCNN's diagnostic accuracy against experienced radiologists.
Main Methods:
- A dCNN was trained and validated on 1744 mammograms from 438 patients.
- Soft tissue opacities were categorized using the ACR BI-RADS atlas.
- The dCNN's diagnostic performance was benchmarked against human readers on a separate dataset.
Main Results:
- The dCNN achieved accuracies ranging from 73.8% to 89.8% on the test dataset.
- The dCNN demonstrated superior specificity (100%) compared to human readers.
- Sensitivity of the dCNN (84.0%) was comparable to human readers.
Conclusions:
- A dCNN can automatically detect and classify soft tissue opacities in mammograms.
- The AI model provides standardized, observer-independent classification.
- Artificial intelligence can effectively replicate human decision-making in BI-RADS classification.
