Related Experiment Video
Updated: Dec 29, 2025
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
43.5K
Automatic detection of simulated motion blur in mammograms
1Department of Biomedical Engineering, The George Washington University, Washington, DC, 20052, USA.
Medical Physics
|February 5, 2020
Summary
Machine learning and blur measure operators can automatically detect motion blur in mammograms, improving diagnostic accuracy. This technology supports immediate retakes, preventing missed diagnoses and reducing patient anxiety.
Area of Science:
- Medical Imaging
- Machine Learning
- Radiology
Background:
- Motion blur in mammograms can reduce lesion detection performance.
- Subtle motion blur may go undetected by human observers.
- This can lead to delayed diagnosis of breast cancer.
Purpose of the Study:
- To develop an automated system for detecting motion blur in mammograms.
- To utilize machine learning algorithms and blur measure (BM) operators for blur detection.
- To support clinical decision-making by enabling immediate mammogram retakes.
Main Methods:
- Simulated motion blur using a point-spread-function (PSF) mask based on random motion.
- Investigated effects of tissue elasticity, exposure time, and motion boundary.
- Trained three machine learning classifiers (Ensemble Bagged Trees, SVM, KNN) on blurred and unblurred mammograms from INbreast and DDSM databases.
Main Results:
- Achieved high classification accuracies in detecting simulated motion blur.
- Average accuracies ranged from 85.7% to 93.6% across different classifiers and datasets.
- Ensemble Bagged Trees and fine Gaussian SVM showed strong performance.
Conclusions:
- Machine learning and BM operators show potential for automatic motion blur detection in mammograms.
- This automated detection could enhance diagnostic reliability.
- Further research is needed to quantify the impact of motion blur on radiologist performance.
