Related Experiment Video
Updated: Aug 11, 2026
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
Knowledge-based computer-aided detection of masses on digitized mammograms: a preliminary assessment
Y H Chang1, L A Hardesty, C M Hakim
1Department of Radiology, University of Pittsburgh, Pennsylvania 15261-0001, USA.
Medical Physics
|May 8, 2001
Summary
A new knowledge-based computer-aided detection (CAD) scheme significantly reduces false positives in mammogram mass identification. This approach improves accuracy by learning from known masses, enhancing diagnostic performance.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Medicine
- Radiology
Background:
- Mammography is crucial for early breast cancer detection.
- Computer-aided detection (CAD) systems aim to improve mammogram interpretation accuracy.
- Rule-based CAD schemes can generate false positives, necessitating further refinement.
Purpose of the Study:
- To develop and evaluate a knowledge-based computer-aided detection (CAD) scheme for enhancing mass identification in mammograms.
- To improve the specificity of existing CAD systems by reducing false-positive findings.
- To create a system that learns from a database of known masses to better distinguish true positives from false positives.
Main Methods:
- A knowledge-based approach was developed to prune suspicious regions identified by a rule-based CAD scheme.
- A learning process established a knowledge base by quantitatively characterizing known masses.
- Similarity measures and a composite likelihood were derived to assess suspicious regions.
- Receiver-operating characteristic (ROC) analyses were used to evaluate performance.
Main Results:
- The knowledge-based CAD scheme achieved an area under the ROC curve of 0.83 on the development set (600 regions).
- Fifty-one percent of false-positive regions were eliminated while maintaining 90% sensitivity.
- An area under the ROC curve of 0.80 was achieved on an independent test set (1,000 regions).
Conclusions:
- Knowledge-based approaches can significantly reduce false-positive detections in mammography CAD systems.
- This method maintains reasonable sensitivity, offering potential improvements over existing rule-based CAD schemes.
- The developed CAD scheme shows promise for enhancing the overall performance and clinical utility of mammogram analysis.
