Related Experiment Video
Updated: Jun 6, 2026
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
Exploring the potential of context-sensitive CADe in screening mammography
Georgia D Tourassi1, Maciej A Mazurowski, Brian P Harrawood
1Department of Radiology, Carl E. Ravin Advanced Imaging Laboratories, Duke University Medical Center, Durham, North Carolina 27705, USA. georgia.tourassi@duke.edu
Medical Physics
|December 17, 2010
Summary
A new context-sensitive computer-assisted detection (CADe) system personalizes decision support for radiologists interpreting screening mammograms. This advanced CADe system improves sensitivity and reduces errors compared to conventional methods.
Area of Science:
- Radiology
- Medical Imaging
- Artificial Intelligence in Healthcare
Background:
- Conventional computer-assisted detection (CADe) systems offer uniform decision support in screening mammography.
- Personalized decision support tailored to individual user patterns is lacking in current systems.
Purpose of the Study:
- To investigate the efficacy of a context-sensitive CADe system.
- To evaluate decision support guided by radiologists' focus of attention and reporting patterns.
Main Methods:
- An observer study involving six radiologists evaluating 20 screening mammograms with eye-tracking.
- Collection of eye-position data and diagnostic decisions.
- Analysis using a knowledge-based CADe system in conventional and context-sensitive modes.
Main Results:
- Conventional CADe achieved 85.7% sensitivity with 3.15 false positives per image (FPsI).
- Context-sensitive CADe provided 85.7%-100% sensitivity with 0.35-0.40 FPsI.
- Context-sensitive CADe improved radiologist sensitivity and reduced performance gaps.
Conclusions:
- Context-sensitive CADe shows potential for improving diagnostic interpretation of screening mammograms.
- This system effectively delineates and reduces perceptual and cognitive errors in radiologists.
- Personalized decision support enhances mammogram interpretation accuracy.
