Related Experiment Videos
Mass discrimination in mammography: experiments using hybrid images
Arthur Burgess1, Francine Jacobson, Philip Judy
1Department of Radiology, Brigham and Women's Hospital, Harvard Medical School, 75 Francis St, Boston, MA 02115, USA.
Academic Radiology
|November 25, 2003
Summary
This study investigated how mass size affects the contrast needed to discriminate between different types of breast lesions in mammograms. Higher contrast is required for larger masses, suggesting complex visual processing is involved.
Area of Science:
- Medical Imaging
- Radiology
- Human Visual Perception
Background:
- Previous research indicated mammogram contrast requirements for mass detection increase with size.
- Understanding contrast-detail relationships is crucial for accurate mammographic interpretation.
Purpose of the Study:
- To quantify the amplitude threshold for discriminating between breast masses as a function of lesion size.
- To compare discrimination thresholds with previously reported detection thresholds.
Main Methods:
- A hybrid imaging approach using digitized masses added to normal mammograms.
- Observer experiments employed a two-alternative forced-choice method on a computer monitor.
- Tasks included discriminating between ductal carcinoma and fibroadenoma, and between two ductal carcinomas, across lesion sizes from 1 to 16 mm.
Main Results:
- Discrimination contrast-detail diagrams showed minimum threshold amplitudes for lesion sizes around 4 mm.
- For masses larger than 4 mm, discrimination thresholds were 1.5-2 times higher than detection thresholds.
- Detection results exhibited an unusual contrast-detail form where thresholds increased with lesion size, contrary to typical image noise behavior.
Conclusions:
- A simple signal detection model with visual system characteristics explained detection results but not discrimination findings.
- Additional complexity is needed in observer models to account for discrimination performance variations with lesion size.