Related Experiment Videos
Primary breast abnormalities: selective pixel sampling on dynamic gadolinium-enhanced MR images
S Mussurakis1, P Gibbs, A Horsman
1Centre for MR Investigations, University of Hull, England.
Radiology
|February 11, 1998
Summary
Selective sampling of enhancing areas in breast MRI may not improve diagnostic accuracy. Analyzing the entire lesion with region-of-interest (ROI) methods is as effective as thresholding for detecting breast cancer.
Area of Science:
- Radiology
- Medical Imaging
- Oncology
Background:
- Magnetic resonance imaging (MRI) is crucial for breast cancer diagnosis.
- Region-of-interest (ROI) analysis is a common method for evaluating breast abnormalities on MRI.
- Thresholding, a selective sampling technique, is sometimes used within ROIs.
Purpose of the Study:
- To evaluate the diagnostic utility of a region-of-interest (ROI) analytic method using selective pixel sampling (thresholding) on contrast-enhanced breast MRI.
- To compare thresholding techniques with standard ROI analysis for primary breast abnormalities.
Main Methods:
- Dynamic gadolinium-enhanced MRI scans were acquired for 105 women.
- ROIs were drawn to encompass entire lesions.
- Pixel-by-pixel analysis determined relative signal intensity increase at 1 and 2 minutes postcontrast.
- Thresholding was applied within ROIs, varying lower boundaries from 0% to 100%.
Main Results:
- Analysis included 71 invasive carcinomas and 37 benign lesions.
- Narrower thresholding increased enhancement ratios but also variability.
- Enhancement ratios at 1 minute differed between benign and malignant lesions (P < .0005).
- Thresholding did not improve diagnostic performance in receiver operating characteristic analysis.
Conclusions:
- Selective sampling of highly enhancing pixels within breast lesions does not offer a diagnostic advantage over analyzing the entire lesion with ROIs.
- Standard, lesion-encompassing ROIs are sufficient for diagnostic evaluation of breast abnormalities on contrast-enhanced MRI.