Related Experiment Videos
Feature extraction and classification of dynamic contrast-enhanced T2*-weighted breast image data
G Torheim1, F Godtliebsen, D Axelson
1Department of Anesthesia and Medical Imaging, Norwegian University of Science and Technology, and MR-Centre, Trondheim. geir.torheim@no.nycomed-amersham.com
IEEE Transactions on Medical Imaging
|January 29, 2002
Summary
Rapid T2*-weighted magnetic resonance imaging (MRI) offers a robust method for breast cancer detection. Semi-automatic region of interest (ROI) analysis combined with noise reduction improves classification accuracy, with minimum enhancement indicating malignancy.
Area of Science:
- Radiology
- Medical Imaging
- Oncology
Background:
- Dynamic contrast-enhanced T1-weighted MRI for breast cancer has limited specificity.
- Investigating alternative imaging techniques like rapid T2*-weighted MRI is crucial.
- Susceptibility artifacts and motion complicate T2*-weighted breast MRI analysis.
Purpose of the Study:
- To evaluate the efficacy of rapid T2*-weighted imaging for breast tumor analysis.
- To compare different classification methods, including semi-automatic and manual region of interest (ROI) definitions.
- To assess the impact of noise reduction on image analysis.
Main Methods:
- 127 breast tumor patients underwent rapid, single-slice T2*-weighted MRI.
- Leave-one-out cross-validation was used to test classification methods.
- Semi-automatic ROI definition was compared to manual ROI definition; pixel-based analysis was also performed.
Main Results:
- Minimum enhancement parameter demonstrated high robustness and accuracy.
- Semi-automatic ROI definition was efficient and yielded comparable results to manual methods.
- Noise reduction enhanced sensitivity and specificity, though not significantly; pixel-based analysis did not improve classification.
Conclusions:
- Rapid T2*-weighted breast MRI analysis is feasible using semi-automatic ROI tools and noise reduction.
- Minimum enhancement serves as a reliable indicator of malignancy in T2*-weighted imaging.
- This approach offers a rapid and robust method for breast cancer assessment.