Related Experiment Videos
Computerized detection of clustered microcalcifications: evaluation of performance on mammograms from multiple
R M Nishikawa1, K Doi, M L Giger
1Kurt Rossmann Laboratories for Radiologic Image Research, Department of Radiology, University of Chicago, IL 60637, USA.
Summary
A computerized method for detecting clustered microcalcifications in mammograms showed lower performance on diverse screening center images. Customization to local image characteristics is crucial for optimal clinical performance.
Area of Science:
- Medical Imaging
- Radiology
- Computer-Aided Diagnosis
Background:
- Automated detection of clustered microcalcifications is vital for early breast cancer diagnosis.
- Digital mammography performance can vary based on image acquisition and processing across different screening centers.
Purpose of the Study:
- To evaluate the performance of an automated clustered microcalcification detection algorithm using digitized mammograms from various screening sites.
- To compare the algorithm's performance on real-world screening data versus a standardized database.
Main Methods:
- Radiologists submitted mammograms (n=43) from 14 sites to a scientific exhibit for digitization and analysis.
- Algorithm performance was assessed on these diverse cases and compared to a standard database (n=39).
- Image characteristics such as contrast and noise were considered as potential influencing factors.
Main Results:
- The detection algorithm's performance was lower on mammograms from the diverse screening centers compared to the standard database.
- Subtle microcalcifications and variations in image characteristics (contrast, noise) contributed to reduced performance.
- Algorithm performance demonstrated dependence on center-specific image properties.
Conclusions:
- The computerized detection algorithm is considered robust and accurate for potential clinical testing.
- Clinical implementation requires customization of the algorithm to the specific image characteristics of each screening center for optimal results.
- Further research should focus on adaptive algorithms that can account for inter-center variability in mammogram quality.