Related Experiment Videos
Automated segmentation of digitized mammograms
U Bick1, M L Giger, R A Schmidt
1Kurt Rossmann Laboratories for Radiologic Image Research, Department of Radiology, University of Chicago, IL, USA.
Academic Radiology
|January 1, 1995
Summary
A new automated algorithm accurately segments digital mammograms, identifying breast regions for computer-aided diagnosis. This reliable segmentation is crucial for enhancing mammography analysis and improving diagnostic accuracy.
Area of Science:
- Medical Imaging
- Computer-Aided Diagnosis
- Biomedical Engineering
Background:
- Accurate segmentation of digital mammograms into breast and non-breast regions is essential for subsequent image analysis.
- Existing segmentation methods may be limited by digitizing system, image orientation, or projection type.
Purpose of the Study:
- To develop a fully automated segmentation algorithm for digital mammograms.
- To ensure the algorithm operates independently of digitizing system, image orientation, and projection.
Main Methods:
- The algorithm identifies unexposed and direct-exposure image regions.
- It generates a border defining the valid breast region for further analysis.
- Tested on 740 digitized mammograms and evaluated by expert mammographers and medical physicists.
Main Results:
- The segmentation algorithm achieved acceptable results for computer-aided diagnosis in 97% of mammograms.
- Segmentation issues in 2.9% of cases were primarily due to digitization artifacts or poor mammographic technique.
Conclusions:
- The developed algorithm is a viable component for intelligent workstations in mammography.
- It facilitates automated and reliable image analysis for computer-aided diagnosis.