Automated Spot Mammography for Improved Imaging of Dense Breasts
Abstract
We are developing an automated stereo spot mammography technique to improve imaging of lesions within dense breast tissue. During the fourth year of this project, our work was devoted primarily to: I) completing our observer study comparing the suspicious spot regions selected by radiologists to those detected by a computer (CAD), 2) performing a study comparing the radiologist selected regions and CAD selected regions to true regions determined by a radiologist from mammograms, biopsy images and pathology results, 3) defining the geometric requirements for the automated collimator to insure the suspicious region in a full-field digital mammogram is adequately covered with stereo spot image acquisition, and 4) obtaining our first images of a modular breast phantom that was manufactured for our spot mammography experiments. The images of the phantom showed it did not satisfy our design requirements with respect to dense regions overlapping the simulated masses, and with respect to the phantom producing an image texture similar to that of an actual mammogram. We therefore were not able to complete our planned phantom experiments. The manufacturer has promised to build us a correct phantom that we will employ in experiments during an extended year of the project.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 2003
- Accession Number
- ADA421767
Entities
People
- Mitchell M. Goodsitt
Organizations
- University of Michigan