Computer-Aided Characterization of Breast Masses on Volumetric Ultrasound Images: An Adjunct to Mammography

Abstract

The purpose of this project is to develop computer techniques for the analysis of breast ultrasound images, and to combine computerized sonographic and mammographic analyses. The techniques developed in this project are aimed at providing a second opinion to the radiologists in the task of making a biopsy recommendation. In the no-cost time extension year, we have completed and analyzed the observer performance studies proposed for the project. Our results indicate that our CAD system can significantly improve radiologists' likelihood of malignancy estimates. This result has been consistent in the two observer studies that we conducted, the first study involving US volumes only, and the second study involving both US images and mammograms. Our results also indicate that if we select a likelihood of malignancy cutoff threshold for biopsy recommendation with CAD so that the sensitivities with and without CAD are equal, 4 out of 5 radiologists showed significant improvement in their specificity in the first observer study. In the second study 3 out of 10 radiologists showed significant improvement in their specificity under the same condition. Further improvement of our methods can provide radiologists with a powerful aid for decision making, which may help reduce benign biopsies and improve patient care.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Oct 01, 2005
Accession Number
ADA443058

Entities

People

  • Berkman Sahiner

Organizations

  • University of Michigan

Tags

Communities of Interest

  • Biomedical

DTIC Thesaurus Topics

  • Acquisition
  • Algorithms
  • Biomedical Research
  • Breast Cancer
  • Cancer
  • Carcinoma
  • Computers
  • Data Sets
  • Department Of Defense
  • Graphical User Interface
  • Health Services
  • Medical Personnel
  • Neoplasms
  • Observers
  • Three Dimensional
  • Two Dimensional
  • Ultrasounds

Fields of Study

  • Medicine
  • Physics

Readers

  • Medical Imaging.
  • Regression Analysis.
  • Software Engineering.