Automated Method for Analysis of Mammographic Breast Density - A Technique for Breast Cancer Risk Estimation

Abstract

The goal of this proposed project is to develop an automated technique to assist radiologists in estimating mammographic breast density. The computerized image analysis tool can provide a consistent and reproducible estimation of percent dense area on routine clinical mammograms, thereby contributing to the understanding of the relationship of mammographic density to breast cancer risk, detection, and prognosis, and the prevention and treatment of breast cancer. During this project year, we completed the observer performance study that compared the mammographic density on digital mammograms (DMs) and digitized film mammograms fDFMs) manually segmented by experienced radiologists. The average mammographic density on digital mammograms was found to be significantly lower than that on digitized mammograms as perceived by radiologists. The lower density on DMs may improve the mammographic sensitivity for lesion detection on dense breasts. However, for patients with DFMs and DMs taken over time, comparison of serial mammograms for breast density changes will be problematic. We also performed a study to compare mammographic density segmented from DMs by our automated density segmentation program, referred to as Mammographic Density ESTimator (MDEST), with the average mammographic density manually segmented by four experienced radiologists. The mammographic density from DMs obtained by our automated MDEST program is highly correlated with the average mammographic density manually segmented by radiologists. The study therefore indicates the feasibility of using our MDEST program for estimation of mammographic density, although improvement is needed to further increase the segmentation accuracy. We will continue to improve the image analysis methods in the coming year.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jul 01, 2004
Accession Number
ADA428513

Entities

People

  • Heang-Ping Chan

Organizations

  • University of Michigan

Tags

Communities of Interest

  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Accuracy
  • Algorithms
  • Breast Cancer
  • Computer Programs
  • Computers
  • Data Sets
  • Detection
  • Electronic Mail
  • Estimators
  • Graphical User Interface
  • Image Processing
  • Medical Personnel
  • Neoplasms
  • Three Dimensional
  • Two Dimensional
  • United States
  • User Interface

Fields of Study

  • Medicine
  • Physics

Readers

  • Computer Vision.
  • Oncology and Biomarker-Based Cancer Detection.
  • Statistical inference.