Cancers Missed on Mammography.

Abstract

Our goal is to decrease the estimated 30% of breast cancers missed on screening mammograms by radiologists, by applying computer-aided diagnosis (CAD) techniques to mammogram interpretation. A database of 100 screening mammograms containing cancers that were observational misses in routine practice will be collected and analyzed for pathology, mammographic lesion type, breast density, location, subtlety, size, growth rate, and reasons for having been overlooked. These cases will be digitized along with 300 normals, and presented to 12 general radiologists and 3 experts, to evaluate observer performance, once with and once without CAD prompts pointing at suspicious areas. The radiologists' readings will be compared for sensitivity, specificity, and positive and negative predictive value. The hypothesis is that CAD can decrease the number of misses by up to 50%, and that we can measure this using a unique database of cancers containing a large fraction of cancers that a radiologist might miss without CAD help. In evaluating interobserver variability of readers, we anticipate that CAD second opinions can help average readers perform closer to experts. The results will be used to justify and plan the future implementation of CAD in the form of commercial products for practical use in clinical mammographic practice.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Oct 01, 1997
Accession Number
ADA338762

Entities

People

  • Robert Schmidt

Organizations

  • University of Chicago

Tags

DTIC Thesaurus Topics

  • Biomedical Research
  • Breast Cancer
  • Cancer
  • Computer Programs
  • Computer-Aided Diagnosis
  • Computers
  • Databases
  • Detection
  • Health Care
  • Mammography
  • Materials
  • Medical Personnel
  • Neoplasms
  • Observers
  • Physicians
  • Sensitivity
  • Standards

Fields of Study

  • Medicine
  • Physics

Readers

  • Oncology and Biomarker-Based Cancer Detection.