Computer-Assisted Visual Search/Decision Aids as a Training Tool for Mammography

Abstract

The primary goal of the project is to develop a computer-assisted visual search (CAVS) mammography training tool that will improve the perceptual and cognitive skills of trainees leading to mammographic expertise. In the first two years we carried out two experiments. The first equated experience by comparing perceptual skills of expert radiologists with lay people searching non-medical pictorial scenes for hidden targets. Results show that expert radiology search and detection strategies do not transfer to the non-medical search and detection tasks. In the second study, a 75-case mammogram test set was administered to mammographers, residents and mammography technologists. The results compared effectiveness of experience and training at different levels of expertise. Not surprisingly, resident performance in detecting and classifying breast lesions was significantly inferior to experts, and no better than that of mammography technologists. In the third year we carried out an experiment to determine if retrospectively-visible cancers in mammograms attract visual attention and are accurately recognized in a blinded review. Results comparing a test set of 40 retrospectively-visible vs. directly-visible cancer cases indicated that not only did retrospectively-visible cancers fail to attract visual attention, but that they also led to higher false-positive error rates than did directly-visible cancers.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jul 01, 2000
Accession Number
ADA388032

Entities

People

  • Calvin Nodine

Organizations

  • University of Pennsylvania

Tags

Communities of Interest

  • Energy and Power Technologies
  • Human Systems

DTIC Thesaurus Topics

  • Artificial Intelligence
  • Artificial Intelligence Software
  • Automata Theory
  • Cognition
  • Cognitive Science
  • Computer Languages
  • Computer Vision
  • Computers
  • Databases
  • Health Services
  • Human-Machine Interaction
  • Information Processing
  • Information Science
  • Machine Learning
  • Medical Personnel
  • Psychology
  • Reasoning

Readers

  • Artificial Intelligence
  • Brain and Cognitive Science; Experimental Psychology; Cognitive Neuroscience
  • Oncology and Biomarker-Based Cancer Detection.