Computer-Aided Detection of Mammographic Masses in Dense Breast Images

Abstract

This document describes the research tasks and educational activities in which the PI has been engaged during the first phase of this post-doctoral work. The Pr has selected 342 dense breast masses (175 cancer cases and 167 benign cases) for testing using an automated segmentation algorithm which combines region growing with cost function analysis. This method has been validated on 198 cases using overlap, accuracy, sensitivity and specificity statistics by one expert radiologist and has been validated on 46 cases by a second expert radiologist. All statistics have values that range from 0 to 1. The mean overlap and accuracy values for cancer cases are approximately 0.46 and 0.78, respectively. The mean overlap and accuracy values for benign cases are approximately 0.52 and 0.87, respectively. The PI has attended cancer imaging workshops, attended two scientific meetings, made two oral presentations, team taught an Imaging Technologies course, and published two manuscripts during this time.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jun 01, 2004
Accession Number
ADA427116

Entities

People

  • Lisa M. Kinnard

Organizations

  • Howard University

Tags

Communities of Interest

  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Accuracy
  • Algorithms
  • Artificial Intelligence
  • Boundaries
  • Computational Science
  • Computer Science
  • Computer Vision
  • Computers
  • Databases
  • Detection
  • Image Segmentation
  • Information Science
  • Medical Personnel
  • Neural Networks
  • Physicians
  • Statistics
  • Two Dimensional

Fields of Study

  • Medicine
  • Physics

Readers

  • Approximation Theory.
  • Oncology and Biomarker-Based Cancer Detection.
  • Technical Research and Report Writing.