Computer-Aided Classification of Malignant and Benign Lesions on Mammograms

Abstract

New methods have been investigated for computerized characterization of mammographic masses and microcalcifications. A mass segmentation method based on an active contour model was developed. The resulting segmentation algorithm was shown to be within the inter-observation variation of radiologists' hand segmentation. Morphological, texture, and spiculation features were extracted from the segmented mass and its margins. New classifiers based on statistical methods, genetic algorithms and neural networks were developed. The high-sensitivity classifier developed in this project was shown to have a significantly higher sensitivity than competing classifiers at the same specificity levels. The effect of the mass classification algorithm on radiologists' classification was evaluated using an observer study. It was shown that the radiologists' classification was significantly improved when they were aided by the computerized classification scores. A microcalcification detection algorithm was applied for automated detection of individual microcalcifications with a region of interest. The individual microcalcifications were segmented from the background using an automated algorithm. Morphological and texture features were extracted from computer-detected microcalcifications, and were used in a statistical classifier to distinguish between malignant and benign microcalcifications. Using an observer performance study, it was shown that the developed automated microcalcification characterization method was significantly more accurate than experienced radiologists.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
May 01, 2001
Accession Number
ADA400645

Entities

People

  • Berkman Sahiner

Organizations

  • University of Michigan

Tags

Communities of Interest

  • Energy and Power Technologies
  • Human Systems

DTIC Thesaurus Topics

  • Artificial Intelligence
  • Artificial Intelligence Software
  • Computational Science
  • Computer Vision
  • Computers
  • Data Mining
  • Data Science
  • Databases
  • Feature Extraction
  • Health Services
  • Information Processing
  • Information Science
  • Knowledge Management
  • Machine Learning
  • Medical Personnel
  • Neural Networks
  • Self Organizing Systems

Readers

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

Technology Areas

  • AI & ML
  • Biotechnology
  • Biotechnology - Cancer Biotech