Ultrasonic Morphological Analyzers for Breast Cancer Diagnosis

Abstract

The goal of this research is to improve ultrasonic classification of breast lesions and guide decisions regarding biopsy requirements, especially for small lesions and those in young, dense breasts, which are particularly difficult to evaluate with mammography. The research is developing a set of complementary ultrasonic morphological analysis procedures (UMAPs) that analyze digital ultrasonic echo data. Each UMAP extracts a particular quantitative lesion feature that is now subjectively described; these include "echogenicity," "heterogeneity," "shadowing," and lesion boundary characteristics. The set of complementary UMAP features will be analyzed with statistical procedures to derive reliable and objective lesion classification. UMAP processing is being applied to digitized radio-frequency echo data previously acquired in 140 clinical breast examinations with linear-array ultrasound systems. Anonymous ancillary patient data (including reports from subsequent biopsies) and system calibration data are stored in archival data files. Also stored are the clinicians' levels-of- suspicion that a lesion is cancerous; these LOS values were based on conventional, subjective scoring of ultrasonograms. After refining each UMAP procedure, lesion classification will be evaluated using multi-parameter discriminant functions. To quantify incremental benefits for breast cancer identification, ROC curves for UMAP classification will be compared to ROC curves based on the clinicians' LOS that a lesion is cancerous.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jul 01, 1999
Accession Number
ADA384099

Entities

People

  • Frederic L. Lizzi

Tags

Communities of Interest

  • Biomedical

DTIC Thesaurus Topics

  • Algorithms
  • Analyzers
  • Boundaries
  • Breast Cancer
  • Calibration
  • Classification
  • Digital Data
  • Frequency
  • Heterogeneity
  • Identification
  • Materials
  • Medical Personnel
  • Neoplasms
  • New York
  • Radio Frequency
  • Spectra
  • Spectrum Analysis

Fields of Study

  • Medicine

Readers

  • Computational Modeling and Simulation
  • Medical Imaging.
  • Oncology and Biomarker-Based Cancer Detection.

Technology Areas

  • AI & ML