Date of Award

12-1994

Document Type

Thesis

Degree Name

Master of Science

Department

Department of Electrical and Computer Engineering

First Advisor

Steven K. Rogers, PhD

Abstract

This research advances computer-aided breast cancer diagnosis. More than 50 million women over the age of 40 are currently at risk from this disease in the United States. Computer-aided diagnosis is offered as a second opinion to radiologists to aid in decreasing the number of false readings of mammograms. This automated tool is designed to enhance detection and classification. New feature extraction methods are presented that provide increased classification power. Angular second moment, a second-order gray-level histogram statistic, provides baseline accuracy. Two novel extraction methods, eigenmass and wavelets, are introduced to the field. Based on the Karhunen-Loeve Transform, eigenmass features are developed using eigenvectors to alter the data set into new coefficients. Wavelets, previously only exploited for their segmentation benefits, are explored as features for classification. Daubechies-4, Danbechies-2O, and biorthogonal wavelets are each investigated. Applied to 94 difficult-to-diagnosis digitized microcalcification cases, performance is 74 percent correct classifications. Feature selection techniques are presented which further improve performance. Statistical analysis, neural and decision boundary-based methods are implemented, compared, and validated to isolate and remove useless features. The contribution from this analysis is an increase to 88 percent correct classification in system performance.

AFIT Designator

AFIT-GSO-ENG-94D-03

DTIC Accession Number

ADA289213

Comments

The author's Vita page is omitted.

Download

Share

COinS