Accession Number : ADA321763

Title :   Computer-Aided Diagnosis and Automated Screening of Digital Mammogram.

Descriptive Note : Annual rept. 1 Sep 95-31 Aug 96,

Corporate Author : UNIVERSITY OF SOUTH FLORIDA TAMPA

Personal Author(s) : Woods, Kevin S.

PDF Url : ADA321763

Report Date : SEP 1996

Pagination or Media Count : 16

Abstract : The primary objective of our on-going research is to develop computer techniques for detecting abnormalities in digital mammograms which can be used by radiologists to increase diagnostic accuracy and decrease the rate of undetected cancers. Our efforts in the second full year of research have focused on the initial stages of such a system. First, a general view of detection algorithms is presented. The basic framework involves two phases, pixel-level segmentation and region-level classification, each composed of a few elementary steps. By viewing several recent detection algorithms in the context of this general framework, we are able to show some fundamental advantages to concentrating efforts on the early pixel-level analysis of the segmentation phase. Secondly, we are developing an automated technique for quantifying and characterizing breast tissue structure. Unlike most previous efforts, the purpose of our work will be to assign a "difficulty index" to the mammogram image based on texture analysis and density of breast parenchymal tissue. The difficulty index will be incorporated as an additional parameter in a mammogram image analysis system so that decisions may be effectively conditioned on the density measure.

Descriptors :   *DIGITAL SYSTEMS, *COMPUTER AIDED DIAGNOSIS, *BREAST CANCER, *MAMMOGRAPHY, ALGORITHMS, IMAGE PROCESSING, DENSITY, MEASUREMENT, TISSUES(BIOLOGY), AUTOMATION, DETECTION, CLASSIFICATION, INDEXES, PIXELS, TEXTURE, ABNORMALITIES, SEGMENTED, MAMMARY GLANDS, CONNECTIVE TISSUE.

Subject Categories : Medicine and Medical Research
      Numerical Mathematics
      Cybernetics
      Optics

Distribution Statement : APPROVED FOR PUBLIC RELEASE