3-D Digital Imaging of Breast Calcifications: Improvements in Image Quality, and Development of Automated Reconstruction Methods

Abstract

In our work to date, we have generated a manually segmented and paired dataset of 110 patients images, which we have used as a "gold standard' in the evaluation of computer algorithms for identifying, segmenting and correlating calcifications. We have been able to develop two separate computer algorithms, one for identification and segmentation of potential calcifications, the other to find calcification triplets that should be paired. Both algorithms are quite robust. There are a number of significant findings from this work that will be published. First, the use of Euler's number to determine connectivity in an automated fashion is unique. Secondly, the simultaneous correction of patient motion and the determination of correspondence between the views is unique, and will be published. At this time, the algorithms are as good as the human observer. However, it now appears that the algorithm could be relatively easily improved. Similarly, the comparison to the human observer can be improved.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Feb 01, 2000
Accession Number
ADA391387

Entities

People

  • Andrew D. Maidment

Organizations

  • Thomas Jefferson University

Tags

Communities of Interest

  • Biomedical

DTIC Thesaurus Topics

  • Algorithms
  • Breast Cancer
  • Cancer
  • Computer Vision
  • Computers
  • Data Sets
  • Detectors
  • Geometry
  • Health Services
  • Identification
  • Image Reconstruction
  • Materials
  • Medical Personnel
  • Observers
  • Physicians
  • Three Dimensional
  • Two Dimensional

Fields of Study

  • Medicine

Readers

  • Computational Modeling and Simulation
  • Medical Imaging.
  • Regression Analysis.