Advertisement

Masses

  • Ralph Highnam
  • Michael Brady
Part of the Computational Imaging and Vision book series (CIVI, volume 14)

Abstract

This chapter presents a number of preliminary explorations of applying automated image analysis to masses, abnormal anatomical structures that often indicate breast pathology. In particular, we present results for three important applications:
  • Matching temporal pairs: “ Salient” regions are extracted independently in two mammograms of the same breast and the same view taken at two different times. The salient regions in the later mammogram are then matched with the isointensity regions in the first, and those that have either appeared in the later mammogram but not in the earlier one, or which have changed significantly between the two mammograms are drawn to the attention of the radiologist.

  • Matching bilateral pairs: In a similar fashion, salient regions are extracted from the same-view left and right breast mammograms of the same woman, taken at approximately the same time. They are matched by an algorithm that draws the radiologist’s attention to those that appear in one breast but not in the other.

  • Locating abnormal regions: Salient regions are extracted from a breast image and for each a number of features are computed. These features are used to assess the “normality” of the region, as defined by a probabilistic model learned automatically from approximately 5000 bright regions found in the Mammographic Image Analysis Society (MIAS) database [231] of images. If a salient region is determined to be in the tails of the probabilistic model, it is declared to be “abnormal” and is drawn to the attention of the radiologist.

Keywords

Current Image Bright Region Salient Region Previous Image Novelty Detection 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Copyright information

© Springer Science+Business Media Dordrecht 1999

Authors and Affiliations

  • Ralph Highnam
    • 1
  • Michael Brady
    • 1
  1. 1.Department of Engineering ScienceOxford UniversityOxfordUK

Personalised recommendations