Classification of microcalcifications in digitised mammograms
Date
2014-04-30
Authors
Kramer, Dani
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
In this investigation ^number of image texture analysis techniques for the classification
of microcalcifications in digitised mammograms are presented. Microcalcifications
are often an early indication of breast cancer, and computer-aided diagnostic
techniques are capable of improving diagnostic accuracy. Three categories of image
texture features are extracted from regions of interest surrounding clusters of
microcalcifications. These comprise a set of statistical texture features based on
the co-occurrence matrix, a set of wavelet-based texture signatures and a propose^
third set of texture features. This set, referred to as multiscale statistical texture
features, is based on a combination of the other two approaches to texture analysis.
The multiscale statistical texture features outperform the other types of texture
features in tests using two separate datasets and a k-nn classifier for classification.
Improved classification accuracy is also achieved using an artificial neural network
for classification.