Anomaly detection in digital mammography using neural networks
Keywords:
Mammography, Anomaly Detection, Neural Network, Gaussian Mixture Model.Abstract
A PC supported recognition and finding the framework for bosom malignant growth, the most well-known type of disease
among ladies, utilizing mammography. The framework depends on the Generalized Regression Neural Networks
worldview, which has demonstrated value for therapeutic choice help in past works from our group. In the proposed
system, bosoms are first divided adaptively into districts. The GLCM Features are extricated from wavelet subgroups. At
that point highlights got from the recognition of injuries (masses and smaller scale calcifications) just as textural highlights,
are removed from every district and joined so as to group mammography examinations as "Benign", "Malignant" or
"Normal". At whatever point other than the ordinary record is recognized, the areas that actuated that mechanized finding
can be featured. Two techniques are assessed to characterize this irregularity locator. In a first situation, manual divisions
of sores are utilized to prepare a NN that appoints an irregularity list to every district; neighbourhood peculiarity lists are
then joined into a worldwide inconsistency list.
