Malignancy and abnormality detection of mammograms using discrete wavelet transformed features and neural network

Abstract

Mammograms can be used to check for breast cancer in women. In this paper, we have proposed breast cancer detection into two stages. In the first stage, mammograms have to classify into malignant and benign. While in second stage, the type of abnormality is detected. Features have been extracted using Discrete Wavelet Transform. These wavelet based features has been reduced using Principle Component Analysis. Those images which have been classified as malignant in the first stage are further classified into six classes to check its abnormality. It has been observed that the accuracy of classification of abnormalities is more than 90%. Mammographic Institute Society Analysis dataset is used for experimentation.