Digital mammography: towards pectoral muscle removal via independent component analysis

Abstract

The extraction of features for automated assessment for breast cancer detection and diagnosis requires identification of the breast tissue. The pectoral muscle in medio-lateral oblique (MLO) mammogram images is one of the few landmarks in the breast. Yet, it can bias and affect the results of any mammogram processing method. To avoid such effects it is often necessary to automatically identify and segment the pectoral muscle prior to breast tissue image analysis. We propose the use of Independent Component Analysis (ICA) for identification and subsequent removal of the pectoral muscle. The identification is posed as classification of image subsections corresponding to pectoral muscle and breast tissue as represented by a set of ICA basis functions. Average classification rates 97.3% and 83.3% for pectoral muscle and breast tissue respectively have been obtained.