Classification of Architectural Distortion from Other Abnormalities in Mammograms

Screening mammograms is a repetitive task that causes fatigue and eye strain. For thousands of cases analyzed by a radiologist, not more than ten are cancerous and thus an abnormality may be overlooked. Computer-aided detection (CAD) algorithms are developed to assist radiologists in detecting mammographic abnormalities. In this paper, a CAD system is developed to classify Architectural Distortion abnormality from other malignant abnormalities and normal mammogram samples. Gabor features and Law’s Texture Energy measures are used to detect architectural distortion. Classification is done using Support Vector Machines (SVM). SVM identifies the architectural distorted sample from other samples. SVM is implemented using RBF kernel function. Algorithm is tested for two datasets one set which includes all abnormalities except speculated masses and other set includes all abnormalities along with speculated masses.