Classification of Breast Lesions based on Laws’ Feature Extraction Techniques

Abstract

Breast lesions are characterized into three classes which include primary benign, primary malignant and secondary malignant. In the present work Laws’ mask texture features are computed from the ultrasound images of the breast lesions. These Laws’ masks of various resolutions i.e., of length 3, 5, 7 and 9 have been used to extract the statistical features (Mean, Standard Deviation, Kurtosis, Skewness and Energy) from Laws’ texture images. In the present work using the SVM classifier, an overall classification accuracy of 88.3% and the individual classification accuracy values of 95.2% , 88.6% and 91.6% have been obtained for primary benign, primary malignant and secondary malignant classes respectively.