Elastogram features selection and classification based on mRMR and SVM

Abstract:

For evaluating elastogram objectively, image processing and pattern recogniton techniques are pro- posed. First the real elasticity information encoded in color was extracted by transform the image from RGB color space to HSV space. Then the statistical features and texture features were extracted from region of inter- est on the elastogram. The important and reliable features were selected by using Minimum-Redundancy-Maxi- mum-Relevance ( mRMR) algorithm. Finally the selected features were input to the SVM classifier to classify the thyroid nodules into benign and malignant. The experiment results confirmed the method had higher accu- racy (92% ). It is helpful to improve the clinical accuracy by using CAD techniques