Hybrid transforms domain-based mammogram analysis using C-SVM classifier
B Prathibha, V.R. Sadasivam · International Journal of Medical Engineering and Informatics · 2012
Early detection is considered as the only best way to avoid breast cancers and mammography is the most effective method for early detection and screening of breast cancers. Digital mammograms with the help of computer aided diagnostic systems help radiologists in taking fair decisions for low mortality rate. The proposed system combines wavelet features and spectral features for better analysis of mammograms. The support vector machines classifier is used to analyse 214 mammogram images from MIAS database pertaining to the severity of abnormality, i.e., benign and malign. The proposed system gives 96.26% accuracy for discrimination between normal-malign samples, 92.52% for normal-benign samples and 90.65% accuracy for benign-malign samples. The novelty of the proposed method is compared with classifiers kNN, Bayesian and kernel density estimation. The study reveals that features extracted in hybrid transform domain with SVM classifier proves to be a promising tool for analysis of mammograms.