Research on automatic recognition of breast tumors based on Principal Component Analysis

Principal Component Analysis (PCA) can be used to describe breast tumor by a feature matrix. So this paper proposed a novel method to detect tumor based on PCA. Firstly, calculate the covariance matrix of breast image and obtain its eigenvectors. Next, select eigenvectors to structure feature space obeying cumulative contribution order. It is effective because the most information is contained in these eigenvectors which correspond to the front few biggest eigenvalues. Thirdly, map train samples and test samples to feature space to get the feature matrix. Lastly, recognize tumors using classifier based on distance combines with support vector machine (SVM). The proposed method is tested on 174 images and the detection sensitivity is 94.0%. Result shows that PCA is useful for tumor detection.