Multi-Class Brain Images Classification Based on Reality-Preserving Fractional Fourier Transform and Adaboost

Abstract

With the development of computer technology, the diagnostic capability of the computer-aided diagnosis systems has improved. It has contributed to classify the brain images into health or other pathological categories automatically and accurately. In this paper, we proposed an improved method by introducing reality-preserving fractional Fourier transform (RPFRFT) and Adaboost to classify brain images into five different categories of health, cerebrovascular disease, neoplastic disease, degenerative disease and inflammatory disease. We used 190 T2-weighted images obtained by magnetic resonance imaging in the experiment. First, we employed RPFRFT to extract spectrum features from each magnetic resonance image. Second, we applied principal component analysis (PCA) to reduce feature dimensionality to only 86. Third, those reduced spectral features of different samples were combined and then were fed into Adaboost to train the classifier. The 10×10-fold cross validation obtained an accuracy of 98.6%. The result confirms the effectiveness of our proposed method.