A Pathological Brain Detection System Based on Radial Basis Function Neural Network

(Aim) It is beneficial to classify brain images as healthy or pathological automatically, since the information set of 3D brain is too large to interpret with manual methods. Among various 3D brain imaging techniques, magnetic resonance (MR) imaging is widely used in daily medical treatment, because it can help in the four ways of diagnosis, prognosis, pre-surgical, and postsurgical procedures. Although there are automatic detection methods, they suffer from low accuracy. (Method) Therefore, we proposed a novel approach, which employed 2D discrete wavelet transform (DWT), and calculated the entropy as features. Then, a radial basis function neural network (RBFNN) was trained to classify images as pathological or healthy. A 10 x 10-fold cross validation was conducted to prevent overfitting. (Result) The method achieved a sensitivity of 95.89%, a specificity of 92.78%, and an overall accuracy of 95.44% over 125 MR brain images. (Conclusion) The performance suggests the proposed classifier is robust and effective in comparison with recently state-of-the-art approaches.