作者: Yudong Zhang , Yi Sun , Preetha Phillips , Ge Liu , Xingxing Zhou
DOI: 10.1007/S10916-016-0525-2
关键词:
摘要: This work aims at developing a novel pathological brain detection system (PBDS) to assist neuroradiologists interpret magnetic resonance (MR) images. We simplify this problem as recognizing brains from healthy brains. First, 12 fractional Fourier entropy (FRFE) features were extracted each image. Next, we submit those multi-layer perceptron (MLP) classifier. Two improvements proposed for MLP. One improvement is the pruning technique that determines optimal hidden neuron number. compared three techniques: dynamic (DP), Bayesian boundaries (BDB), and Kappa coefficient (KC). The other use adaptive real-coded biogeography-based optimization (ARCBBO) train biases weights of experiments showed FRFE?+?KC-MLP?+?ARCBBO achieved an average accuracy 99.53 % based on 10 repetitions K-fold cross validation, which was better than 11 recent PBDS methods.