Alzheimer’s diseases diagnosis using fusion of high informative BiLSTM and CNN features of EEG signal

Electroencephalography (EEG) signals are low cost and available data for diagnosis of mental disorders such as Alzheimer's diseases (AD). Each EEG signal contains information about electrical brain activities during the time. In addition to temporal features, different locations of brain show regional information in each time instance. Both of the mentioned features are used for AD diagnosis in this work. Bidirectional long short-term memory (BiLSTM) networks are used for analyzing the time sequences while the convolutional neural network (CNN) is used for exploration of relationship among EEG signals recorded by different channels located in different parts of the brain. The temporal and regional features are then fused through a fully connected neural network. Moreover, channel selection using entropy measure and data augmentation using autoencoder networks are implemented in this work to improve the diagnosis accuracy. The proposed framework is assessed in different cases and compared with several state-of-the-art methods in various experiments done by different EEG recording channels. The experiments show that the proposed method can reach to 100% accuracy for AD diagnosis using the studied EEG database.