Automatic Diagnosis of Schizophrenia in EEG Signals Using CNN-LSTM Models

• DOI: 10.3389/fninf.2021.777977
• Publication Date: 2021-12-01T02:46:37Z
• Authors: Afshin Shoeibi; Delaram Sadeghi; Parisa Moridian; Navid Ghassemi; Jónathan Heras; Roohallah Alizade...; Ali Khadem; Yinan Kong; Saeid Nahavandi; Yu-Dong Zhang; Juan Manuel Gorriz
• Keywords: Normalization; F1 score

Schizophrenia (SZ) is a mental disorder whereby due to the secretion of specific chemicals in brain, function some brain regions out balance, leading lack coordination between thoughts, actions, and emotions. This study provides various intelligent deep learning (DL)-based methods for automated SZ diagnosis via electroencephalography (EEG) signals. The obtained results are compared with those conventional methods. To implement proposed methods, dataset Institute Psychiatry Neurology Warsaw, Poland, has been used. First, EEG signals were divided into 25 s time frames then normalized by z -score or norm L2. In classification step, two different approaches considered this was first carried machine e.g., support vector machine, k -nearest neighbors, decision tree, naïve Bayes, random forest, extremely randomized trees, bagging. Various DL models, namely, long short-term memories (LSTMs), one-dimensional convolutional networks (1D-CNNs), 1D-CNN-LSTMs, used following. models implemented activation functions. Among CNN-LSTM architecture had best performance. architecture, ReLU L2-combined normalization model achieved an accuracy percentage 99.25%, better than most former studies field. It worth mentioning that perform all simulations, -fold cross-validation method = 5