Abstract
Schizophrenia is a serious mental illness that causes great harm to patients, so timely and accurate detection is essential. This study aimed to identify a better feature to represent electroencephalography (EEG) signals and improve the classification accuracy of patients with schizophrenia and healthy controls by using EEG signals. Our research method involves two steps. First, the EEG time series is preprocessed, and the extracted time-domain and frequency-domain features are transformed into a sequence of red–green–blue (RGB) images that carry spatial information. Second, we construct hybrid deep neural networks (DNNs) that combine convolution neural networks and long short-term memory to address RGB images to classify schizophrenic patients and healthy controls. The results show that the fuzzy entropy (FuzzyEn) feature is more significant than the fast Fourier transform (FFT) feature in brain topography. The deep learning (DL) method that we propose achieves an average accuracy of 99.22% with FuzzyEn and an average accuracy of 96.34% with FFT. These results show that the best effect is to extract fuzzy features as input features from EEG time series and then use a hybrid DNN for classification. Compared with the most advanced methods in this field, significant improvements have been achieved.
Highlights
Schizophrenia is a serious mental illness that causes great harm to patients, so timely and accurate detection is essential
We attempted to compare which feature of fuzzy entropy (FuzzyEn) and fast Fourier transform (FFT) is better and improve the accuracy of classification of schizophrenia in EEG signals
We use a hybrid deep neural networks (DNNs) that consists of convolutional neural network (CNN) and Long-short-term memory (LSTM) components to address the RGB images and differentiate schizophrenic patients and healthy controls
Summary
Schizophrenia is a serious mental illness that causes great harm to patients, so timely and accurate detection is essential. The deep learning (DL) method that we propose achieves an average accuracy of 99.22% with FuzzyEn and an average accuracy of 96.34% with FFT These results show that the best effect is to extract fuzzy features as input features from EEG time series and use a hybrid DNN for classification. Some feature extraction methods (such as those employed for time-domain features and frequency-domain features7) have been proposed to quantify EEG signals for studying state changes in the brain. In a relatively new development, DL algorithms have been extensively applied in medical image and signal processing and have shown high research potential In most cases, their performance exceeds traditional machine learning techniques. Long-short-term memory (LSTM) networks are composed of recurrent networks that include memory to model temporal dependencies in time series problems This approach gives us a way to structure our research
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
