Schizophrenia Detection In Eeg Applying Artificial Intelligence

Abstract

Schizophrenia, bipolar disorder, and depression are psychiatric illness that causes negative effects on individuals as well as society. Due to the significant overlap in the clinical symptoms of depression (DD), bipolar disorder (BD), and schizophrenia (SZ), using conventional guidelines is hard to discriminate between them. A proper differentiation between depression, bipolar disorder, and schizophrenia patients is critical to lessen their effects. Therefore, an artificial intelligence approach based on electroencephalography (EEG) data is suggested as an automated clinical system to assist psychiatrists in the diagnosis of mental conditions. This system consists of three stages. Firstly, the data is up-sampled by SMOTE to produce balanced data. Then, feature extraction and selection are used to reduce the number of irrelevant features. The final stage is a classification complied with machine learning and deep learning. The recommended models include K-Nearest Neighbor, Logistic Regression, Gradient Boosting, Random Forest, Neural network, VGG19, and Efficient Net B0. The proposed method achieves the highest accuracy of 97.50%, in a data from 199 people with major depressive disorder, 117 participants with schizophrenia, and 67 subjects with bipolar depression. The proposed methodology has the potential to be a valuable supplemental tool for medical professionals with future improvement.