Detection of Schizophrenia using Machine Learning

Schizophrenia is a mental illness that impacts approximately 1% of the global population. The condition can make it challenging for individuals with schizophrenia to differentiate between internally generated and externally generated stimuli due to abnormalities in the corollary discharge process of the nervous system. This phenomenon provides an explanation for certain symptoms associated with schizophrenia. Exploring this process and its connection to the disease's symptoms can contribute to our understanding of the abnormal brain processes observed in affected individuals. Advancements in the study of electrophysiological brain activity, along with improved access to EEG data, offer new possibilities for investigating and diagnosing neuronal disorders in real-world settings. In this study, a larger dataset of EEG data was utilized, including data from previous research published in Schizophrenia Bulletin Advance Access on July 10, 2013. The dataset combined EEG records from 22 healthy individuals and 36 individuals with schizophrenia, as well as additional data from 10 healthy individuals and 13 schizophrenia patients from a prior study. The classification phase involved considering two different methods for diagnosing schizophrenia using EEG signals. Traditional machine learning techniques, such as xgboost, decision trees, naïve Bayes, random forest, Long short-term memories (LSTMs), support vector machines, two-dimensional convolutional networks (2D-CNNs), and two-dimensional convolutional networks-LSTMs, were employed to classify the EEG signals. Deep learning models were also implemented, and various activation functions were compared. Among the proposed deep learning models, the SVM architecture demonstrated the highest performance level. The SVM model utilized the RBF Kernel with cross-validation values (CV) of 3, 6, and 8. Remarkably, the SVM model achieved a precision rate of 100 percent.