Efficient Schizophrenia Detection Through Transfer Learning for Robust Diagnostics

Schizophrenia, a severe mental illness characterized by emotional, behavioral, and cognitive abnormalities, presents challenges in accurate diagnosis and effective treatment. Patients often experience delusions, hallucinations, and difficulties in thinking and social interaction. The impact on one's reality perception and emotional control is substantial. While typically emerging in late adolescence or early adulthood, post-COVID-19 cases of schizophrenia have shown an increase, emphasizing the need for improved diagnostic tools and treatments. Current methods involve antipsychotic drugs, psychotherapy, and mental health specialist assistance. However, these approaches vary in quality and accuracy. Electroencephalogram (EEG) signals, measuring brain electrical activity, provide insights into schizophrenia but lack direct diagnostic capabilities. To address these challenges, a deep learning model is proposed, utilizing transfer learning with Inception V3, Xception, and DenseNet121 models. Recognizing the limitations of a singular model, the study explores employing multiple models to enhance performance in schizophrenia detection. The investigation acknowledges and addresses inherent drawbacks, aiming to improve the accuracy of identification through the application of deep learning methods on specific datasets. In the context of the increasing post-COVID-19 schizophrenia cases, the urgency for advanced and accurate diagnostic approaches becomes even more pronounced.