Detection of Schizophrenia Using Attention Mechanism with Convolutional Neural Network on EEG Signal

Schizophrenia (SZ) is a challenging mental disorder, requiring accurate diagnosis for effective treatment. Early diagnosis of schizophrenia can facilitate a normal life for patients. Electroencephalogram (EEG) data contains valuable insights into brain network connectivity, enabling the detection of abnormalities associated with schizophrenia. Hence, this study proposes a methodology employing a deep learning-based model utilizing deep convolutional neural networks (CNNs) with an Attention mechanism to classify schizophrenia patients. In this research, each EEG signal undergoes decomposition with Continuous Wavelet Transform (CWT) to extract relevant features that capture the underlying neurophysiological characteristics of brain activity. These features are then utilized to distinguish schizophrenia patients from healthy individuals using the proposed method, which is evaluated on an EEG dataset comprising 45 schizophrenic patients and 39 healthy subjects. To provide a comparison, various deep learning models like ResNet-50, CNN and DenseNet Model are applied. The results indicate that the proposed model outperforms all other methods tested.