A Pyramidal Spatial-Based Feature Attention Network for Schizophrenia Detection Using Electroencephalography Signals

• Published in: IEEE transactions on cognitive and developmental systems Vol. 16; no. 3; pp. 935 - 946
• Main Authors: Karnati, Mohan, Sahu, Geet, Gupta, Abhishek, Seal, Ayan, Krejcar, Ondrej
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.06.2024; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Attention; Brain modeling; classification; Computational modeling; Convolutional neural networks; Electroencephalography; electroencephalography (EEG) signals; Feature extraction; Feature maps; Hypothesis testing; Industrial applications; Neurological diseases; Recording; Schizophrenia; schizophrenia (SZ) detection; Signal classification; Source code
• ISSN: 2379-8920; 2379-8939
• DOI: 10.1109/TCDS.2023.3314639

Automatic signal classification is utilized in various medical and industrial applications, particularly in schizophrenia (SZ) diagnosis, one of the most prevalent chronic neurological diseases. SZ is a significant mental illness that negatively affects a person's behavior by causing things like speech impairment and delusions. In this study, electroencephalography (EEG) signals, a noninvasive diagnostic technique, are being investigated to distinguish SZ patients from healthy people by proposing a pyramidal spatial-based feature attention network (PSFAN). The proposed PSFAN consists of dilated convolutions to extract multiscale deep features in a pyramidal fashion from 2-D images converted from 4-s EEG recordings. Then, each level of the pyramid includes a spatial attention block (SAB) to concentrate on the robust features that can identify SZ patients. Finally, all the SAB feature maps are concatenated and fed into dense layers, followed by a Softmax layer for classification purposes. The performance of the PSFAN is evaluated on two data sets using three experiments, namely, the subject-dependent, subject-independent, and cross-dataset. Moreover, statistical hypothesis testing is performed using Wilcoxon's rank-sum test to signify the model performance. Experimental results show that the PSFAN statistically defeats 11 contemporary methods, proving its effectiveness for medical industrial applications. Source code: https://github.com/KarnatiMOHAN/PSFAN-Schizophrenia-Identification-using-EEG-signals .