Causal ordinal connections based characterization of weighted effective brain network for schizophrenia detection

The human brain is responsible for a wide range of a person's behavioral and cognitive capabilities. The functionality of the brain is affected by various disorders like schizophrenia, epilepsy, and Alzheimer. This study presents a novel framework for the automated detection of schizophrenia using EEG-based Weighted Effective Brain Connectivity Networks (WEBCNs). The proposed method introduces a new network descriptor called Weighted Directed Ordinal Connection (WDOC) that integrates causal directionality, connection strength, and ordinal relation between edges to capture complex brain dynamics. EEG signals from schizophrenia patients and healthy controls are preprocessed and transformed into WEBCNs using four causal connectivity estimators: Directed Transfer Function (DTF), Granger Causality (GC), Partial Directed Coherence (PDC), and Transfer Entropy (TE). WDOC-based features are extracted and classified using multiple machine learning algorithms, including KNN, SVM (linear, polynomial, RBF), and Random Forest. Among all models, the SVM with RBF kernel achieved the best performance, yielding 94.44% accuracy, 95% precision, 94% recall, and 89% kappa score for PDC-based networks. Structural and statistical analyses confirm distinct topological alterations in the causal flow between frontal and parietal regions in schizophrenia. The results demonstrate that WDOC-based characterization enhances discriminative power and interpretability in effective brain network analysis.