A Taxonomy and Comprehensive Survey of Brain Graph based Approaches for Schizophrenia Detection with Graph Neural Networks

Schizophrenia is a long term complex neuropsychiatric condition characterized by abnormal functional interactions between the parts of the brain. Brain can be modelled as a graph structure with brain regions as nodes and functional connections as the edges. Since the abnormalities causing Schizophrenia are not restricted to a single brain region, many latest developments in neuroimaging and graph learning for classification and detection of schizophrenia from brain graph structure connectivity patterns focused on applying Graph Neural Networks (GNNs). This work is a comprehensive review of developments in schizophrenia detection using brain graphs and GNN architectures including GCNs, GATs and hybrid models. The authors have systematically analysed recent studies published between 2019 to 2026, comparing the performance and methodological trends with focus on the data modalities such as structural MRI, functional MRI and resting state functional MRI. The key challenges for accurate prediction of schizophrenia include dataset heterogeneity, domain shift, interpretability limitations and external validation barriers. The analysis reveals that GNN based approaches to intelligent schizophrenia detection using resting state functional connectivity MRI are promising.