Impact of Connectivity Granularity: A Comparison of ROI and Network-Level Approaches for Early Schizophrenia Classification

While schizophrenia diagnosis relies on clinical interviews, there is growing interest in neuroimaging-based computational tools to aid classification. In particular, resting-state fMRI-derived functional connectivity has been explored as a potential biomarker, with applications not only in supporting clinical assessment but also in research contexts such as patient stratification and probing disease mechanisms. Here, we compare two common approaches to computing functional connectivity - region of interest (ROI)-level and brain network-level - and evaluate their predictive power for classifying first-episode schizophrenia patients, in contrast to most prior work focusing on chronic patients. We show that ROI-level features consistently outperform network-level features. Despite the simplicity of our classification models, we achieved accuracies up to 83.15% using the AAL90 atlas. We also found that non-lagged functional connectivity generally outperforms lagged variants, suggesting that added temporal complexity may introduce noise rather than improve predictive power. Overall, our findings highlight region-based connectivity from a medium-resolution atlas as a promising representation for early-stage schizophrenia classification, while emphasising the need for validation on independent datasets to confirm generalisability.