Schizophrenia Biomarkers from fMRI Data Using Forman-Ricci Curvature on Hypergraphs

In this work, subject-level weighted 3-uniform hypergraph adjacency tensors are constructed from rs-fMRI BOLD time series using O-information as hyperedge weights, and Forman--Ricci curvature (FRC) is computed at hyperedge and node levels. Global mean curvature and node-wise profiles are compared between schizophrenia patients and healthy controls (HC) across 90 AAL regions via Mann--Whitney $U$ and Welch's $t$-tests with Benjamini--Hochberg correction. Results reveal significantly more negative curvature in healthy controls (\(p = 2.88 \times 10^{-5}\), \(d \approx 0.58\)) and widespread regional differences, supporting global and node-level FRC as sensitive and interpretable neuroimaging biomarkers for schizophrenia.