A quantum-inspired machine learning framework has identified stable brain regions that act as hubs in schizophrenia and bipolar disorder, with bipolar features forming a subset of schizophrenia's — a network-based signature of hierarchical brain changes.
The research
Researchers led by Domenico Pomarico at the University of Bari Aldo Moro and colleagues across institutions in Italy, Chile, and the UK applied tensor network machine learning — a method originally developed for quantum many-body physics — to structural MRI data. The team, including Alessandro Grecucci, Loredana Bellantuono, Jesus M. Cortes, Marianna La Rocca, Alfonso Monaco, Marlis Ontivero-Ortega, Alessandro Scarano, Massimo Stella, Roberto Bellotti, Sebastiano Stramaglia, and Nicola Amoroso, encoded gray-matter features into a Matrix Product State representation. After training, they extracted quantum connected correlations between features and mapped them onto a weighted graph.
Using repeated train-test sampling across two classification tasks — healthy controls versus schizophrenia, and versus bipolar disorder — they tracked both global network properties and node-level centrality measures. The analysis revealed a stable set of gray-matter features, most prominently Heschl gyrus, insular cortex, and frontal regions, that consistently acted as hubs across multiple centrality measures and resamplings. These centralities showed lower variability than Shapley values, suggesting the network representation offers more reliable interpretability. Notably, the bipolar feature set emerged as a subset of the schizophrenia one, aligning with the hierarchically organized neuroanatomical alterations reported in prior neuroimaging studies.
The paper, arXiv:2608.21368, was submitted on 19 June 2026 and is categorized under Neurons and Cognition (q-bio.NC).
Why it matters
This work bridges quantum physics, machine learning, and neuroscience to deliver more interpretable maps of brain disorders. For anyone interested in cognition, it underscores that complex psychiatric conditions are not random but involve specific, hierarchically organized brain networks. Understanding these hubs — like the Heschl gyrus (auditory processing) and insular cortex (interoception, emotion) — could lead to better diagnostics and targeted interventions. More broadly, it shows how advanced computational methods can reveal stable biological signatures that simpler models miss, potentially improving how we classify and study brain health.
What you can do
- Stay curious about how lifestyle factors (sleep, exercise, social engagement) influence brain network integrity; evidence links them to cognitive resilience.
- If you have concerns about your own cognitive health or mood, consult a qualified healthcare professional — this research is not a diagnostic tool.
- Engage in mentally stimulating activities that challenge memory, reasoning, and social cognition; they may help maintain gray-matter networks.
Source: arXiv q-bio.NC
Curious about your own brain? Take our free adaptive IQ test or try 306 brain training levels.