Exploring the Predictive Value of Structural Covariance Networks for the Diagnosis of Schizophrenia

Clara Sophie Vetter^1,2,3Annika Bender¹Dominic Dwyer^1,4,5Maxime Montembeault⁶Anne Ruef¹Katharine Chisholm⁷Lana Kambeitz-Ilankovic⁸Linda A. Antonucci⁹Stephan Ruhrmann⁸Joseph Kambeitz⁸Marlene Rosen⁸Theresa Lichtenstein⁸Anita Riecher-Rössler¹⁰Rachel Upthegrove¹¹Raimo Kalevi Rikhard Salokangas¹²Jarmo Hietala¹²Christos Pantelis^13,14Rebekka Lencer^15,16Eva Meisenzahl¹⁷Stephen Wood^18,19Paolo Brambilla^20,21Stefan Borgwardt¹⁶Peter Falkai^1,22Alessandro Bertolino⁹Nikolaos Koutsouleris^1,22*

• ¹Department of Psychiatry and Psychotherapy, Ludwig-Maximilian-University, Munich, Germany, Munich, Germany
• ²Munich Center of Machine Learning (MCML), Munich, Germany
• ³Helmholtz Association - Munich School for Data Science (MUDS), Munich, Germany
• ⁴Centre for Youth Mental Health, Faculty of Medicine, Dentistry and Health Sciences, University of Melbourne, Parkville, Victoria, Australia
• ⁵The National Centre of Excellence for Youth Mental Health, Orygen, The University of Melbourne, Parkville, Victoria, Australia
• ⁶Douglas Hospital Research Centre, Department of Psychiatry, Faculty of Medicine and Health Sciences, McGill University, Montréal, Quebec, Canada
• ⁷School of Psychology, University of Sussex, Sussex, United Kingdom
• ⁸Clinic for Psychiatry and Psychotherapy, University Hospital of Cologne, Cologne, North Rhine-Westphalia, Germany
• ⁹Department of Translational Biomedicine and Neuroscience, University of Bari Aldo Moro, Bari, Italy
• ¹⁰Faculty of Medicine, University of Basel, Basel, Basel-Stadt, Switzerland
• ¹¹Department of Psychiatry, Medical Sciences Division, University of Oxford, Oxford, England, United Kingdom
• ¹²Department of Psychiatry, Faculty of Medicine, University of Turku, Turku, Finland
• ¹³Melbourne Neuropsychiatry Centre, Department of Psychiatry, Melbourne Medical School, Faculty of Medicine, Dentistry and Health Sciences, University of Melbourne, Carlton, Victoria, Australia
• ¹⁴north western metal health, Royal Melbourne Hospital, Parkville, Melbourne, Australia
• ¹⁵Institute for Translational Psychiatry, University of Muenster, Muenster, Germany
• ¹⁶Translational Psychiatry Unit, Department of Psychiatry and Psychotherapy, University of Lübeck, Lübeck, Germany
• ¹⁷Department of Psychiatry and Psychotherapy, University Hospital Düsseldorf, Heinrich Heine University Düsseldorf, Düsseldorf, North Rhine-Westphalia, Germany
• ¹⁸Orygen Youth Health, Parkville, Victoria, Australia
• ¹⁹Institute for Mental Health, School of Psychology, University of Birmingham, Birmingham, England, United Kingdom
• ²⁰Department of Neurosciences and Mental Health, Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico, University of Milan, Milan, Italy
• ²¹Department of Pathophysiology and Transplantation, Faculty of Medicine and Surgery, University of Milan, Milan, Lombardy, Italy
• ²²Max Planck Institute for Psychiatry, Munich, Bavaria, Germany

Introduction: Schizophrenia is a psychiatric disorder hypothesized to result from disturbed brain connectivity. Structural covariance networks (SCN) describe the shared variation in morphological properties emerging from coordinated neurodevelopmental processes, This study evaluates the potential of SCNs as diagnostic biomarker for schizophrenia. Methods: We compared the diagnostic value of two SCN computation methods derived from regional gray matter volume (GMV) in 154 patients with a diagnosis of first episode psychosis or recurrent schizophrenia (PAT) and 366 healthy control individuals (HC). The first method (REF-SCN) quantifies the contribution of an individual to a normative reference group's SCN, and the second approach (KLS-SCN) uses a symmetric version of Kulback-Leibler divergence. Their diagnostic value compared to regional GMV was assessed in a stepwise analysis using a series of linear support vector machines within a nested cross-validation framework and stacked generalization, all models were externally validated in an independent sample (NPAT=71, NHC=74), SCN feature importance was assessed, and the derived risk scores were analyzed for differential relationships with clinical variables.Results: We found that models trained on SCNs were able to classify patients with schizophrenia and combining SCNs and regional GMV in a stacked model improved training (balanced accuracy (BAC)=69.96%) and external validation performance (BAC=67.10%). Among all unimodal models, the highest discovery sample performance was achieved by a model trained on REF-SCN (balanced accuracy (BAC=67.03%). All model decisions were driven by widespread structural covariance alterations involving the somatomotor, default mode, control, visual, and the ventral attention networks. Risk estimates derived from KLS-SCNs and regional GMV, but not REF-SCNs, could be predicted from clinical variables, especially driven by body mass index (BMI) and affect-related negative symptoms. Discussion: These patterns of results show that different SCN computation approaches capture different aspects of the disease. While REF-SCNs contain valuable information for discriminating schizophrenia from healthy control individuals, KLS-SCNs may capture more nuanced symptom-level characteristics similar to those captured by PCA of regional GMV.