Machine Learning Study on Dissociative Symptoms and the Relationship with Structural Brain Regions

463 

Abstract Submission 

John Fanning¹, Caroline Plett¹, Anne Ruef¹, Joseph Kambeitz², Raimo K. Salokangas³, Jarmo Hietala³, Alessandro Bertolino⁴, Stefan Borgwardt⁵, Paolo Brambilla⁶, Rachel Upthegrove⁷, Stephen J. Wood⁸, Rebecca Lencer⁹, Eva Meisenzahl¹⁰, Peter Falkai¹¹, Lisa Hahn¹, Nikolaos Koutsouleris¹

¹Department of Psychiatry and Psychotherapy, LMU University Hospital, LMU Munich, Munich, Bavaria, ²Department of Psychiatry and Psychotherapy, University of Cologne, Cologne, Bavaria, ³Department of Psychiatry, University of Turku, Turku, Finland, ⁴Department of Basic Medical Science, Neuroscience and Sense Organs, University of Bari Aldo Moro, Bari, Bari, ⁵Department of Psychiatry and Psychotherapy, University of Lübeck, Lübeck, Germany, ⁶Department of Neurosciences and Mental Health, Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policli, Milan, Italy, ⁷Institute of Mental Health, University of Birmingham,, Birmingham, United Kingdom, ⁸Centre for Youth Mental Health, University of Melbourne, Melbourne, Australia, ⁹Institute for Translational Psychiatry, University Münster, Münster, Germany, ¹⁰Department of Psychiatry and Psychotherapy, Medical Faculty, Heinrich Heine University, Düsseldorf, Germany, ¹¹Department of Psychiatry and Psychotherapy, LMU University Hospital, LMU Munich, Munich, Germany

John Fanning  
Department of Psychiatry and Psychotherapy, LMU University Hospital, LMU Munich
Munich, Bavaria

Caroline Plett  
Department of Psychiatry and Psychotherapy, LMU University Hospital, LMU Munich
Munich, Bavaria

Anne Ruef  
Department of Psychiatry and Psychotherapy, LMU University Hospital, LMU Munich
Munich, Bavaria

Joseph Kambeitz  
Department of Psychiatry and Psychotherapy, University of Cologne
Cologne, Bavaria

Raimo K. Salokangas  
Department of Psychiatry, University of Turku
Turku, Finland

Jarmo Hietala  
Department of Psychiatry, University of Turku
Turku, Finland

Alessandro Bertolino  
Department of Basic Medical Science, Neuroscience and Sense Organs, University of Bari Aldo Moro
Bari, Bari

Stefan Borgwardt  
Department of Psychiatry and Psychotherapy, University of Lübeck
Lübeck, Germany

Paolo Brambilla  
Department of Neurosciences and Mental Health, Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policli
Milan, Italy

Rachel Upthegrove  
Institute of Mental Health, University of Birmingham,
Birmingham, United Kingdom

Stephen J. Wood  
Centre for Youth Mental Health, University of Melbourne
Melbourne, Australia

Rebecca Lencer  
Institute for Translational Psychiatry, University Münster
Münster, Germany

Eva Meisenzahl  
Department of Psychiatry and Psychotherapy, Medical Faculty, Heinrich Heine University
Düsseldorf, Germany

Peter Falkai  
Department of Psychiatry and Psychotherapy, LMU University Hospital, LMU Munich
Munich, Germany

Lisa Hahn  
Department of Psychiatry and Psychotherapy, LMU University Hospital, LMU Munich
Munich, Bavaria

Nikolaos Koutsouleris  
Department of Psychiatry and Psychotherapy, LMU University Hospital, LMU Munich
Munich, Bavaria

Depersonalization and derealization are highly distressing symptoms that are not only confined to trauma-related diagnoses, such as post-traumatic stress disorder, but also frequently emerge in psychotic and affective conditions. Structural brain imaging studies have identified associations between regional brain volumes and these dissociative symptoms; however, there is a lack of evidence whether the presence of these symptoms in early-stage affective and psychotic patients can be predicted at the individual level using brain imaging data. In this study we attempt to predict the presence of the two dissociative symptoms using structural imaging as well as explore how closely related the measured brain patterns relate to childhood trauma and/or life events.

The current study employed a machine learning-based approach to predict the presence of derealization (n = 605) and depersonalization (n = 602) using structural brain imaging of patients collected from the multisite PRONIA study cohort. The sample consisted of healthy controls and patients with recent-onset psychosis, recent-onset depression, and clinical high-risk of developing psychosis. We created support vector machines with a 10-fold nested cross validation scheme to facilitate the prediction of our sample for each of the symptoms seperately. Additionally, the decision scores derived from these models were correlated with the Childhood Trauma Questionnaire and the Cologne Chart of Life Events.

Two stratifications of the sample were created to independently analyze derealization and depersonalization symptom presence as assessed by SPI-A O8 and SPI-A F6 scores, respectively. Balanced accuracy in the support vector machine models using structural imaging for derealization symptoms was 58.7% (sensitivity = 80.0%, specificity = 37.6%), and 54.5% for depersonalization symptoms (sensitivity: 66.4%, specificity: 42.5%). Furthermore, the decisions scores from the models showed no significant relationships between either of the trauma-based questionnaires.

The machine learning analyses identified a signal using structural brain imaging to identify dissociative symptoms in a trans-diagnostic sample of early-stage affective and psychotic patients. However, both models showed poor performance predicting the presence of either symptom, which suggests that the SPI-A may be an ineffective tool to fully assess these symptoms. This notion is supported from the correlational analyses, in which no significant relationships were found. Future investigations could use different questionnaires like the Dissociative Experiences Scale to find a more robust measure for the prediction of these symptoms.

Psychiatric (eg. Depression, Anxiety, Schizophrenia) ¹

Classification and Predictive Modeling ²

Affective Disorders

Modeling

Psychiatric Disorders

Schizophrenia

STRUCTURAL MRI

Other - Depersonalization; Derealization