Poster No:
608
Submission Type:
Abstract Submission
Authors:
Madalina-Octavia Buciuman1, Paris Alexandros Lalousis2, Shalaila Haas3, Linda Antonucci4, Lana Kambeitz-Ilankovic5, Anne Ruef6, Stefan Borgwardt7, Joseph Kambeitz8, Christos Pantelis9, Rebecca Lencer10, Alessandro Bertolino11, Paolo Brambilla12, Rachel Upthegrove13, Stephen J. Wood14, Peter Falkai15, Anita Riecher-Rössler16, Stephan Ruhrmann5, Frauke Schultze-Lutter17, Eva Meisenzahl18, Jarmo Hietala19, Raimo K. Salokangas19, Dominic Dwyer20, Nikolaos Koutsouleris6
Institutions:
1Department of Psychiatry and Psychotherapy, LMU University Hospital, LMU Munich, Munich, BAYERN, 2Institute of Psychiatry, Psychology & Neuroscience, King's College London, London, United Kingdom, 3Icahn School of Medicine at Mount Sinai, New York, NY, 4University of Bari Aldo Moro, Milan, Italy, 5University of Cologne, Cologne, Germany, 6Department of Psychiatry and Psychotherapy, LMU University Hospital, LMU Munich, Munich, Bavaria, 7Department of Psychiatry and Psychotherapy, University of Lübeck, Lübeck, Germany, 8Department of Psychiatry and Psychotherapy, University of Cologne, Cologne, Bavaria, 9Melbourne Neuropsychiatry Centre, Carlton, Victoria, 10Institute for Translational Psychiatry, University Münster, Münster, Germany, 11Department of Basic Medical Science, Neuroscience and Sense Organs, University of Bari Aldo Moro, Bari, Bari, 12Department of Neurosciences and Mental Health, Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policli, Milan, Italy, 13Institute of Mental Health, University of Birmingham,, Birmingham, United Kingdom, 14Centre for Youth Mental Health, University of Melbourne, Melbourne, Australia, 15Department of Psychiatry and Psychotherapy, LMU University Hospital, LMU Munich, Munich, Germany, 16Faculty of Medicine, University of Basel, Basel, Basel, 17Heinrich-Heine University Düsseldorf, Düsseldorf, Germany, 18Department of Psychiatry and Psychotherapy, Medical Faculty, Heinrich Heine University, Düsseldorf, Germany, 19Department of Psychiatry, University of Turku, Turku, Finland, 20Orygen, the National Centre of Excellence for Youth Mental Health, Melbourne, Victoria
First Author:
Co-Author(s):
Anne Ruef
Department of Psychiatry and Psychotherapy, LMU University Hospital, LMU Munich
Munich, Bavaria
Stefan Borgwardt
Department of Psychiatry and Psychotherapy, University of Lübeck
Lübeck, Germany
Joseph Kambeitz
Department of Psychiatry and Psychotherapy, University of Cologne
Cologne, Bavaria
Rebecca Lencer
Institute for Translational Psychiatry, University Münster
Münster, Germany
Alessandro Bertolino
Department of Basic Medical Science, Neuroscience and Sense Organs, University of Bari Aldo Moro
Bari, Bari
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
Peter Falkai
Department of Psychiatry and Psychotherapy, LMU University Hospital, LMU Munich
Munich, Germany
Eva Meisenzahl
Department of Psychiatry and Psychotherapy, Medical Faculty, Heinrich Heine University
Düsseldorf, Germany
Jarmo Hietala
Department of Psychiatry, University of Turku
Turku, Finland
Dominic Dwyer
Orygen, the National Centre of Excellence for Youth Mental Health
Melbourne, Victoria
Nikolaos Koutsouleris
Department of Psychiatry and Psychotherapy, LMU University Hospital, LMU Munich
Munich, Bavaria
Introduction:
Impairments in daily functioning are prevalent in the early stages of psychosis and depression and are of increasing interest as clinical outcome metrics, due to their high impact on patients' quality of life. However, individualized risk stratification of functional outcomes is currently limited and would benefit from the exploration of novel methods and data modalities. In this context, group-level brain functional abnormalities have been extensively reported in psychotic and depressive disorders, but their value as early prognostic tools within the framework of precision psychiatry has been insufficiently explored. In the current work, we aimed to assess the value of the resting-state fractional amplitude of low frequency fluctuations (fALFF) as a longitudinal risk stratification tool of poor functioning in clinical high-risk for psychosis (CHR) and recent-onset depression (ROD).
Methods:
A total number of 218 clinical high-risk (CHR) individuals and 197 recent-onset depression (ROD) patients coming from eight European sites within the multi-center European Personalised Prognostic Tools for Early Psychosis Management (PRONIA) study were used for the current analyses.
Functional outcomes were measured using the Global Assessment of Functioning: Disability scale (Pedersen et al., 2007), with high functioning defined as more than 80 points on this scale.
All machine learning analyses were conducted using an open source MATLAB-ased software (NeuroMiner 1.2, https://github.com/neurominer-git/NeuroMiner_1.2). We trained support vector machine classifiers within a leave-site-out nested cross-validated framework to predict functional outcomes at the 9-month follow-up, 18-month follow-up, as well as the persistence of high functioning from the 9-month to the 18-month follow-up based on multiband functional amplitude of low frequency fluctuations (fALFF). The fALFF data was computed for three different frequency sub-bands, in line with literature showing distinct sources and potential functional information captured at different frequencies (Yu et al., 2014; Wang et al., 2016): slow-5 (0.01 - 0.027 Hz), slow-4 (0.027 - 0.073), and slow-3 (0.073 – 0.0198 Hz). The statistical significance of the models' performance relative to chance level was assessed using 1000 permutation of the labels and the predictive features were visualized using sign-based consistency of feature weights (Gómez-Verdejo et al, 2019).
Furthermore, we evaluated differences in accuracy between transdiagnostic and diagnosis-specific models (CHR vs ROD), associations of predictive patterns with broader clinical characteristics.
Results:
Distributed fALFF increases/decreases were transdiagnostically predictive of functioning outcomes at the 9-month follow-up above chance level, with a maximal balanced accuracy obtained for the slow-5 frequency sub-band (Balanced accuracy = 63.4%, Sensitivity = 55.2%, Specificity = 71.6%, P<.001). For the 18-month follow-up time point, functioning outcomes were best predicted based on the activity in the slow-3 sub-band (Balanced accuracy = 63.8%, Sensitivity = 68.4%, Specificity = 59.2%). Lastly, the persistence of high functioning from the 9-month to the 18-month follow-up time point could be predicted by baseline fALFF data with higher accuracy than the individual time-points, based on both the slow-5 (Balanced accuracy = 67.8%, Sensitivity = 71.7%, Specificity = 64.0%) and slow-3 (Balanced accuracy = 69.7%, Sensitivity = 75.3%, Specificity = 64.0%) frequency sub-bands. Predictive patterns were predominantly spanned regions of the default-mode, frontoparietal and salience networks.
Conclusions:
We provide first evidence for the relevance of rs-fMRI data as a transdiagnostic longitudinal biomarker of functioning outcomes in early psychotic and affective stages. These results could facilitate the development of multimodal risk stratification tools that incorporate functional brain biomarkers in the patient's prognostic assessment.
Disorders of the Nervous System:
Psychiatric (eg. Depression, Anxiety, Schizophrenia) 1
Modeling and Analysis Methods:
Classification and Predictive Modeling 2
Keywords:
FUNCTIONAL MRI
Machine Learning
Other - clinical high-risk for psychosis, recent-onset depression, longitudinal functioning outcomes
1|2Indicates the priority used for review
Provide references using author date format
Pedersen, G., Hagtvet, K. A. & Karterud, S. Generalizability studies of the Global Assessment of Functioning–Split version. Compr. Psychiatry 48, 88–94 (2007).
Yu, R. et al. Frequency-specific alternations in the amplitude of low-frequency fluctuations in schizophrenia. Hum. Brain Mapp. 35, 627–637 (2014).
Wang, L. et al. Frequency-dependent changes in amplitude of low-frequency oscillations in depression: A resting-state fMRI study. Neurosci. Lett. 614, 105–111 (2016).
Gómez-Verdejo, V., Parrado-Hernández, E., & Tohka, J. (2019). Sign-consistency based variable importance for machine learning in brain imaging. Neuroinformatics, 17(4), 593-609.