Multimodal Clustering Analysis of meta-analytically derived brain regions in Schizophrenia
Poster No:
470
Submission Type:
Abstract Submission
Authors:
Maxim Korman1, Keith Smith2, Lukas Roell3, Daniel Keeser4
Institutions:
1LMU Hospital Munich, Munich, Germany, 2Computer and Information Sciences, University of Strathclyde, Glasgow, United Kingdom, 3LMU Hospital Munich / University of Melbourne, Munich /Melbourne, Bavaria, 4Department of Psychiatry and Psychotherapy, LMU University Hospital, Munich, Germany
First Author:
Co-Author(s):
Introduction:
Schizophrenia spectrum disorders (SSD) are severe and complex mental health conditions characterized by diverse clinical presentation, biological heterogeneity and a lack of reliable biomarkers. Although neuroimaging studies consistently reveal brain structural and functional alterations in SDD, interindividual variability complicates subgroup identification and the development of targeted treatments.
·Workflow.
Methods:
We developed a novel unsupervised clustering workflow that integrates T1w, DTI, fMRI and EEG data. Background and framework of this study were pre-registered (https://osf.io/k7wja/). First, we analyzed data from 146 SSD patients and 129 healthy controls (HC) to replicate alterations of gray matter volume (GMV) in regions of interest (ROIs) identified in meta-analyses. Subsequently, clustering analysis across multiple modalities was conducted using the complete data from 104 SSD patients. Grounded in meta-analysis, our approach employs advanced non linear dimensionality reduction techniques (PaCMAP, UMAP) followed by Gaussian mixture model (GMM) clustering and extensive validation of the robustness of the identified subgroups (e.g. bootstrapping, significance testing).
Results:
The replication of gray matter volume alterations was successful for 30/33 of all regions of interest in our cohort. About half of these regions showed significant alterations across all neuroimaging modalities in case-control comparisons, with the thalamus demonstrating the most substantial effects. A trending association with polygenic risk scores for thalamic volume was observed, highlighting a potential genetic contribution. Correlations with clinical variables differed significantly depending on the neuroimaging modality, pointing to the increased explanatory power provided by multimodality. Clustering analysis identified five distinct SSD subgroups, characterized by divergent brain-symptom-genetics profiles, including patterns of fractional anisotropy-cognition-general functioning (GAF) and GMV-resilience-polygenic risk score for Intracranial Volume (ICV).
·Multimodal Clusters, Mean Values for neuroimaging modalities, correlations between modalities and clinical scores
Conclusions:
Our multimodal clustering approach based on meta-analyses-derived brain regions identified clinically meaningful subgroups within our cohort with distinct brain-symptom-genetic profiles. These findings warrant further replication studies that may identify additional subgroups, enhancing our understanding of the complex heterogeneity underlying mental illness and SSD, and thus contribute to the development of more tailored therapeutic interventions.
Disorders of the Nervous System:
Psychiatric (eg. Depression, Anxiety, Schizophrenia) 1
Genetics:
Genetics Other
Modeling and Analysis Methods:
Multivariate Approaches
Novel Imaging Acquisition Methods:
Multi-Modal Imaging 2
Keywords:
Cognition
Electroencephaolography (EEG)
FUNCTIONAL MRI
Meta- Analysis
MRI
Multivariate
Psychiatric Disorders
Schizophrenia
Thalamus
Workflows
1|2Indicates the priority used for review