Longitudinal Multimodal Analysis of Structural Brain Changes and Psychopathology Using Data-Driven Fusion Approaches

Understanding the longitudinal trajectory of brain structure and its relationship with psychopathological measures is essential for understanding mechanisms in neurodevelopmental and affective disorders. In this study, we analyze multimodal neuroimaging and behavioral data to identify joint patterns of structural and psychological changes over time. Using the data-driven fusion approach—multiset canonical correlation analysis (mCCA) and joint independent component analysis (jICA)–we aim to uncover meaningful associations between cortical features and symptom measures.

We utilize longitudinal data from the FOR2107 cohort, including baseline and first 2-year follow-up assessments. Our dataset consists of FreeSurfer-derived features (surface area, cortical thickness) and self-reported psychopathological measures (SCL questionnaire). To extract latent multimodal patterns, we apply a joint ICA-based fusion approach, integrating structural brain measures and psychopathological scores. We then compute subject-specific loading parameters and examine their longitudinal correlations across modalities.

Preliminary findings reveal distinct multimodal patterns that capture shared variance between structural brain changes and symptom severity over time. We identify components with significant cross-modal correlations, suggesting a robust link between cortical alterations and psychological measures. Further analyses of individual components provide insights into brain regions most associated with these longitudinal effects.

Our findings demonstrate the potential of multimodal ICA-based fusion analysis in capturing complex structural-psychopathological relationships. This approach may contribute to a better understanding of the neurobiological mechanisms underlying psychiatric symptoms. Future work will focus on expanding the analysis to additional follow-up data and exploring predictive modeling for clinical applications.

Publication History

Article published online:
30 April 2025