Title:	Brain network dynamics across psychiatric conditions
DNr:	Berzelius-2026-204
Project Type:	LiU Berzelius
Principal Investigator:	Håkan Olausson <hakan.olausson@liu.se>
Affiliation:	Linköpings universitet
Duration:	2026-06-26 – 2027-01-01
Classification:	30215
Keywords:

Abstract

The proposed project investigates neural mechanisms of self–other distinction using functional MRI data acquired with a validated self-touch versus other-touch paradigm across multiple clinical populations. The ability to distinguish between self-generated and externally generated sensory signals is a fundamental component of bodily self-awareness and social cognition, and is known to be altered in several psychiatric conditions. We will analyze an existing dataset collected at the Center for Social and Affective Neuroscience (CSAN), comprising multiple cohorts measured with closely comparable paradigms, including task-based and resting-state fMRI. These cohorts include adults with ADHD (n = 28, controls n = 30), anorexia nervosa (e.g. n = 40, controls n = 40), and psychosis-spectrum disorders (n = 30, controls n = 30), and autism spectrum condition (n = 29, controls n = 30). Previous analyses of these datasets have demonstrated systematic but distinct alterations in self–other distinction across disorders. For example, individuals with ADHD show an enhanced differentiation between self- and other-touch at the neural level, suggesting a sharper bodily self-boundary. In contrast, individuals with psychosis show reduced differentiation between self- and externally generated stimuli, including decreased sensory attenuation and impaired predictive processing already at early processing stages. In anorexia nervosa, altered processing of self-generated and social touch has been linked to disturbances in body perception and prediction errors, often reflected in increased neural responses to touch. The central objective of the project is to determine whether the disorder-specific alterations identified in previous studies correspond to distinct diagnostic profiles or instead reflect continuous, transdiagnostic dimensions of altered bodily self-processing. To this end, we will use NeuroSTORM, a recently developed fMRI foundation model (Wang et al., 2026), to derive high-dimensional, non-linear representations of brain activity from resting-state and self–other-touch task fMRI. These representations will allow us to test whether brain activity patterns are organised according to diagnostic categories, shared latent dimensions, or context-dependent differences between rest and active self–other processing. The analysis will combine representation learning with clustering, dimensionality reduction, and classification approaches to identify both shared and disorder-specific neural signatures. This will provide a unified computational framework for comparing conditions that show qualitatively different patterns in classical analyses (e.g. increased vs. reduced self–other distinction). The expected outcome is a more precise characterization of how psychiatric populations differ in fundamental aspects of bodily self-processing. The project will contribute to ongoing efforts to move beyond categorical diagnostic frameworks towards dimension-based models of psychiatric disorders, grounded in neurobiological data. By leveraging large-scale representation learning, it aims to uncover structure in neural data that is not accessible using conventional linear analysis methods.