Embodied Dynamic Causal Modelling: Brain and Cardiac Effective Connectivity

Abstract

Brain-body interactions underpin embodied inference and interoception, and are central to both general health and the pathophysiology of mental health disorders. These interactions are orchestrated across a hierarchy extending from the cortex, through the brainstem, to peripheral autonomic effectors. Existing methods either reduce this communication to scalar summaries or lack an invertible biophysical model, and so cannot resolve where in this extended architecture an observed effect originates. To address this, we introduce Embodied Dynamic Causal Modelling, an extension of the generative causal modelling framework that formalises hierarchical message passing across the full brain-body axis. We instantiate the framework for brain-heart interactions with the Brainstem Cardiac Network, a closed-loop neural mass model of brainstem autonomic circuitry and sinoatrial node dynamics. By incorporating five brainstem nuclei, a viscerosensory thalamic relay and the cardiac pacemaker, the network enables estimation of directed effective connectivity across the cortical-brainstem-cardiac axis. As an illustrative case study, representative of the experiments and extended physiological systems that Embodied DCM and the BCN can now interrogate, we applied the framework to simultaneously recorded electroencephalography and electrocardiography during an auditory Go/No-Go paradigm. A multimodal observation model permits joint inversion of electrophysiological and cardiac responses, so that neural and bodily measurements together constrain an account of central and peripheral dynamics. The case study shows how the framework can arbitrate between competing hypotheses about functional cytoarchitecture, distinguishing granular from agranular microcircuits within the viscero-sensorimotor hierarchy. It also shows how time-varying estimates of synaptic efficacy can be recovered alongside continuous latent neuronal dynamics, over extended windows spanning anticipation, target processing and motor inhibition. By linking neural and cardiac signals to a single generative architecture, this work moves the study of brain-heart interactions beyond correlation, providing a framework for testing theories of embodied inference and for identifying circuit-level disruptions in disorders of interoceptive and autonomic regulation.