Effective connectivity markers of hyposmia in parkinson’s disease based on olfactory network interactions

Olfactory dysfunction is one of the earliest and most prevalent non-motor symptoms of Parkinson’s disease (PD), yet the directional network mechanisms linking hyposmia to disease progression remain poorly understood. In this study, we investigated stage-dependent alterations in effective connectivity within the core olfactory network using resting-state functional MRI and dynamic causal modeling (DCM). Effective connectivity among the olfactory bulb, orbitofrontal cortex, amygdala, and hippocampus was examined in healthy controls, PD patients with preserved or mild olfactory function, and PD patients with severe hyposmia. Bayesian model selection favored a fully connected network architecture, revealing a non-linear trajectory of network reorganization. Patients with preserved olfaction exhibited compensatory hyperconnectivity, characterized by increased directed coupling and network density relative to controls, whereas patients with severe hyposmia showed a marked collapse of effective connectivity, particularly along sensory–limbic and frontal integration pathways. Graph-theoretical analyses confirmed a progressive decline in network strength and efficiency across groups. Importantly, machine-learning classification using DCM-derived connectivity features achieved high accuracy in distinguishing control subjects and PD subgroups, demonstrating strong discriminative and translational potential. To our knowledge, this is the first study to apply DCM to explicitly characterize stage-specific, directional olfactory network alterations in Parkinson’s disease and to integrate these mechanistic markers with machine learning. These findings provide novel insight into compensatory versus degenerative network dynamics underlying hyposmia in PD and establish effective connectivity of the olfactory network as a promising candidate biomarker for disease stratification and early detection.