Latent brain-cognition dimension indicative of neuroinflammatory basis of Long COVID.

Abstract

Long COVID is characterized by persistent multisystemic symptoms with poorly defined pathophysiological substrate. The presented study investigated whether Long COVID is associated with alterations in brain microstructure and cognition by jointly modelling multimodal MRI and neuropsychological data within a shared framework. One hundred nine participants (77 Long COVID, 32 controls) without the history of severe acute COVID infection underwent quantitative MRI (T₁, T₂, T₁ρ, T₂ρ relaxometry, neurite orientation dispersion and density indices derived from diffusion-weighted imaging) and neuropsychological testing. The resulting brain-cognition axis provided a robust group separation across dimensionalities (permutation P ≈ 0.02), with 72.6% of variance attributable to cognition and 27.4% to MRI features. The predominant imaging features included higher isotropic free-water and reduced neurite density, with minimal relaxometry influence. Projection of discriminant weights back to anatomy revealed involvement over cortico-subcortical dorsal attention, salience and limbic circuits, consistent with patients' cognitive complaints. Furthermore, brain-cognition axis score correlated not only with the global Long COVID symptom burden but also with its respiratory, systemic, and neurological domains without clear predominance for any symptom group (partial R ² ≈ 0.10-0.18, q < 0.01). Hence, although the cognitive profile provided the dominant discriminative anchor in group separation, neuroimaging offered complementary information and pathophysiological background for the clinically apparent cognitive phenotype. The prominence of neuroinflammation-sensitive MRI protocols highlights probable immune and vascular contributions rather than overt neurodegeneration as the basis of Long COVID in individuals without a severe acute infection course.