Glymphatic surrogate, choroid plexus remodeling, and exploratory brain network analysis in the post-acute phase of anti-NMDAR encephalitis: multimodal MRI evidence for cognitive associations and differential white matter mediation.

BACKGROUND: Anti-N-methyl-D-aspartate receptor (anti-NMDAR) encephalitis typically presents with acute neuroinflammation, and many patients often experience persistent cognitive impairment during the post-acute/recovery phase. While impaired cerebrospinal fluid (CSF)-related markers-including glymphatic function and choroid plexus (CP) structural remodeling-are well documented in neurodegenerative diseases, their role in acute autoimmune brain injury remains unclear. OBJECTIVE: To evaluate a glymphatic surrogate (diffusion tensor image analysis along the perivascular space (DTI-ALPS) index), choroid plexus structural remodeling (CP/intracranial volume (ICV) ratio), white matter microstructural integrity (global fractional anisotropy (FA) using tract-based spatial statistics (TBSS)), and structural covariance network (SCN) topology in anti-NMDAR encephalitis using multimodal MRI, and to investigate their associations with cognition, with an exploratory mediation analysis examining whether white matter microstructure mediates these associations. METHODS: A total of 103 anti-NMDAR encephalitis patients and 97 healthy controls (HCs) with comparable sex distribution were enrolled. The glymphatic surrogate was evaluated by DTI-ALPS and CP structural remodeling was quantified by CP. White matter microstructural differences were analyzed using TBSS; mean FA values were extracted from all statistically significant clusters within the white matter skeleton (FA threshold = 0.2) after TFCE correction (P < 0.05), and the mean FA across these clusters was used as a global measure. Group-level SCNs were constructed based on 68 cortical regions, and graph theory metrics were compared. Partial correlation and bootstrap mediation analysis (5, 000 resamples) examined relationships among CSF markers, FA, and cognitive scores (MoCA, MMSE). A Monte Carlo sensitivity analysis was performed to determine the minimum indirect effect detectable at the achieved sample size (n = 103). RESULTS: Compared to HCs, patients with anti-NMDAR encephalitis showed significantly decreased bilateral DTI-ALPS indices (left: 1.361 ± 0.196 vs. 1.489 ± 0.178; right: 1.335 ± 0.183 vs. 1.483 ± 0.174; q = 0.0001), increased CP/ICV ratio (0.080 ± 0.026 vs. 0.071 ± 0.019, q = 0.0135), and reduced global FA (0.506 ± 0.052 vs. 0.597 ± 0.099, q< 0.0001). In covariate-adjusted analyses within the patient group (n=103), the CP/ICV ratio showed strong negative correlations with MoCA (partial r = -0.463, q< 0.001) and MMSE (partial r = -0.317, q = 0.0078), while the left DTI-ALPS index showed a significant positive correlation with MoCA (partial r = 0.310, q = 0.0078). Global FA was strongly correlated with the DTI-ALPS index (partial r = 0.394, q< 0.001). However, exploratory bootstrap mediation analysis revealed that global FA significantly mediated the DTI-ALPS-MoCA association (indirect effect = 0.075, 95% CI [0.011, 0.152]; proportion mediated = 70.57%), but did not mediate the CP/ICV-cognition associations (all 95% CIs included zero). Sensitivity analysis indicated adequate power to detect indirect effects ≥ 0.13; the observed effect (0.075) fell below this threshold, warranting caution in interpretation. Exploratory group-level structural covariance network analysis revealed topological changes in patients, including increased clustering coefficient and modularity and decreased global efficiency; however, these network metrics could not be linked to individual cognitive or CSF measures due to the group-level nature of network construction. CONCLUSIONS: Patients with anti-NMDAR encephalitis exhibit concurrent glymphatic function, CP structural remodeling, widespread white matter microstructural damage, and topological reorganization. The CP/ICV ratio emerged as the primary imaging correlate of cognitive dysfunction. Although exploratory mediation suggested that FA partially accounted for the ALPS-cognition association, this likely reflects shared methodological collinearity rather than a distinct biological pathway, reinforcing CP/ICV as the more robust clinically relevant marker.These findings extend the concept of glymphatic-choroid plexus pathology from neurodegenerative diseases to post-acute autoimmune brain injury, providing a framework for non-invasive biomarkers in anti-NMDAR encephalitis.