Fig. 4: Validation cohort analysis: brain injury biomarkers, structural MRI features, and perivascular diffusion metrics.

a, b Levels of glial fibrillary acidic protein (GFAP) and neurofilament light chain (NfL) in the validation cohort across NS-PASC (n = 44), Other-PASC (n = 45), and Cog-PASC (n = 16). c, d Cortical thickness (mm) across brain regions (c) and standardized beta coefficients (β) from regression models relating MoCA scores to thickness in the frontal, cingulate, insular, and parietal cortices NS-PASC (n = 44), Other-PASC (n = 45), and Cog-PASC (n = 16). d Effect sizes are presented as heat maps. Paramagnetic susceptibility (ppm) across cortical regions (e) and standardized β coefficients from regression models linking MoCA scores to iron deposition in the occipital cortex, caudate, putamen, and hippocampus ((NS-PASC (n = 25), Other-PASC (n = 39), and Cog-PASC (n = 14)) (f). Choroid plexus volume (CPV) fraction and DTI-ALPS index by group. DTI-ALPAS (NS-PASC (n = 46), Other-PASC (n = 46), and Cog-PASC (n = 16) (g), and standardized β coefficients from regression models relating these metrics to MoCA scores (h). Significant differences were observed between Cog-PASC and both NS-PASC and Other-PASC in (g) (***p < 0.001). For all box plots, the center line indicates the median, box boundaries the interquartile range (IQR; 25th–75th percentiles), and whiskers extend to the most extreme data points within 1.5 × IQR. Data points beyond whiskers are shown as outliers. Group comparisons were conducted using Two-sided ANCOVA, adjusting for age, sex, and education. Regression models were adjusted for the same covariates. All p-values were corrected for multiple comparisons using the false discovery rate (FDR); significance levels are indicated as *adjusted p < 0.05, **adjusted p < 0.01, and ***adjusted p < 0.001. Exact p-values are provided in the Source Data file.