Establishing a normative model for turbulence dynamics in the brain

Abstract The turbulence dynamics (TD) framework has proven to be informative of whole-brain network synchronicity and rigidity, and is related to neuropsychiatric disorders and antidepressant response prediction. However, studies conducted so far have used small samples and have not accounted for individual differences in demographic and other confounding variables. We aimed to investigate whether demographics were associated with TD read-outs and -if so- built a Normative Model (NM) to account for such demographic variability. We analysed resting-state fMRI data from the UK Biobank ( N = 11,184) and extracted TD features. We analysed associations between TD and age, sex and head motion during scanning across 11 spatial scales (λ:0.01–0.31). We then evaluated the benefit of correcting for these covariates using Normative Modeling (NM) by its association with recent depressive symptoms using the RDS-4 questionnaire. Age, sex and motion during scanning were statistically significantly associated with all TD features, though with generally small effect sizes (η_p 2 < 0.01). Increased network rigidity was seen in older age. While NM-correction did not increase predictive power for depressive symptoms—likely due to a weak phenotype—it successfully shifted the most predictive spatial scale of Amplitude Turbulence to an intermediate scale (λ = 0.16), aligning with established literature. TD features show small but consistent, statistically significant associations with demographics and motion, warranting correction. After applying NM-correction, no improvement in overall depression prediction occurred. However, the most predictive scale of Amplitude Turbulence shifted by applying NM, aligning with previous findings on TD for health and disease. The proposed NM value lies in recovering a more disease-specific, literature-consistent signal, rather than in improving prediction.