MRI-derived brain age as a biomarker of ageing in rats: validation using a healthy lifestyle intervention
MRI data can be used as input to machine learning models to accurately predict brain age in healthy human subjects. A large difference between predicted and chronological brain age (the so-called BrainAGE score) has been associated with disease and neurodegeneration, indicating the potential utility of neuroimaging-based ageing biomarkers. So far, most brain age prediction studies have been carried out on humans. However, it is important for such a biomarker to be validated on laboratory animals too, in order to better account for specific environmental or genetic factors within a more controlled laboratory framework. In this work, we developed a new algorithm for rat brain age prediction based on the combination of Gaussian process regression and a logistic regression classifier. The algorithm was trained on a cohort of 31 normal rats. High prediction accuracy was achieved using leave-one-out cross-validation (mean absolute error = 4.87 weeks, correlation between predicted and chronological age r = 0.92), supporting the validity and potential of the method. Furthermore, the trained model was tested on two independent groups of 24 rats each: a new normal control group and a "healthy lifestyle" group that underwent long-term environmental enrichment and dietary restriction (EEDR) between 3 and 17 months of age. After fitting a linear mixed-effects model, the BrainAGE values were found to increase more slowly with chronological age in the EEDR group than in the controls (slope = 0.52 vs. 0.61; p = 0.015 for the interaction term). When survival analysis was performed with a Cox regression model, the BrainAGE score at 5 months of age had a significant prediction power (p = 0.03). Our results demonstrate that BrainAGE, as computed by the proposed approach, is significantly modulated by EEDR intervention, hence it is a sensitive marker of biological ageing. These findings also support the potential of lifestyle-related prevention approaches to slow down the brain ageing process. Moreover, the results of the survival analysis further demonstrate that BrainAGE is indeed a predictor of ageing outcome.