White matter brain age as a biomarker of cerebrovascular burden in the ageing brain

Abstract

As the brain ages, it almost invariably accumulates vascular pathology, which differentially affects the white matter. The microstructure of the white matter may therefore reveal a brain age reflecting cerebrovascular disease burden and a relationship to vascular risk factors. In this study, a white matter specific brain age was developed from diffusion weighted imaging (DWI) using a three-dimensional convolutional neural network (3D-CNN) deep learning model in both cross-sectional data from UK biobank participants (n = 37327) and a longitudinal subset (n = 1409) with an average of 2.25 years follow up. We achieved a mean absolute error (MAE) of white matter brain age prediction of 2.84 years and a Pearson’s r of 0.902 with chronological age in the test participants. The average white matter brain age gap (WMBAG) of the baseline 1409 participants with repeated scans were 0.36 ± 0.11 years younger than that of other participants in the baseline test sample with single time-point MRI scan (n = 9759). Individual vascular risk factors and the cumulative vascular risk score were significantly correlated with greater WMBAG Obesity was observed to be correlated with WMBAG only in the male participants. Participants with one, two, and three or more vascular risk factors, compared to those without any, showed an elevated WMBAG of 0.54, 1.23, and 1.94 years, respectively. Baseline WMBAG was also associated significantly with processing speed, executive and global cognition after Bonferroni correction. The significant associations of diabetes and hypertension with poor processing speed and executive function were found to be mediated through the WMBAG. However, the vascular risk factors did not associate with the two-year change in WMBAG in the longitudinal dataset. Our analysis suggests that tissue-specific brain age can be successfully targeted for the examination of the most relevant risk factors and cognition, although longer-term longitudinal data are needed to demonstrate its dynamic characteristics. The results suggest an intriguing possibility that a white matter brain age gap can serve as a potential neuroimaging biomarker for an individual’s cerebrovascular ageing process.

Association of Blood Pressure With Brain Ages: A Cohort Study of Gray and White Matter Aging Discrepancy in Mid-to-Older Adults From UK Biobank

“…The similar network architecture (same network design and same convolution filters) as in our previous WM brain age study was adopted. 15 In particular, the feature extractor consisted of 6 3D-convolutional layers falsetrueConv3D(32,-3px3,1,2)−Conv3D(64,-3px3,1,2)−Conv3D(128,-3px3,1,2) true−Conv3D(256,-3px3,-3px1,-3px2)−Conv3D(256,-3px3,-3px1,-3px2)−Conv3D(64,-3px3,-3px1,2) followed by a 2-layer multilayer perceptron with dimensions falsed−100−1, where falseConv3D(m, n, p, q) denoted a 3D convolutional layer with channel number falsem, kernel size …”

“…The similar network architecture (same network design and same convolution filters) as in our previous WM brain age study was adopted. 15 In particular, the feature extractor consisted of 6 3D-convolutional layers…”