Fold opening dynamics during the aging process highlight new sulcus communities

International audienceAging is associated with morphological changes in the brain, which have been studied by many investigators using features such as cortical thickness, surface area or gyrification index. It has been shown, for example, that with age, there is a reduction of grey and white matter volumes, while the volume of cerebrospinal fluid increases. However, these changes are not uniform in the cerebral hemisphere. Moreover, numerous neurodegenerative diseases can cause accelerated brain atrophy. The challenge is to disentangle normal brain atrophy due to healthy ageing from accelerated brain atrophy due to a disease. The idea here is to rekindle the discussion by considering a new fold-related atrophy feature: the sulcal width. Using a large-scale general population cohort (UK Biobank), we first observed an accelerated sulcal opening in most of the sulci beyond a peculiar age, which we suggest calling “tipping-point age” and which we estimated by fitting a gender-stratified piecewise linear model on the age-dependency of the sulcal openings and computing a hundred resamples to reduce overfit bias and to determine 95% confidence intervals. We found that as opposed to men, women tend to display bigger slope changes in atrophy, hinting at the fact that men are subject to accelerated atrophy much earlier than women probably due to more health risk factors. Going on with the data from women only, we discarded sulci which presented too much instability during bootstrapping for the intuitive reason that the variability of the tipping-point age reflects the prominence of the acceleration breakpoint. With the parameters of the piecewise linear modelling of the 61 remaining sulci, we performed a spectral clustering with an optimal number of 9 clusters in order to detect sulcus communities following similar opening trajectory during aging. Moving on to the older, at-risk cohort (Memento), we found that sulci with most stable modelling yield higher correlation with long-term memory test. In fact, we demonstrated a strong correlation (r=0.47, p-value=0.00012) between the length of the confidence interval of the tipping-point age during bootstrapping and the correlation coefficient of the sulcal opening with the free recall test score. In other words, the sulci that are most relevant for the long-term memory test are those which effectively show a neat acceleration point in atrophy. Furthermore, the predictive power on the free recall test scores of the at-risk Memento subjects varies with the sulcus community, as well as correlation with the memory test (Fig 2b). This result suggests that the specific opening trajectory of the community with the highest predictive power are induced by early neurodegenerative processes occurring in the general UK Biobank cohort