ABSTRACT

Sex differences in brain anatomy have been described from early childhood through late adulthood, but without any clear consensus among studies. Here, we applied a machine learning approach to estimate ‘Brain Sex’ using a continuous (rather than binary) classifier in 162 boys and 185 girls aged between 5 and 18 years. Changes in the estimated sex differences over time at different age groups were subsequently calculated using a sliding window approach. We hypothesized that males and females would differ in brain structure already during childhood, but that these differences will become even more pronounced with increasing age, particularly during adolescence. Overall, the classifier achieved a good performance, with an accuracy of 80.4% and an AUC of 0.897 across all age groups. Assessing changes in the estimated sex with age revealed a growing difference between the sexes with increasing age. That is, the very large effect size of d = 1.2 which was already evident during childhood increased even further from age 11 onward, and eventually reached an effect size of d = 1.6 at age 17. Altogether these findings suggest a systematic sex difference in brain structure already during childhood, and a subsequent increase of this difference during adolescence.

 抽象的

从幼儿期到成年晚期，大脑解剖学上的性别差异已经被描述过，但研究之间没有任何明确的共识。在这里，我们应用机器学习方法，使用连续（而不是二元）分类器对 162 名男孩和 185 名年龄在 5 至 18 岁之间的女孩进行“大脑性别”估计。随后使用滑动窗口方法计算不同年龄组的估计性别差异随时间的变化。我们假设男性和女性在童年时期就已经在大脑结构上存在差异，但随着年龄的增长，这些差异将变得更加明显，特别是在青春期。总体而言，该分类器取得了良好的性能，所有年龄段的准确率为 80.4%，AUC 为 0.897。评估估计性别随年龄的变化发现，随着年龄的增长，两性之间的差异越来越大。也就是说，在儿童时期就已经很明显的 d = 1.2 的巨大效应量从 11 岁起进一步增加，并最终在 17 岁时达到 d = 1.6 的效应量。总而言之，这些发现表明大脑中存在系统性的性别差异这种差异在童年时期就已经存在，并且在青春期期间这种差异逐渐增加。