We introduce inherent measures for effective quality control of brain\nsegmentation based on a Bayesian fully convolutional neural network, using\nmodel uncertainty. Monte Carlo samples from the posterior distribution are\nefficiently generated using dropout at test time. Based on these samples, we\nintroduce next to a voxel-wise uncertainty map also three metrics for\nstructure-wise uncertainty. We then incorporate these structure-wise\nuncertainty in group analyses as a measure of confidence in the observation.\nOur results show that the metrics are highly correlated to segmentation\naccuracy and therefore present an inherent measure of segmentation quality.\nFurthermore, group analysis with uncertainty results in effect sizes closer to\nthat of manual annotations. The introduced uncertainty metrics can not only be\nvery useful in translation to clinical practice but also provide automated\nquality control and group analyses in processing large data repositories.\n