This trial showed an obviously higher rate of large and giant aneurysm obliteration with the Tubridge FD over Enterprise stent-assisted coiling. However, this higher obliteration rate came at the cost of a nonsignificantly higher rate of complications. Investigational site comparisons suggested that a learning curve for flow-diverter implantation should be recognized and factored into trial designs.

Accurate and reliable prediction on structural deformation is a powerful means to evaluate the safety state of dams during long-term operation. This work is a contribution to the quantification of the cognitive and stochastic uncertainties in prediction models for the dam performance using measured time series, through constructing the prediction interval (PI) by an innovative model combined with the gradient boosting decision tree (GBDT) and the bootstrap method. The constructed PI, improved by the kernel density estimation (KDA), consists of an upper bound and a lower bound of the interval to provide a confidence level for dam deformation prediction. The bootstrap method combined with multiple GBDT is utilized to estimate the variance of the model bias, mainly derived from the cognitive uncertainties. The variance of random noise can be furtherly estimated through training the combined model to fit the residuals so as to indicate the stochastic uncertainties. The effectiveness of the newly proposed model is validated employing measured data of the Jinping Ι arch Dam. The results indicate that the methodology can obtain high-quality PIs and accurate prediction values and thus can provide strong support for better appraising dam performance during long-term operation.