Abstract: There is increasing concern about statistical power in neuroimaging research. Critically, an underpowered study has poor predictive power. That is, findings from a low powered study are unlikely to be reproducible. Optimal experimental designs and sufficiently large sample sizes are key ingredients to obtain a well-powered study. During this talk, I will present our recently developed statistical methods for sample size calculations and study design optimization. First, I will present a direct way to estimate power based on an existing pilot study, thus allowing to minimize the cost of an fMRI experiment, while attaining a given level of power. The procedure estimates the proportion of active peaks (π1) and the distribution of effects, which is a function of effect size δ and sample size n. Using an evaluation procedure based on simulations and data from the Human Connectome Project, I will show that this method predicts well the effect size and the power curves. Second, I will show recently developed tools to optimize the detection power within subjects. To this end, we use the genetic algorithm to optimize the experimental design, extended to allow complex experimental setups and a mix between blocked and event-related designs. Both methods will be available in a user-friendly web based toolbox available at www.neuropowertools.org.