Diffusion tensor imaging (DTI) and High angular resolution diffusion imaging (HARDI) allow to study axonal fibers in a living brain. Despite being very popular, they produce datasets of imaging signals with notoriously high noise. The goal of tractography is to estimate the geometry of axonal fibers based on these datasets. In this talk, we will look at a particular methodology called statistical tractography, that allows to trace the axonal fibers together with the surrounding confidence ellipsoids. This method quantifies the uncertainty in the fiber locations as it propagates from imaging signals level. It enhances existing brain images that can assist in disease diagnostics. We will discuss the different underlying models that connect the observed imaging signals and the axonal fibers, the asymptotical properties of the estimated fibers and of some test statistics. Extensions to longitudinal studies, growing numbers of gradient directions and some open problems will be discussed if time permits.