Computer-aided diagnosis (CADx) systems have been proposed to overcome the limitations of the radiological reading of multiparametric MRI. Fully automated segmentation of the prostate is a crucial step of CADx systems, which can be successfully performed by atlas-based segmentation of T2-weighted (T2W) MR images. For applications like treatment monitoring and active surveillance, the repeatability of automated segmentation method is highly important. In this work, we investigated the repeatability of several fully automated atlas-based prostate segmentation methods. We found that the repeatability of the investigated methods is excellent, which is promising for the further development of CADx systems following patients with multiple measurements over time.