Prostate cancer is the most common cancer among US men. However, prostate\nimaging is still challenging despite the advances in multi-parametric Magnetic\nResonance Imaging (MRI), which provides both morphologic and functional\ninformation pertaining to the pathological regions. Along with whole prostate\ngland segmentation, distinguishing between the Central Gland (CG) and\nPeripheral Zone (PZ) can guide towards differential diagnosis, since the\nfrequency and severity of tumors differ in these regions; however, their\nboundary is often weak and fuzzy. This work presents a preliminary study on\nDeep Learning to automatically delineate the CG and PZ, aiming at evaluating\nthe generalization ability of Convolutional Neural Networks (CNNs) on two\nmulti-centric MRI prostate datasets. Especially, we compared three CNN-based\narchitectures: SegNet, U-Net, and pix2pix. In such a context, the segmentation\nperformances achieved with/without pre-training were compared in 4-fold\ncross-validation. In general, U-Net outperforms the other methods, especially\nwhen training and testing are performed on multiple datasets.\n