In this work, we present new synthetic vascular models that tries to mimic various portions of the cerebral vascular tree, as acquired from Magnetic Resonance Angiography - Time of Flight modality - acquisitions. Not only are these vascular models able to replicate the cerebral arteries, but also, the bifurcations formed by the arteries, and furthermore, one option within the models allows to embed an intracranial aneurysm. Our goal in designing this set of tools was to train convolutional neural networks for various pattern recognition tasks; namely, we intend to label the main bifurcations forming the Circle of Willis, or to automatically detect intracranial aneurysms. However, to efficiently train a neural network, the fidelity of the mimicked vascular portions is of paramount importance.