We propose a framework for accelerated reconstruction of 2D phase contrast MRI from undersampled K space by using deep convolutional neural networks. The reconstruction problem is considered as a de-aliasing problem in complex spatial domain. A U-net architecture was trained and tested on 4D flow MRI data in 10 patients with aortic stenosis and 4 healthy volunteers. The reconstructed complex two channel image showed that the U-net is able to unaliase the undersampled flow images with resulting magnitude and phase difference images showing good agreement with the fully sampled magnitude and phase images.