In this study, we developed a super resolution deep learning algorithm to enhance spatial resolution of 4D Flow MRI. The neural network was trained using synthetic 4D Flow MRI dataset generated from aortic CFD simulations.
The network was then tested on actual 4D Flow MRI of a phantom and volunteer data.
You can download the pre-trained 4DFlowNet on this site.
The zip file contains the TF1.8 version. For TF2.0 with Keras please use the HDF5 file pre-trained weights.