 |
 |
 |
 |
 |
Accelerated Phase Contrast MRI with Use of Resolution Enhancement Generative Adversarial Neural Network
The complex-difference reconstruction for inline super-resolution of phase-contrast flow (CRISPFlow) model was developed to accelerate phase contrast imaging. The model was built on two modified enhanced super-resolution generative adversarial neural networks (ESRGAN). The trained weights for each of the networks can be downloaded through the Harvard Dataverse.
Velocity compensated and encoded images acquired with low resolution are reconstructed using the vendor reconstruction algorithm pipeline. The low-resolution images are sent to an external sever via a Framework for Image Reconstruction (FIRE) interface. Network 1 is used to enhance complex-difference images, which are obtained using complex-valued subtraction of velocity compensated and encoded images. Network 2 enhances velocity compensated images directly. Both networks enhance real and imaginary parts separately. The resolution-enhanced velocity compensated and encoded images, the latter obtained through complex-valued addition, are returned to the vendor pipeline to reconstruct anatomical and phase-contrast images.