Image-to-Image Translation on Multi-Contrast MR Images

Generative Adversarial Neural Networks (GANs), UNets, and registration based methods for T1w-to-T2w image translation.

[NeuroImage paper]

Code usage

1. Prepare your dataset under the directory 'data' in the CycleGAN folder and set dataset name to parameter 'image_folder' in model init function.

• Directory structure on new dataset needed for training and testing:
  • data/Dataset-name/trainA
  • data/Dataset-name/trainB
  • data/Dataset-name/testA
  • data/Dataset-name/testB

2. Train a model by:

python CycleGAN/CycleGAN.py

3. Generate synthetic images by following specifications under:

• CycleGAN/generate_images/ReadMe.md

Result GIFs - 304x256 pixel images

Left: Input image. Middle: Synthetic images generated during training. Right: Ground truth.
Histograms show pixel value distributions for synthetic images (blue) compared to ground truth (brown).

CycleGAN - T1 to T2

CycleGAN - T2 to T1