This project utilizes cGAN with a RegNet backbone to generate contrast-enhanced MRI from a set of pre-contrast MRI sequences
The code is based on Pytorch lightning
Follow the workflow outlined below.
First use the resemble_n_get_body/get_body_t1.py to generate the mask of T1 or anything you want.
Than use get_body_all_series.py to apply the mask on every sequence.
Use reg_series/resample_t1.py to resample every T1 to a chosen T1. (ensuring consistent size and spacing)
reg_series/itk_resample_all2t1.py to resample every sequence to T1 within a patient
reg_series/kill_empty_slice.py to kill the empty slice if needed
reg_series/preprocess.py to register the sequences of each patient to ensure organ and body alignment.
Edit the training_project/ce_mri_param.py to change the configuration
Than you can run the train_main.py
Edit the inference/test_param.py and run inference/inference_2d_main.py