BrainWeb Training Data, Pre-Trained Model and Cartesian Brain Ultra-Low Field MR Data for the Spatially-Adaptive TV Example in MRpro

This dataset contains the following: A set of 2500 images (brainweb_data.pt) generated from the BrainWeb dataset [Aubert-Broche et al, IEEE TMI 2006] using an inversion recovery signal model from MRpro. A set of pre-trained weights (tv_model.pt) of a network to estimate spatially adaptive regularization parameter weights to be used in the total variation (TV)-minmization example in MRpro. The network was trained on 2250 of the images above, using 100 images for validation and leaving 150 for testing purposes. A small subset (training_images_part.pt, validation_images_part.pt, test_images_part.pt - 16, 8 and 10 images, respectively) of randomly selected images that were used for training, validation and testing of the network mentioned above. These are the images that are used in the notebook example of MRpro in order to avoid to download all 2500 images. A brain MR scan of a healthy volunteer obtained with an OSI2 ONE ultra-low field scanner (https://www.opensourceimaging.org/project/osii-one). For further details, please have a look at the examples of MRpro (https://github.com/PTB-MR/mrpro).