PDNet-OASIS

---

tags: - TensorFlow - MRI reconstruction - MRI datasets: - OASIS

This model can be used to reconstruct single coil OASIS data with an acceleration factor of 4.

Model description

For more details, see https://www.mdpi.com/2076-3417/10/5/1816. This section is WIP.

Intended uses and limitations

This model can be used to reconstruct single coil brain retrospective data from the OASIS database at acceleration factor 4. It cannot be used on multi-coil data.

How to use

This model can be loaded using the following repo: https://github.com/zaccharieramzi/fastmri-reproducible-benchmark. After cloning the repo, git clone https://github.com/zaccharieramzi/fastmri-reproducible-benchmark, you can install the package via pip install fastmri-reproducible-benchmark. The framework is TensorFlow.

You can initialize and load the model weights as follows:

from fastmri_recon.models.functional_models.pdnet import pdnet

model = pdnet()
model.load_weights('model_weights.h5')

Using the model is then as simple as:

model([
    kspace,  # shape: [n_slices, n_rows, n_cols, 1]
    mask,  # shape: [n_slices, n_rows, n_cols]
])

Limitations and bias

The limitations and bias of this model have not been properly investigated.

Training data

This model was trained using the OASIS dataset.

Training procedure

The training procedure is described in https://www.mdpi.com/2076-3417/10/5/1816 for brain data. This section is WIP.

Evaluation results

This model was evaluated using the OASIS dataset.

• PSNR: 33.22
• SSIM: 0.910

Bibtex entry

@article{ramzi2020benchmarking,
  title={Benchmarking MRI reconstruction neural networks on large public datasets},
  author={Ramzi, Zaccharie and Ciuciu, Philippe and Starck, Jean-Luc},
  journal={Applied Sciences},
  volume={10},
  number={5},
  pages={1816},
  year={2020},
  publisher={Multidisciplinary Digital Publishing Institute}
}