Zaccharie Ramzi, Philippe Ciuciu, Jean-Luc Starck
We present a new neural network, the XPDNet, for MRI reconstruction from periodically under-sampled multi-coil data. We inform the design of this network by taking best practices from MRI reconstruction and computer vision. We show that this network can achieve state-of-the-art reconstruction results, as shown by its ranking of second in the fastMRI 2020 challenge.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Image Reconstruction | fastMRI Knee 4x | PSNR | 40.2 | XPDNet |
| Image Reconstruction | fastMRI Knee 4x | SSIM | 0.9287 | XPDNet |
| Image Reconstruction | fastMRI Knee 8x | PSNR | 37.2 | XPDNet |
| Image Reconstruction | fastMRI Knee 8x | SSIM | 0.8893 | XPDNet |
| Image Reconstruction | fastMRI Brain 4x | PSNR | 41.3 | XPDNet |
| Image Reconstruction | fastMRI Brain 4x | SSIM | 0.9581 | XPDNet |
| Image Reconstruction | fastMRI Brain 8x | PSNR | 38.1 | XPDNet |
| Image Reconstruction | fastMRI Brain 8x | SSIM | 0.9408 | XPDNet |
| MRI Reconstruction | fastMRI Knee 4x | PSNR | 40.2 | XPDNet |
| MRI Reconstruction | fastMRI Knee 4x | SSIM | 0.9287 | XPDNet |
| MRI Reconstruction | fastMRI Knee 8x | PSNR | 37.2 | XPDNet |
| MRI Reconstruction | fastMRI Knee 8x | SSIM | 0.8893 | XPDNet |
| MRI Reconstruction | fastMRI Brain 4x | PSNR | 41.3 | XPDNet |
| MRI Reconstruction | fastMRI Brain 4x | SSIM | 0.9581 | XPDNet |
| MRI Reconstruction | fastMRI Brain 8x | PSNR | 38.1 | XPDNet |
| MRI Reconstruction | fastMRI Brain 8x | SSIM | 0.9408 | XPDNet |