Subspace DWI datasets

This dataset contains raw diffusion-weighted MRI data acquired ex vivo on mouse brains and can be used to validate the subspace reconstruction pipeline available here: https://github.com/CRMSB/PAPER_subspace_DWI. The data were acquired using a custom 3D diffusion spin-echo sequence with prospective undersampling. A fully sampled central region of k-space was preserved, while the outer k-space was sampled using a variable-density Poisson distribution along the \(k_y-k_z\) directions. Independent undersampling masks were generated for the \(b_0\) images and diffusion-weighted directions. The acquisition includes 5 \(b_0\) images and 30 diffusion directions with a diffusion weighting of \(b = 3000\ \mathrm{s/mm^2}\). The nominal isotropic spatial resolution is approximately \(93\ \mu\mathrm{m}\). Two acquisitions are provided in the dataset: * **Dataset 10** contains the fully sampled reference acquisition.* **Dataset 15** contains a prospectively undersampled acquisition with an acceleration factor of \(AF = 12\). These data can therefore be used to compare accelerated reconstructions against the corresponding fully sampled reference and to evaluate reconstruction strategies such as wavelet, locally low-rank, and subspace-based approaches. The acquisition and reconstruction framework were developed at the Centre de Résonance Magnétique des Systémes Biologiques (CRMSB), UMR 5536, CNRS / University of Bordeaux, Bordeaux, France.