Raw cardiac MRI datasets from the paper titled "Accelerating whole-heart 3D T2 mapping: impact of undersampling strategies and reconstruction techniques", Zhu D et al., published in PLOS One, 2021.Fully sampled, multi-coil, multi contrast 3D cardiac MRI data with whole-heart coverage for N=3 swine and N=8 normal human subjects is included in the .mat (MATLAB) files.Each .mat file contains one dataset corresponding to either swine (rawMR_kdata_swineX.mat) or human (rawMR_kdata_HumanX.mat) raw complex multicoil k-space data contained within a 5-dimensional array (type: complex singles, named to match filename). The data was acquired using the technique described by Ding et al. (Magn Reson Med 2015 Sep;74(3):803-16. doi: 10.1002/mrm.25458). This dataset was used in the publication by Zhu et al. (PLOS One 2021, in press) in its entirety.The first dimension of the 5-dimensional array indexes individual readout points, and is fractionally sampled. The second and third dimensions index into phase and slice encoding, respectively, and are fully sampled. The fourth dimensions indexes in to volume/T2 contrast, varying with data varying in degree of T2-weighting as achieved by T2-preparation (TEs: 0, 25,35,45 msec respectively). The final dimension indexes into coil/channel with each dataset typically having been acquired with a 32-channel cardiac array. A second variable ( type: cell array of string, name: dims) summarizes the what data is contained along each dimension. Fourier transformation along the first three dimensions (with appropriate shifting) should yield images, and succeeding root-sum-squares combination along the coil/channel dimension should produce the final 3-dimensional cardiac MRI data with changing contrast.

Date Submitted: 2021-07-17