Single breath-hold MRCP at 3T using a 3D T2-weighted TSE (3D SPACE) sequence for zero-shot learning reconstruction

Description This dataset is the sample data for ZS_BH-MRCP (GitHub). Place this dataset in the "sample_data/" after clone the repository. Data The provided a single breath-hold MRCP data was acquired at 3T (Lumina, Siemens Healthineers AG, Erlangen) using the 3D-SPACE (3D T2w TSE) sequence for a single healthy volunteer. The 3D MRCP data was decoupled along the fully-sampled readout direction using Fourier transform. The data in the HDF5 format contains the following structures: Datasets kdata: Acquired k-space data decoubled along the fully-sampled readout direction using Fourier transform (nCoil x Phase Encoding x Readout (decoupled) x Partition Encoding) sm_espirit: ESPIRiT-based sensitivity maps (nCoil x Phase Encoding x Readout (decoupled) x Partition Encoding) cs: L1-wavelet compressed sensing reconstruction (Oversampled-Cropped Phase Encoding x Readout x Partition Encoding) Attribute recon_crop_size: The size of the reconstructed image cropping over-sampled in the phase encoding direction Citation This work has been submitted for publication and is currently under revision. Until the peer-reviewed version is available, please cite the preprint on arXiv (Kim et al., 2025). Kim, J., Nickel, M. D., & Knoll, F. (2025). Zero‑shot self‑supervised learning of single breath‑hold magnetic resonance cholangiopancreatography (MRCP) reconstruction. arXiv. https://doi.org/10.48550/arXiv.2508.09200