Synthetic data from: A tissue-informed deep learning-based method for positron range correction in 68Ga PET imaging

Description This dataset contains preclinical PET phantom data stored in HDF5 format, derived from synthetic or simulated acquisitions based on the Digimouse anatomical model [1]. It includes three tracer distributions — F18, Ga68, and MU maps — for 15 numerical phantoms. Activity distributions were manually defined and assigned to relevant anatomical regions. Using these maps, positron range effects were simulated for both 18F and 68Ga with an adapted PenEasy [2]. The resulting annihilation distributions were then processed with MCGPU-PET [3] to simulate realistic PET sinograms, which were reconstructed using the GFIRST 3D-PSF-OSEM reconstructor [4] to obtain the final synthetic PET images, based on the Inveon PET/CT preclinical scanner. The images were cutted to reduce the FOV. For more information about the workflow to generate the dataset, check the paper (Submitted): File Format DATA.h5: HDF5 file containing all dynamic PET volumes under the group Digimouse. Inside Digimouse, you will find 3 datasets: F18: shape (15, 159, 50, 50), dtype=float32 Ga68: shape (15, 159, 50, 50), dtype=float32 MU: shape (15, 159, 50, 50), dtype=float32 Each dataset contains: 15 numerical phantoms 159 axial slices 50×50 image resolution per slice. The original simulation was 154x154 image resolution, but was reduce to 50x50 to avoid excess empty space. Optional .raw In addition to the .h5 file, all individual 3D volumes are also provided as .raw files to open with programs such as Amide [ref]: Format: Binary float32, no header Shape: (z, y, x) = (159, 50, 50) Naming convention: F18_0.raw, Ga68_2.raw, etc.