README: tensor-denoising-2DCSI This contains code for denoising of single-timepoint 2D CSI data using Tucker decomposition. Thus far, the datasets that this code uses are Varian .fid files, which are read in using supplementary MATLAB functions (see individual files - they are listed in the header of each file. You can likely find them in agentmess's hyperpolarized-MRI-toolbox). Additional notes: - The 13C imaging data matrix size is 128x8x8 (spectralxspatialxspatial) and is of hyperpolarized bicarbonate (161 ppm) and CO2 (125.5 ppm) for pH imaging - You will likely need to change some of the variables (listed in VARIABLES TO CHANGE section) pertaining in particular to data paths. As far as data files go, I don't have access to them anymore (since I don't work at UCSF any longer) Brief summary of files and directories: Within /Simulation_with_phantom_data: CSI3xTucker.m This is the main function that I would suggest working with. This adds white Gaussian noise to a specified dataset and then denoises with the specified settings multiple times. You can also run tests to identify optimal denoising parameters. It performs a Tucker decomposition and denoising 3 different times along the processing pipeline. CSImultiTucker.m Similar to the file above, but gradually apodizes bit by bit and denoises along the way. It seems to work just about the same. Within /Mouse_data_processing: mouseproc_TVD: I believe this is the main file to work with. This does a 2x or 3x Tucker decomposition on mouse data and then analyzes the data with user-specified ROIs as well. ROIs can also be loaded from .mat files (again, I'd have to direct you to them). mouseproc_TVDmulti: I wouldn't bother with this one. This is analogous to the multi-Tucker and doesn't seem to work well. Any questions, please contact me. Thanks! Dave K