A public manifold left/right ventricular database

A public manifold left/right ventricular database This database contains the 44 postprocessed meshes based on the public databases 10.5281/zenodo.4590294 and 10.5281/zenodo.3890034. This database was used to train the DeepSDF of our work Shape of my heart: Cardiac models through learned implicitly defined functions 10.48550/arXiv.2308.16568. Further information on the postprocessing can be found in this article. Structure The database is split into the closed surfaces for each chamber between endo- and epicardium. lv endo epi rv endo epi The metadata files metadata.pkl.gz and metadata.csv contain a path to all meshes and their respective sources. Here's an example how the files can be read: import pandas as pd metadata = pd.read_pickle("metadata.pkl.gz") #or metadata = pd.read_csv("metadata.csv", sep=";") The file point_clouds.pkl.gz contains a DataFrame of the point cloud data that was used for generated for all 44 meshes and used in the validation and training. The data can be read similar to the metadata: import pandas as pd point_clouds = pd.read_pickle("metadata.pkl.gz") The data contains the following fields (SDF = signed distance function, LV = left ventricular, RV = right ventricular) mesh_id xyz sdf_lv_endo sdf_lv_epi sdf_rv_endo sdf_rv_epi ID of the mesh from which the point cloud was generated Spatial coordinate SDF to the LV endocardium SDF to the LV epicardium SDF to the RV endocardium SDF to the RV epicardium All provided meshes contain the following fields: The fiber and sheet orientation of the original meshes, computed using a rule-based approach. fiber sheet A true/false, (0/1) mask marking the valves of the meshes. valve The universal ventricular coordinates from the original meshes to provide orientation inside the ventricles. See 10.1016/j.media.2018.01.005 for more details. uvc_intraventricular uvc_longitudinal uvc_rotational uvc_transmural Acknowledgements This work was supported by the Swiss National Science Fund [Cardiotwin, Weave/Lead Agency, Project number 214817], by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) [EXC-2047/1-390685813, and EXC2151-390873048], PRIN-PNRR [project no. P2022N5ZNP], and CSCS [production grant no. s1074].