Regional heterogeneity in left atrial stiffness impacts passive deformation in a cohort of patient-specific models

We present the first database of left atrial models with patient-specific wall thickness, suitable for electro-mechanical simulations. Our database consists of ten left atrial models generated from end-diastolic CT acquired from heart failure patients recruited for cardiac resynchronization therapy upgrade. We also provide models of the epicardial adipose tissue (EAT) surrounding each left atrium, also segmented from end-diastolic CT. The initial left atrial segmentation obtained at end-diastole was upsampled and smoothed. The final multi-label segmentation was used to generate a tetrahedral mesh. The resulting meshes had an average edge length of 0.6 mm. The elements of all ten meshes are labelled as follows: 1) Left atrial myocardium 2) Mitral valve ring 3) Mitral valve plane 4) Right superior pulmonary vein ring 5) Right superior pulmonary vein valve plane 6) Right inferior pulmonary vein ring 7) Right inferior pulmonary vein valve plane 8) Left superior pulmonary vein ring 9) Left superior pulmonary vein valve plane 10) Left inferior pulmonary vein ring 11) Left inferior pulmonary vein valve plane 12) Left atrial appendage ring 13) Left atrial appendage valve plane.Left atrial fibres were generated using a rule-based method based on histological data. We defined a system of coordinates on the meshes: a longitudinal coordinate varying continuously from 0 at the atrial roof to 1 at the mitral valve; a transmural coordinate varying continuously from 0 at the endocardium to 1 at the epicardium; a rotational coordinate varying continuously from – π at the lateral wall, 0 at the septum and then back to + π at the lateral wall. This coordinate system was assigned to the left atrial myocardium in the meshes and all the other labels were assigned with -100. Universal atrial coordinates were used to divide the left atrium in 5 regions: anterior, posterior, septum, lateral and roof. The region labels used are defined on the left atrial mesh elements where a label of 1 refers to the anterior region, 2 to the posterior region, 3 to the septum, 4 to the lateral region and 5 to the roof.The EAT meshes are also tetrahedral meshes with an average edge length of 0.6 mm. All the elements of the EAT mesh are labelled as 20. This database also contains all training datasets for the Gaussian processes emulators (GPEs) trained in the study entitled ” Regional heterogeneity in left atrial stiffness impacts passive deformation in a cohort of patient-specific models”, submitted to PLOS Computational Biology.Every simulation_data folder contains subfolders for each wave of history matching carried out in this study. For each wave, the subfolder contains: -X.txt: the rows are the samples, and the columns represent the parameters that were varied in the analysis-Y.txt: the rows are the samples, and the columns represent the values for the output features simulated for each sample-x_labels.txt: contains the labels of the parameters varied in the analysis-y_labels.txt: contains the labels of the output features examined in the analysis The observed_data folders contain the values of the image-derived output features for each case.-exp_mean: contains the values of the output features estimated using image registration.-exp_std: contains an estimate of the uncertainty attached to the image-based output features, derived from feature tracking.-y_labels.txt: contains the labels of the output features examined in the analysis For calibration, we used Markov Chain Monte Carlo (MCMC) after the final wave of history matching in each case. The MCMC samples for each case are provided as .h5 files. The target output values used for MCMC can be found within the observed_data foldersThe details about the simulation model can be found in the original publication.