In silico electrocardiograms of 1.8 million ventricular extrasystoles and corresponding activation maps (part 1)

1.8 million ECGs derived from multiscale simulations of cardiac electrophysiology of ventricular extrasystoles. 1000 anatomical variants of a bi-ventricular mesh x 600 excitation origins x 3 heart positions in the torso.

1.8 million ECGs derived from multiscale simulations of cardiac electrophysiology of ventricular extrasystoles. 1000 anatomical variants of a bi-ventricular mesh x 600 excitation origins x 3 heart positions in the torso.

This dataset contains about 1.8 million body surface potentials (BSPs) simulated using 1000 heart models generated using a statistical shape model.
It has been used in [1]. Here, only the noise-free BSPs are provided.

Due to its size, this is a multi-part dataset:
- Part 1: https://doi.org/10.5445/IR/1000156139
- Part 2: https://doi.org/10.5445/IR/1000156554
- Part 3: https://doi.org/10.5445/IR/1000156555
- Part 4: https://doi.org/10.5445/IR/1000156556
- Part 5: https://doi.org/10.5445/IR/1000156557

Each archive XXXX-YYYY.tar contains 20 heart models and corresponding signals. Each subdirectory within the archive contains:

- heart.vtp:
  A triangle mesh of the heart including the point data:
  - ab, rt, rtCos, rtSin, tm, tv: Consistent biventricular coordinates [2].
  - class: Boundary regions used as input for the computation of fiber orientations [3].
  - trigger: 1-based indices of the ca. 600 foci (-1000 if not a focus).

- heart_transform_matrices.mat:
  A 1 x 3 cell array containing 4 x 4 transformation matrices that, together with heart_alignment_matrix.mat (see below),
  describe the pose of the heart within the torso.
  First apply the matrix from heart_alignment_matrix.mat and then the matrix from heart_transform_matrices.mat to the nodes in heart.vtp.

- actTimes.mat:
  A numNodes x numFoci matrix of activation times computed using the fast iterative method [4,5]
  (conduction velocity in fiber direction: 1 m/s, perpendicular to fiber direction: 1/2.7 m/s).

- bsp.mat:
  - bsp: A numElectrodes x numTimeSamples x numHeartPoses x numFoci matrix of BSPs computed by aligning a transmembrane voltage template with
         scaled activation times (see actTimeScalings.mat below) and solving the second bidomain equation using the boundary element method [6].
  - bspEnd: Time index of the end of depolarization (largest scaled activation time).


The archive general.tar contains heart-model-independent data and parameters used to generate the individual heart models:

- torso.vtp:
  A triangle mesh of the torso including the point data:
  - electrodes: 1-based indices of the 200 electrodes (-1000 if not an electrode).

- heart_meanshape.vtp:
  A triangle mesh of the mean shape of the statistical shape model [7,8].  

- heart_shapemodel.mat:
  - pc: A 3*numNodes x numModes matrix of principal components (numModes = 100).
  - var: A numModes x 1 vector of variances.
  - weights: A numModes x numModels matrix of weights used to generate the 1000 heart models.

- heart_alignment_matrix.mat:
  A 4 x 4 transformation matrix describing the alignment of the mean shape with the torso-specific heart.

- heart_transform_params.mat:
  A struct containing roll, pitch, yaw angles and x, y, z translations used to generate the heart_transform_matrices.mat (see above).

- fiber_angles.mat:
  - alphaEndo: numModels x 1 vector of endocardial fiber angles used to generate fiber orientations.
  - alphaEpi: numModels x 1 vector of epicardial fiber angles used to generate fiber orientations.

- actTimeScalings.mat:
  - A numModels x numFoci matrix of factors used to scale the activation times.

- tmv_template.mat:
  The transmembrane voltage time course used to compute BSPs.

- heart_classes.vtp:
  A coarse triangle mesh of the mean shape used for fuzzy classification.

- heart_classes_subdiv.vtp:
  A subdivided version of the coarse triangle mesh of the mean shape used to convert between Cobiveco and barycentric coordinates.


[1] https://doi.org/10.48550/arXiv.2209.08095
[2] https://doi.org/10.1016/j.media.2021.102247
[3] https://github.com/KIT-IBT/LDRB_Fibers
[4] https://github.com/KIT-IBT/FIM_Eikonal
[5] https://doi.org/10.1137/120881956
[6] https://doi.org/10.1016/j.cmpb.2007.09.004
[7] https://doi.org/10.5281/zenodo.4506463
[8] https://doi.org/10.1016/j.media.2015.08.009

This dataset contains about 1.8 million body surface potentials (BSPs) simulated using 1000 heart models generated using a statistical shape model.
It has been used in [1]. Here, only the noise-free BSPs are provided.
Due to its size, this is a multi-part dataset:

Part 1: https://doi.org/10.5445/IR/1000156139
Part 2: https://doi.org/10.5445/IR/1000156554
Part 3: https://doi.org/10.5445/IR/1000156555
Part 4: https://doi.org/10.5445/IR/1000156556
Part 5: https://doi.org/10.5445/IR/1000156557

Each archive XXXX-YYYY.tar contains 20 heart models and corresponding signals. Each subdirectory within the archive contains:


heart.vtp:
A triangle mesh of the heart including the point data:

ab, rt, rtCos, rtSin, tm, tv: Consistent biventricular coordinates [2].
class: Boundary regions used as input for the computation of fiber orientations [3].
trigger: 1-based indices of the ca. 600 foci (-1000 if not a focus).



heart_transform_matrices.mat:
A 1 x 3 cell array containing 4 x 4 transformation matrices that, together with heart_alignment_matrix.mat (see below),
describe the pose of the heart within the torso.
First apply the matrix from heart_alignment_matrix.mat and then the matrix from heart_transform_matrices.mat to the nodes in heart.vtp.


actTimes.mat:
A numNodes x numFoci matrix of activation times computed using the fast iterative method [4,5]
(conduction velocity in fiber direction: 1 m/s, perpendicular to fiber direction: 1/2.7 m/s).

bsp.mat:

bsp: A numElectrodes x numTimeSamples x numHeartPoses x numFoci matrix of BSPs computed by aligning a transmembrane voltage template with
scaled activation times (see actTimeScalings.mat below) and solving the second bidomain equation using the boundary element method [6].
bspEnd: Time index of the end of depolarization (largest scaled activation time).


The archive general.tar contains heart-model-independent data and parameters used to generate the individual heart models:


torso.vtp:
A triangle mesh of the torso including the point data:

electrodes: 1-based indices of the 200 electrodes (-1000 if not an electrode).



heart_meanshape.vtp:
A triangle mesh of the mean shape of the statistical shape model [7,8].  


heart_shapemodel.mat:

pc: A 3*numNodes x numModes matrix of principal components (numModes = 100).
var: A numModes x 1 vector of variances.
weights: A numModes x numModels matrix of weights used to generate the 1000 heart models.



heart_alignment_matrix.mat:
A 4 x 4 transformation matrix describing the alignment of the mean shape with the torso-specific heart.


heart_transform_params.mat:
A struct containing roll, pitch, yaw angles and x, y, z translations used to generate the heart_transform_matrices.mat (see above).


fiber_angles.mat:

alphaEndo: numModels x 1 vector of endocardial fiber angles used to generate fiber orientations.
alphaEpi: numModels x 1 vector of epicardial fiber angles used to generate fiber orientations.



actTimeScalings.mat:

A numModels x numFoci matrix of factors used to scale the activation times.



tmv_template.mat:
The transmembrane voltage time course used to compute BSPs.


heart_classes.vtp:
A coarse triangle mesh of the mean shape used for fuzzy classification.

heart_classes_subdiv.vtp:
A subdivided version of the coarse triangle mesh of the mean shape used to convert between Cobiveco and barycentric coordinates.

[1] https://doi.org/10.48550/arXiv.2209.08095
[2] https://doi.org/10.1016/j.media.2021.102247
[3] https://github.com/KIT-IBT/LDRB_Fibers
[4] https://github.com/KIT-IBT/FIM_Eikonal
[5] https://doi.org/10.1137/120881956
[6] https://doi.org/10.1016/j.cmpb.2007.09.004
[7] https://doi.org/10.5281/zenodo.4506463
[8] https://doi.org/10.1016/j.media.2015.08.009