Fig. 3

Segmenation Workflow (a)-(h) and the Rotation workflow (i)-(l). (a) The experimental data from the tomography as the electron density. (b) 3D sobel filter used on the experimental data. (c) a data specific threshold is used on the 3D sobel data, so that the edges of the heart are highlighted but noise is excluded and the hearts edge can have holes which will be filled later. (d) the image size is reduced by differentiating between not connected components in the image and taking the component with the heart and then the image borders are reduced to the edges of the chosen component. (e) watershed algorithm results from data created from d and blurred with a gaussian filter. (f) heart tissue mask from e where the tissue and the ventricles are in the mask. (g) the mask from f was used on the data of b and the following steps are done to get the watershed results of the heart and the ventricles. (h) mask from the g results where tissue has value 1 and the ventricle 2, rest 0. Rotation workflow now (i) The inertia tensor is caclulated and the Eigenvalues and eigenvectors are determined. The mask is shown from h with the two eigenvectors. (j) the euler angles are determined from the eigenvectors and the mask is rotated according to the Euler angles. the rotated mask is shown with the rotated eigenvectors. (k) side view of the rotated mask. (l) again rotated mask so that the tip of the ventricle is directly below the centre of mass of the ventricle.