We have selected a parallelepiped around the heart made by 450*450 pixel rectangles cropped from 100 slices. Each cropped slice underwent edge-detection and we built a volumetric distribution image of the heart volume. Superimposition of 3D grid to the constructed volumetric image of the heart allows construction of the heart model; each grid point may be "occupied", if any cardiac tissue "touches" that point or "free" otherwise. Each occupied grid point correspond to that we called node. In order to attribute a cardiac-tissue-type to each node, we have associated a different color to different regions. To keep point equidistant (1mm=3*0.33 mm) we sampled these images (Temporal sample factor = 3). Grid dimension became 150*150*100.

Original Image	Edge Detected Image
Cardiac Tissue Type Image (450*450 pixel)	Cardiac Tissue Type Image (150*150 pixel)

Legenda: ---------Ventricle ---------A/V node-------- -Atria

During these operations of assignment of cardiac-tissue-types we have found several difficulties. VHD represents sections of the body in the horizontal plane; heart leans to the left side of the body. Doctors are usually thinking the heart in its plane. Furthermore photographs contains some substances, such gel, that sometimes overlap heart parts, and made the recognize process more complex. We need also to recognize microscopic parts, such His bundle and Purkinje fibres, that are not visible in the VHD. We have developed an ad-hoc editor to override this problem. Each slice may be edited and an assignment of cardiac-tissue-type is possible by selecting a type and clicking in correspondence of the position of the node.

Example of an edited slice: