Assessment of regions at risk from coronary X-ray imaging by Kohonen's map

In previous studies, we have reported on: a) the three-dimensional reconstruction of coronary trees from bi-plane angiography, and b) the recovery of heart surface from epicardial vessels. On the ground of such works, we have developed a computational model of the spatial distribution of myocardial vascularization. The proposed model is based a on self-organizing neural network of the Kohonen's type. This allows the computation of an optimal topology-preserving map of the arterial tree onto the epicardium. The availability of such a map permits the computation of the areas at risk directly from X-ray angiography. Preliminary results from clinical data are illustrated.