Showing papers on "3D reconstruction published in 1970"

Left Ventricle 3D Reconstruction Using GibbsRandom Fields And Simulated Annealing

Abstract: This paper proposes a left ventricle tridimensional reconstruction method from two orthogonal X-rays angiographic projections. The algorithm works under the assumption of having two segmented parallel projections and an homogeneous mixture of blood and contrast agent, in order to develop a binary reconstruction based on a Markov Random Field model and Simulated Annealing. The 3D ventricular object is considered as a stacked bidimensional slice set and each slice is reconstructed from the two one-dimensional profiles corresponding to a pair of rows obtained from the segmented projections. Each bidimensional slice is described in a polar coordinate reference system as a function n = R(& ,•) that describes each point in the slice contour. This discrete onedimensional function describing the two-dimensional slice is modeled as a non-causal Markov Random Field, where the conditional probability of one point given the rest of points is equivalent to the conditional probability of the same point given the points belonging to a neighborhood. The slice joint probability distribution is deduced by considering the equivalence between the Gibbs and Markov Random Fields. This joint probability is defined by an energy function including the local potential interaction between the sites included in a neighborhood. The energy function depends on the projections errors of the reconstructed slice, its connexity and the 3D spatial regularity. The proposed algorithm starts with a provided initial approximate reconstruction that is then appropriately deformed to obtain the most probable slice form. Such deformation process is performed by using the probabilistic Gibbs model and the Simulated Annealing in order to minimize the energy function. Performance of the reconstruction method was evaluated with preprocessed ventricular angiographic images and a 3D binary database. The results are promising as the reconstruction error is less than 7%. Transactions on Biomedicine and Health vol 3, © 1996 WIT Press, www.witpress.com, ISSN 1743-3525 240 Simulation Modelling in Bioengineering