An AI Approach to Computer Assisted Tomography Extended Abstract

Computer assisted tomography (CAT) systems demand large amounts of time and space. In this paper, we describe an approach to solving the CAT problem using several AI techniques including representation selection and pruning. Rather than backproject intensity data onto a pixel grid or voxel lattice, we solve the problem geometrically — a backprojection region is stored as a convex polygon in a planar decomposition. An algorithm for intersecting two planar decompositions is described below. This algorithm allows us to merge intensity information from multiple sources into a single coherent structure. Knowledge of the task can provide pruning information in the form of stabilized regions that need not to be decomposed.