Cylindrical approximation of a neuron from reconstructed polyhedron

Abstract

In this paper, we investigate the problem of approximating a neuron (which is a disconnected polyhedron P reconstructed from points sampled from the surface of a neuron) with minimal cylindrical segments. The problem is strongly NP-hard when we take sample points as input. We present a general algorithm which combines a method to identify critical vertices of P and useful user feedback to decompose P into desired components. For each decomposed component Q, we present an algorithm which tries to minimize the radius of the approximate enclosing cylindrical segment. Previously, this process can only be done manually by researchers in computational biology. Empirical results show that the algorithm is very efficient in practice. © Springer-Verlag Berlin Heidelberg 2004.