Three-Dimensional Thinning Algorithm that Peels the Outmost Layer with Application to Neuron Tracing

A novel thinning algorithm for three-dimensional (3D) binary images is presented, with applications in the study of neuronal micro-anatomy by light microscopy. This algorithm satisfies properties important to many biological applications, including (a) connectivity preservation, (b) thinness, and (c) geometry preservation. It is fast to execute (a few minutes for typical 3D data sizes) on a personal computer. The algorithm addresses many challenges that are presented by 3D data. Algorithm improvements, over precursory algorithms, include the following: (1) Stricter and more exhaustive constraints on identifying outmost-layer border points are applied. (2) Border points are deleted by a novel ascending order of weighted neighbor count approach. The algorithm is robust in that it retains the above three properties (a–c) in the presence of relatively severe noise, uneven dye uptake, and nonuniform background.