* Corresponding author. Tel.:#1-979-845-8870; fax:#1-979-847-8578. E-mail address: mccormick@cs.tamu.edu (B.H. McCormick). Neurocomputing 38}40 (2001) 1643}1650 Three-dimensional reconstruction of neuronal forests Brent P. Burton, Bruce H. McCormick*, Reidun Torp, James H. Fallon NVIDIA, Austin, TX 78727, USA Department of Computer Science, Scientixc Visualization Laboratory, Texas A&M University, College Station, TX 778-3112, USA University of Oslo, Norway University of California, Irvine, CA, USA Abstract Three-dimensional reconstruction of neuronal forests from a coherent volumetric data set representing Golgi-stained cortical material is demonstrated. The system automates in parallel neuron feature extraction and reconstruction, replacing largely manual techniques for tracing individual neurons. Serial sections were digitized at 10241024 pixels at a linear resolution of 0.37 m, and then digitally aligned into a coherent volume data set. Our reconstruction program, Recon,determinedregionsofinterest(ROIs)ineachimageandculleddataaggressive- ly, reducing the original volumetric data 70-fold into an ROI-based aligned image stack. The neuronal forest of Fig. 1 below was treaded from this reduced data set.  2001 Published by Elsevier Science B.V. Keywords: Brain microstructure; 3D reconstruction of neurons; Physical sectioning; Brain tissue scanning 1. Introduction Virtually, our entire knowledge of the geometry and statistics of brain microstruc- ture has been drawn from two-dimensional (planar) measurements [1]. From two dimensions, the three-dimensional environment of neurons (their somata, and de- ndritic and axonal arbors) and their interconnections can be only partially inferred. 0925-2312/01/$-see front matter  2001 Published by Elsevier Science B.V. PII:S0925-2312(01)00523-9