Volumetric modeling of neuronal populations

Abstract

Three-dimensional digital images created from serial thin sections are being used extensively in biomedical applications. These images range from 3-D displays of computed tomography d a t a t o fully animated interactive systems. Delineation of areas of interest in these images has usually been done through hand digitization or feature-based analysis.This paper describes a method for the automatic identification of labeled neuronal cell populations, coupled with the subsequent three-dimensional reconstruction and volumetric rendering of a set of serial thin sections. This was achieved by applying morphological image processing operations to each of the serial sections. Surface fitting techniques were then developed for the neuronal cell population delineation. The coupling of volumetric rendering and interior surface fitting adds to the ease of visual interpretation of this complicated three-dimensional structure.