Reconstruction and display of curvilinear objects from optical section data using 3-D curve fitting algorithms

Biological objects resembling filaments are often highly elongated while presenting a small cross-sectional area. Examination of such objects requires acquisition of images from regions large enough to contain entire objects, but at sufficiently high resolution to resolve individual filaments. These requirements complicate the application of conventional optical sectioning and volume reconstruction techniques. For example, objective lenses used to acquire images of entire filaments or filament networks may lack sufficient depth (Z) resolution to localize filament cross-sections along the optical axis. Because volume reconstruction techniques consider only the information represented by a single volume element (voxel), views of filament networks reconstructed from images obtained at low Z-resolution will not accurately represent filament morphology. A possible solution to these problems is to simultaneously utilize all available information on the path of an object by fitting 3-D curves through data points localized in 2-D images. Here, we present an application of this approach to the reconstruction of microtubule networks from 2-D optical sections obtained using confocal microscopy, and to synthesized curves which have been distorted using a simple mathematical model of optical sectioning artefacts. Our results demonstrate that this strategy can produce high resolution 3-D views of filamentous objects from a small number of optical sections.