Contour-based segmentation and visualization of electron tomographic volumes
Electron tomography is increasingly employed to obtain highly-detailed 3-D structural information from complex biological specimens, when such information cannot be obtained otherwise. Thick sections (ca. 0.25-3 μm), which require an intermediate- or high-voltage EM, are often used so that an appreciable portion of the structure of interest can be contained within the section. Because of this relatively large amount of material, volumes made by tomographic reconstruction are often filled with a tremendous amount of detail, not all of which is of interest. Often, in the absence of special stains, the structures of interest have low contrast compared to surrounding or adjoining structures. Typically, all the structures are embedded in the "ground cytoplasm", which has nonuniformlydispersed small structural elements that have the same density and spatial-frequency range as the structures of interest. In addition, the density of a given structure may vary within an image due to non-uniform staining. In these situations, the segmentation of the volume to isolate the interesting portions of the structure presents a formidable problem.