Quantitative analysis of microscopic images often starts with segmentation. Particularly in the biomedical field, active contours were widely for that purpose. For a successful application, they require an initial contour close to the true shape. In the 2D case, the initial contour can trivially be drawn interactively by the user. In 3D, however, initialization requires a surface, which is more intricate to draw. Available software packages are therefore commonly restricted to simple geometrical shapes. We propose a novel method for interactive segmentation of 3D images. Our software implements an automatically growing surface, originating from user-selected seed points and growing according to local gradients. The result of the growing surface provides a close initial surface for a 3D active surface model. Segmentation is optionally supported by using template surfaces. We therefore modify the original formulation of the active surface to include a rigidity force counteracting shape deformation. The whole segmentation process is performed in an interactive 3D environment. We use our software to segment confocal images of adult Drosophila brains. Software and documentation is freely available online at http://3dviewer.neurofly.de/IntSeg_3D.
interactive, segmentation, 3D
Benjamin Schmid is a post-doctoral researcher at the Max Planck Institute of Molecular Cell Biology and Genetics in Dresden. He received a degree in Bioinformatics in 2006. He then joined the Heisenberg lab at the University of Wuerzburg for a PhD, working on the segmentation and registration of confocal images of fly brains. Since 2011 he is working in the Huisken lab in Dresden where he develops software for new microscope prototypes and real-time image processing.
Presenting author: Benjamin Schmid
Organisation: Max Planck Institute of Molecular Cell Biology and Genetics, Dresden