Limits...
Parameter-free binarization and skeletonization of fiber networks from confocal image stacks.

Krauss P, Metzner C, Lange J, Lang N, Fabry B - PLoS ONE (2012)

Bottom Line: The size and intensity pattern of the template is automatically adapted to the input data so that the method is scale invariant and generic.Furthermore, the template matching threshold is iteratively optimized to ensure that the final skeletonized network obeys a universal property of voxelized random line networks, namely, solid-phase voxels have most likely three solid-phase neighbors in a 3 x 3 x 3 neighborhood.This optimization criterion makes our method free of user-defined parameters and the output exceptionally robust against imaging noise.

View Article: PubMed Central - PubMed

Affiliation: Department of Physics, Biophysics Group, Friedrich-Alexander University, Erlangen, Germany.

ABSTRACT
We present a method to reconstruct a disordered network of thin biopolymers, such as collagen gels, from three-dimensional (3D) image stacks recorded with a confocal microscope. The method is based on a template matching algorithm that simultaneously performs a binarization and skeletonization of the network. The size and intensity pattern of the template is automatically adapted to the input data so that the method is scale invariant and generic. Furthermore, the template matching threshold is iteratively optimized to ensure that the final skeletonized network obeys a universal property of voxelized random line networks, namely, solid-phase voxels have most likely three solid-phase neighbors in a 3 x 3 x 3 neighborhood. This optimization criterion makes our method free of user-defined parameters and the output exceptionally robust against imaging noise.

Show MeSH

Related in: MedlinePlus

Statistical properties of a real and a surrogate image stacks.(A) Comparison of the voxel intensity distributions in the real and surrogate image stacks. Both distributions are similar. (B) and (C) show angular distributions of the fiber segments. (B) Typical distributions of azimuthal angles  in a real and a surrogate data set. The distributions are almost indistinguishable. The peaks are a result of voxelization. The principal directions, corresponding to the x- and y-direction, as well as the principal diagonals are over-represented in short fiber segments and lead to maxima at  (C) Typical distributions of polar angles  in a real and a surrogate data set. Again, the distributions are similar. Compared to an ideal isotropic network with , polar angles smaller than  are increasingly suppressed due to the blind spot effect of confocal reflection microscopy [15].
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3351466&req=5

pone-0036575-g002: Statistical properties of a real and a surrogate image stacks.(A) Comparison of the voxel intensity distributions in the real and surrogate image stacks. Both distributions are similar. (B) and (C) show angular distributions of the fiber segments. (B) Typical distributions of azimuthal angles in a real and a surrogate data set. The distributions are almost indistinguishable. The peaks are a result of voxelization. The principal directions, corresponding to the x- and y-direction, as well as the principal diagonals are over-represented in short fiber segments and lead to maxima at (C) Typical distributions of polar angles in a real and a surrogate data set. Again, the distributions are similar. Compared to an ideal isotropic network with , polar angles smaller than are increasingly suppressed due to the blind spot effect of confocal reflection microscopy [15].

Mentions: We start with an image stack recorded by confocal reflection microscopy. Let us assume that the intensities of the image stack are coded with 8 bits, i.e. all brightness values B are in the range [0,255], with B = 0 corresponding to completely dark (black) and B = 255 to maximum bright (white) voxels. In our setup (Leica SP5X confocal microscope in reflection mode), a typical distribution p(B) of brightness values has a sharp peak around B = 15±5 and a flat tail towards large values (Fig. 2A). The reasonable range of binarization thresholds is located somewhere within this tail. However, the distribution p(B) itself offers no hint for choosing the optimum threshold.


Parameter-free binarization and skeletonization of fiber networks from confocal image stacks.

Krauss P, Metzner C, Lange J, Lang N, Fabry B - PLoS ONE (2012)

Statistical properties of a real and a surrogate image stacks.(A) Comparison of the voxel intensity distributions in the real and surrogate image stacks. Both distributions are similar. (B) and (C) show angular distributions of the fiber segments. (B) Typical distributions of azimuthal angles  in a real and a surrogate data set. The distributions are almost indistinguishable. The peaks are a result of voxelization. The principal directions, corresponding to the x- and y-direction, as well as the principal diagonals are over-represented in short fiber segments and lead to maxima at  (C) Typical distributions of polar angles  in a real and a surrogate data set. Again, the distributions are similar. Compared to an ideal isotropic network with , polar angles smaller than  are increasingly suppressed due to the blind spot effect of confocal reflection microscopy [15].
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3351466&req=5

pone-0036575-g002: Statistical properties of a real and a surrogate image stacks.(A) Comparison of the voxel intensity distributions in the real and surrogate image stacks. Both distributions are similar. (B) and (C) show angular distributions of the fiber segments. (B) Typical distributions of azimuthal angles in a real and a surrogate data set. The distributions are almost indistinguishable. The peaks are a result of voxelization. The principal directions, corresponding to the x- and y-direction, as well as the principal diagonals are over-represented in short fiber segments and lead to maxima at (C) Typical distributions of polar angles in a real and a surrogate data set. Again, the distributions are similar. Compared to an ideal isotropic network with , polar angles smaller than are increasingly suppressed due to the blind spot effect of confocal reflection microscopy [15].
Mentions: We start with an image stack recorded by confocal reflection microscopy. Let us assume that the intensities of the image stack are coded with 8 bits, i.e. all brightness values B are in the range [0,255], with B = 0 corresponding to completely dark (black) and B = 255 to maximum bright (white) voxels. In our setup (Leica SP5X confocal microscope in reflection mode), a typical distribution p(B) of brightness values has a sharp peak around B = 15±5 and a flat tail towards large values (Fig. 2A). The reasonable range of binarization thresholds is located somewhere within this tail. However, the distribution p(B) itself offers no hint for choosing the optimum threshold.

Bottom Line: The size and intensity pattern of the template is automatically adapted to the input data so that the method is scale invariant and generic.Furthermore, the template matching threshold is iteratively optimized to ensure that the final skeletonized network obeys a universal property of voxelized random line networks, namely, solid-phase voxels have most likely three solid-phase neighbors in a 3 x 3 x 3 neighborhood.This optimization criterion makes our method free of user-defined parameters and the output exceptionally robust against imaging noise.

View Article: PubMed Central - PubMed

Affiliation: Department of Physics, Biophysics Group, Friedrich-Alexander University, Erlangen, Germany.

ABSTRACT
We present a method to reconstruct a disordered network of thin biopolymers, such as collagen gels, from three-dimensional (3D) image stacks recorded with a confocal microscope. The method is based on a template matching algorithm that simultaneously performs a binarization and skeletonization of the network. The size and intensity pattern of the template is automatically adapted to the input data so that the method is scale invariant and generic. Furthermore, the template matching threshold is iteratively optimized to ensure that the final skeletonized network obeys a universal property of voxelized random line networks, namely, solid-phase voxels have most likely three solid-phase neighbors in a 3 x 3 x 3 neighborhood. This optimization criterion makes our method free of user-defined parameters and the output exceptionally robust against imaging noise.

Show MeSH
Related in: MedlinePlus