3D Image Segmentation

6. 3D Image Segmentation

Image segmentation is a way to electronically dissect significant biological components from a 3D map of a macromolecule, and thereby obtain a clearer view into the macromolecules architectural organization. For instance, it is often helpful to segment an asymmetric local subunit (structure unit) out of an icosahedral virus such that further structural interpretation can be conducted only on the structure unit instead of the entire map without loss of any structural information [5]. Automated segmentation is still recognized as one of the challenge problems in image processing. Commonly used semi-automatic methods include segmentation based on edge detection, region growing and/or region merging, active curve/surface motion and model based segmentation (see for example [1,2]). Publications [3,4] present steps towards an automatic approach for asymmetric subunit detection and segmentation of 3D maps of icosahedral viruses. The approach is a multi-seeded, multi-object boundary tracking variant of the well-known fast marching method of Malladi, Sethian. The traditional fast marching method are designed for a single object boundary segmentation. In order to segment multiple targets, such as 60-component virus capsids or a 3-component molecular trimeric subunit, one has to choose a seed for each of the components. However, assigning only one seed to each component may cause appropriate boundary detection problems, as demonstrated in [4], and hence a multi-seeded initialization scheme becomes necessary. There are several steps to achieve automatic structure unit segmentation for icosahedral virus maps: (i) Decide the global (icosahedral) symmetry; (ii) Segment the capsid layers; (iii) Detect the local symmetries; (iv) Segment the subunits based on the global/local symmetries; (v) If needed, segment the monomers from the segmented subunits (trimers, pentons, hexons, etc.). The following pictures show the segmentation results of the Rice Dwarf Virus (RDV) and the Bacteriophage (P22).

Figure 1

Segmentation of rice dwarf virus (RDV) at resolution 6.8Å.

RDV 1 RDV 2 RDV 3
Original map. Segmented outer capsid layer. Symmetry axes (both global icosahedra symmetry and local 6-fold and 3-fold symmetries).
RDV global RDV local RDV trimer segmentRDV monomer segment
Segmented asymmetric units
(60 copies in total).
Each subunit consists of trimers.
Segmentation into trimers (260 in total). There are five types of trimers colored differently. One of the segmented trimers and the segmentation into monomers.

Figure 2

Segmentation of Bacteriophage P22 after the capsid maturation at resolution 9.5Å.

P22 Original P22 capsid P22 symmetry
Original map. Segmented outer capsid layer. Symmetry axes (both global icosahedra symmetry and local 6-fold symmetries).
P22 global P22 global local out P2 segment global local in
Segmented asymmetric units (60 copies in total). Each subunit consists of one hexon and penton. Since no local symmetry is enforced, the hexon does not appear to have exactly 6-fold symmetry. Segmentation into pentons (12 in total and colored in red) and hexons (60 in total and colored randomly). Segmentation into pentons and hexons (looking from inside).
P22 FivefoldP22 Fivefold segmentP22 Fivefold monomer P22 SixfoldP22 Sixfold segmentP22 Sixfold monomer
One of the segmented pentons (left) and the segmentation (middle) into monomers (right). One of the segmented hexons (left) and the segmentation (middle) into monomers (right).

Publications

  1. Zeyun Yu and Chandrajit Bajaj, “Image Segmentation Using Gradient Vector Diffusion and Region Merging”, Proceedings of the 16th International Conference on Pattern Recognition (ICPR’02), Vol. 2, pp. 941-944, Quebec, Canada, August 2002.
  2. Zeyun Yu and Chandrajit Bajaj, “Normalized Gradient Vector Diffusion and Image Segmentation”, Proceedings of the 7th European Conference on Computer Vision (ECCV’02), Lecture Notes in Computer Science, Vol. 2352, pp. 517-530, Copenhagen, Denmark, May 2002.
  3. Chandrajit Bajaj, Zeyun Yu, and Manfred Auer, “Volumetric Feature Extraction and Visualization of Tomographic Molecular Imaging”, Journal of Structural Biology, Vol. 144, No. 1-2, pp. 132-143, October 2003.
  4. Zeyun Yu and Chandrajit Bajaj, “Automatic Ultra-structure Segmentation of Reconstructed Cryo-EM Maps of Icosahedral Viruses”, IEEE Transactions on Image Processing: Special Issue on Molecular and Cellular Bioimaging, Vol. 14, No. 9, pp. 1324-1337, September 2005.
  5. Chandrajit Bajaj and Zeyun Yu, “Geometric Processing of Reconstructed 3D Maps of Molecular Complexes”, in Handbook of Computational Molecular Biology, Edited by S. Aluru, Chapman & Hall/CRC Press, Computer and Information Science Series, October 2005.