Click
here to close Hello! We notice that
you are using Internet Explorer, which is not supported by Echinobase
and may cause the site to display incorrectly. We suggest using a
current version of Chrome,
FireFox,
or Safari.
Integr Comp Biol
2019 Dec 01;596:1700-1712. doi: 10.1093/icb/icz117.
Show Gene links
Show Anatomy links
High-Throughput Segmentation of Tiled Biological Structures using Random-Walk Distance Transforms.
Baum D
,
Weaver JC
,
Zlotnikov I
,
Knötel D
,
Tomholt L
,
Dean MN
.
???displayArticle.abstract???
Various 3D imaging techniques are routinely used to examine biological materials, the results of which are usually a stack of grayscale images. In order to quantify structural aspects of the biological materials, however, they must first be extracted from the dataset in a process called segmentation. If the individual structures to be extracted are in contact or very close to each other, distance-based segmentation methods utilizing the Euclidean distance transform are commonly employed. Major disadvantages of the Euclidean distance transform, however, are its susceptibility to noise (very common in biological data), which often leads to incorrect segmentations (i.e., poor separation of objects of interest), and its limitation of being only effective for roundish objects. In the present work, we propose an alternative distance transform method, the random-walk distance transform, and demonstrate its effectiveness in high-throughput segmentation of three microCT datasets of biological tilings (i.e., structures composed of a large number of similar repeating units). In contrast to the Euclidean distance transform, the random-walk approach represents the global, rather than the local, geometric character of the objects to be segmented and, thus, is less susceptible to noise. In addition, it is directly applicable to structures with anisotropic shape characteristics. Using three case studies-tessellated cartilage from a stingray, the dermal endoskeleton of a starfish, and the prismatic layer of a bivalve mollusc shell-we provide a typical workflow for the segmentation of tiled structures, describe core image processing concepts that are underused in biological research, and show that for each study system, large amounts of biologically-relevant data can be rapidly segmented, visualized, and analyzed.
Fig. 1. General workflow for the segmentation of tiled biological tissues. Details are explained in the “Segmentation Workflow” section. The images show the processing of the prismatic shell layer of a bivalve mollusc (Atrina rigida). It should be noted that, while the depicted images are 2D, the data segmented are 3D (see the online version for color figure).
Fig. 2. Comparison of Euclidean distance transform and random-walk distance transform showing the susceptibility of the Euclidean distance transform to noise (bottom row). A portion of the tessellated cartilage dataset (see Fig. 3) is used as an example. Top row: distance transforms and resulting segmentations for a binary segmentation without noise. Bottom row: same series as in the top row, but for a binary segmentation with noise (note the black pores inside the foreground objects in the binary segmentation). The segmentation resulting from the Euclidean distance transform contains errors (see arrows) that cannot be overcome by altering the merge threshold. In contrast, the random-walk distance transform is able to produce a correct segmentation even when noise is present and represent the global rather than the local shape of the objects to be segmented (see the online version for color figure).
Fig. 3. Shape, size, and density quantification of stingray tesserae. (A) A microCT scan of the skeleton of a stingray (Urobatis halleri) shows the hyomandibula (B), the skeletal element investigated. Tesserae covering the surface of the skeleton (B-inset) were segmented (C) using the proposed workflow, allowing quantification of all tesserae in a high-throughput fashion in terms of (D) tesseral volume, number of neighbors/sides, and (E) average intensity value. Tesserae in D and E are color-coded according to scales at the bottom of each image; the pie chart in D shows the proportion of each tile shape in the full dataset. In D, note the under-segmented label containing two connected tesserae, marked by the white arrow in the inset; this segmenting error was easily identified as an outlier in volume and neighbor analyses, allowing targeted repair in manual proofreading (see the online version for color figure).
Fig. 4. Endoskeleton segmentation and ossicle-type color-coding in starfish. Using µCT data of the entire dermal endoskeleton of Pisaster giganteus (A), the individual ossicles can be readily identified (B). Using morphology-based classification schemes, the different ossicle classes can be segregated and their average electron density profiles (proxy for porosity and mineral density) calculated (C). The resulting ossicle groupings can then be color-coded for the entire skeletal system (D) (see the online version for color figure).
Fig. 5. Growth analysis of the prismatic ultrastructure in (A) the mineralized shell of the bivalve Atrina rigida. (B) A 3D reconstruction obtained from the segmented microtomography data of the prismatic ultrastructure. The growth direction of the shell in thickness is denoted by the z axis. (C) The radius of all segmented prisms as a function of thickness of the prismatic layer, z. The black curve represents the average prism radius in the entire prismatic tissue. The radius of each prism was calculated as prism-area-equivalent circles using R=√(A/π). (D) 2D microtomography sections obtained perpendicular to the growth direction of the prismatic layer at different thicknesses, z. (E) A 3D reconstruction obtained from the segmented microtomography data of only the shrinking prisms. (F) Rate of radius change of the shrinking prisms as a function of their relative curvature. Despite the large data spread, a linear trend is observed. In C and D, blue and orange represent growing and shrinking prisms, respectively (see the online version for color figure).
Bayerlein,
Self-similar mesostructure evolution of the growing mollusc shell reminiscent of thermodynamically driven grain growth.
2014, Pubmed
Bayerlein,
Self-similar mesostructure evolution of the growing mollusc shell reminiscent of thermodynamically driven grain growth.
2014,
Pubmed
Blowes,
Body wall structure in the starfish Asterias rubens.
2017,
Pubmed
,
Echinobase
Dean,
Ontogeny of the tessellated skeleton: insight from the skeletal growth of the round stingray Urobatis halleri.
2009,
Pubmed
Gorelick,
Shape representation and classification using the poisson equation.
2006,
Pubmed
Jayasankar,
Mechanical behavior of idealized, stingray-skeleton-inspired tiled composites as a function of geometry and material properties.
2017,
Pubmed
Jones,
3D distance fields: a survey of techniques and applications.
2006,
Pubmed
Knötel,
Automated segmentation of complex patterns in biological tissues: Lessons from stingray tessellated cartilage.
2017,
Pubmed
Lin,
A hybrid 3D watershed algorithm incorporating gradient cues and object models for automatic segmentation of nuclei in confocal image stacks.
2003,
Pubmed
Malpica,
Applying watershed algorithms to the segmentation of clustered nuclei.
1997,
Pubmed
Nudelman,
Spiers Memorial Lecture. Lessons from biomineralization: comparing the growth strategies of mollusc shell prismatic and nacreous layers in Atrina rigida.
2007,
Pubmed
Oberlaender,
Automated three-dimensional detection and counting of neuron somata.
2009,
Pubmed
Reich,
Morphological and textural evolution of the prismatic ultrastructure in mollusc shells: A comparative study of Pinnidae species.
2019,
Pubmed
Schoeppler,
Biomineralization as a Paradigm of Directional Solidification: A Physical Model for Molluscan Shell Ultrastructural Morphogenesis.
2018,
Pubmed
Seidel,
Ultrastructural and developmental features of the tessellated endoskeleton of elasmobranchs (sharks and rays).
2016,
Pubmed
